Molecular phylogenetics, vocalizations, and species limits in Celeus woodpeckers (Aves: Picidae)

Molecular Phylogenetics and Evolution 61 (2011) 29–44

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution journal homepage: www.elsevier.com/locate/ympev

Molecular phylogenetics, vocalizations, and species limits in Celeus woodpeckers (Aves: Picidae) Brett W. Benz ⇑, Mark B. Robbins Department of Ecology and Evolutionary Biology and Biodiversity Institute, University of Kansas, Dyche Hall, 1345 Jayhawk Blvd., Lawrence, KS 66045, USA

a r t i c l e

i n f o

Article history: Received 12 August 2010 Revised 27 April 2011 Accepted 2 May 2011 Available online 11 May 2011 Keywords: Picidae Celeus Neotropical woodpeckers Phylogeny Species limits Vocalizations

a b s t r a c t Species limits and the evolutionary mechanisms that have shaped diversification of woodpeckers and allies (Picidae) remain obscure, as inter and intraspecific phylogenetic relationships have yet to be comprehensively resolved for most genera. Herein, we analyzed 5020 base pairs of nucleotide sequence data from the mitochondrial and nuclear genomes to reconstruct the evolutionary history of Celeus woodpeckers. Broad geographic sampling was employed to assess species limits in phenotypically variable lineages and provide a first look at the evolution of song and plumage traits in this poorly known Neotropical genus. Our results strongly support the monophyly of Celeus and reveal several novel relationships across a shallow phylogenetic topology. We confirm the close sister relationship between Celeus spectabilis and the enigmatic Celeus obrieni, both of which form a clade with Celeus flavus. The Mesoamerican Celeus castaneus was placed as sister to a Celeus undatus–grammicus lineage, with the species status of the latter drawn into question given the lack of substantial genetic, morphological, and vocal variation in these taxa. We recovered paraphyly in Celeus elegans; however, this result appears to be the consequence of mitochondrial introgression from Celeus lugubris considering the monophyly of elegans at the ß-FIBI7 locus. A second instance of paraphyly was observed in Celeus flavescens with deep genetic splits and substantial phenotypic variation indicating the presence of two distinct species in this broadly distributed lineage. As such, we advocate elevation of Celeus flavescens ochraceus to species status. Our analysis of Celeus vocalizations and plumage characters demonstrates a pattern of lability consistent with a relatively recent origin of the genus and potentially rapid speciation history. Ó 2011 Elsevier Inc. All rights reserved.

1. Introduction Recent molecular phylogenetic investigation of woodpeckers and allies (Piciformes: Picidae) has brought significant advances in understanding higher-level relationships within this diverse near-global radiation, resolving the phylogenetic position of most picid genera and confirming polyphyly in five broadly distributed clades (Webb and Moore, 2005; Benz et al., 2006; Moore et al., 2006; Fuchs et al., 2007). By comparison, intrageneric relationships and the regional evolutionary histories for much of this diversity are little known, as few groups have been examined within a modern phylogenetic context at the species level. Moreover, several instances of plumage convergence evidenced throughout the family suggest traditional phenotype-based taxonomic arrangements may not accurately reflect phylogenetic relationships among some woodpecker lineages (Weibel and Moore, 2002; Benz et al., 2006; Moore et al., 2006).

⇑ Corresponding author. Address: Biodiversity Research Center, Dyche Hall, 1345 Jayhawk Blvd., Lawrence, KS 66045-7561, USA. Fax: +1 785 864 5335. E-mail address: [email protected] (B.W. Benz). 1055-7903/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.ympev.2011.05.001

The Neotropics support by far the highest picid diversity, encompassing 102 of the 216 species currently recognized within the family, including three non-insular endemic genera Piculus, Veniliornis, and Celeus (Winkler and Christie, 2002). Among these, Celeus is the least known and most speciose, comprising 12 species restricted to Central and South America. Independent molecular data have recently confirmed that the Old World purported congeneric Rufous woodpecker (Micropternus [Celeus] brachyurus) is in fact nested within a southeast Asian clade and sister to Meiglyptes (Benz et al., 2006; Fuchs et al., 2007). The present center of Celeus diversity lies in the Amazon basin, where as many as five species may be sympatric through ecological partitioning and differing foraging strategies that specialize on a broad suite of ant and termite species. As of yet, conventional species-level phylogenetic analyses are lacking within Celeus woodpeckers, and taxonomic arrangements remain based principally on plumage characters and bill morphology, traits that exhibit a high degree of inter and intraspecific variability and are potentially homplaseous (Short, 1972, 1982). Consequently, Celeus represents a prime clade in need of detailed molecular phylogenetic investigation to clarify intrageneric relationships and resolve uncertainty surrounding the species status of several widely allopatric lineages.

30

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

In his analysis of Celeus systematic relationships, Short (1972) examined approximately 800 specimens, comparing bill morphology and an extensive suite of plumage characters to arrive at the conclusion that although Celeus castaneus exhibits significant plumage differences from Celeus elegans, Celeus lugubris, and Celeus flavescens, the shared similarities in bill morphology indicated a common evolutionary history among these taxa, and thus recognized the four taxon ‘elegans superspecies’ as a clade distinct from other members of the genus. Short’s comparative analysis further identified two distinct groups within the six subspecies of C. elegans, long-crested ‘elegans’ forms from the Guyana shield, and short-crested ‘jumana’ forms throughout Amazonia; however, explicit phylogenetic hypotheses were not made given the potential hybridization between C. elegans ‘jumana’ and the partially sympatric C. lugubris (Short, 1972). Subsequent phenotype-based taxonomic treatments of Celeus (Short, 1982; Winkler and Christie, 2002) have provided little additional insight on relationships among the remaining congenerics. In the present investigation, we employ a 5 kb molecular data set and model-based phylogenetic methods to test the position of C. castaneus within Short’s hypothesized ‘elegans superspecies’. Through broad geographic taxon sampling in C. elegans and C. flavescens, we explore the relationships among the ‘jumana’ and ‘elegans’ forms, as well as the phylogenetic position of the distinctive Celeus flavescens ochraceus. The result is a well-supported phylogenetic framework of Celeus woodpeckers that addresses the status of the enigmatic Celeus obrieni, clarifies species limits within C. flavescens, and suggests recent mitochondrial introgression between C. elegans and C. lugubris. Lastly, we provide a first look at Celeus vocalizations and plumage characters within a phylogenetic context, and highlight the need for further avenues of research in this poorly known genus. 2. Methods 2.1. Taxonomic sampling We sampled 43 ingroup specimens encompassing all currently recognized species within Celeus and at least two specimens per taxon, with the exception of C. obrieni, represented solely by the holotype (Table 1). Intraspecific genetic samples were selected from distinct geographic regions to examine genetic diversity across well-known biogeographic boundaries and evaluate species limits within broadly distributed and phenotypically variable lineages. Although limited by tissue availability, intraspecific sampling was focused within C. elegans and C. flavescens, both of which exhibit prominent geographic forms of questionable species status. Outgroup taxa were drawn from two related woodpecker genera, Dryocopus lineatus and Piculus chrysochloros, based on recent higher-level phylogenetic studies within the Picidae (Webb and Moore, 2005; Benz et al., 2006). 2.2. Sequencing protocols Whole genomic DNA was extracted from muscle tissue using proteinase K digestion under manufacturer’s protocols (DNeasy tissue kit, Qiagen). Given the relatively shallow genetic divergences within the Picidae, we selected a suite of rapid evolving mtDNA genes (NADH dehydrogenase subunits 2 and 3 [ND2, 1041 bp; ND3, 351], ATP synthase subunits 6 and 8 [ATP6, 684 bp; ATP8, 168 bp], cytochrome c oxidase subunit 3 [COXIII, 192 bp], Control Region [CR, 957 bp]), as well as two nuclear loci (intron 7 of the ß-fibrinogen gene [ß-FIBI7, 911 bp], and a segment of the nonhistone chromosomal protein HMG-17 gene including exon 2 and adjacent mRNAs [HMGN2, 693 bp]), all of which were amplified

Table 1 Summary of specimens included in this study. #

Species

Country of Origin

Source

Voucher #

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus

UNAM USNM KUNHM KUNHM ANSP USNM KUNHM AMNH ANSP LSUMNS USNM FMNH USNM FMNH FMNH LSUMNS KUNHM KUNHM FMNH AMNH AMNH FMNH FMNH FMNH AMNH ANSP ANSP

99–162 1977 5840 5738 2737 6880 1036 242714a 3224 35528 12777 389780 10473 391072 389194 4364 5764 304 63975a 278666a 242703a 191173a 344388a 208004a 242688a 3253 2477

28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

Celeus grammicus Celeus grammicus Celeus grammicus Celeus loricatus Celeus loricatus Celeus lugubris Celeus lugubris Celeus lugubris Celeus obrieni Celeus spectabilis Celeus spectabilis Celeus torquatus Celeus torquatus Celeus torquatus Celeus undatus Celeus undatus Dryocopus lineatus Piculus chrysochloros

Mexico Panama: Bocas del Toro Guyana Guyana Ecuador: Sucumbios Brazil: Pará Peru: Loreto Brazil: Bahia Ecuador: Sucumbios Brazil: Mato Grosso Guyana Brazil: Rondônia Guyana Bolivia: La Paz Brazil: Roraima Peru: Loreto Guyana Paraguay Brazil: Maranhão Brazil: Pará Brazil: Maranhão Brazil: Minas Gerais Brazil: São Paulo Brazil: Espírito Santo Brazil: Bahia Ecuador: Napo Ecuador: MoronaSantiago Bolivia: Santa Cruz Brazil: Rondônia Peru: Loreto Panama: Colon Ecuador: Esmeraldas Argentina: Corrientes Paraguay Bolivia: Santa Cruz Brazil: Piauí Peru: Madre de Díos Peru: Ucayalí Guyana Brazil: Amazonas Bolivia: Pando Guyana Guyana Peru Paraguay

LSUMNS FMNH LSUMNS LSUMNS LSUMNS USNM KUNHM LSUMNS AMNH LSUMNS LSUMNS KUNHM LSUMNS LSUMNS KUNHM KUNHM KUNHM KUNHM

105252 389782 6892 28510 11832 5899 3204 6534 242687a 45460 10664 1305 25574 9422 5829 5765 799 2966

castaneus castaneus flavus flavus flavus flavus flavus flavus elegans elegans elegans elegans elegans elegans elegans elegans elegans flavescens flavescens flavescens flavescens flavescens flavescens flavescens flavescens grammicus grammicus

Tissue sources: KUNHM, University of Kansas Natural History Museum and Biodiversity Research Center; LSUMNS, Louisiana State University Museum of Natural Science; UNAM, Museo de Zoologıa, Universidad Nacional Autónoma de México; USNM, United States National Museum of Natural History; FMNH, Field Museum of Natural History. a Museum specimens sequenced from toepad samples.

via polymerase chain reaction (PCR) in 25 ll reactions using PureTaq RTG PCR beads (GE Healthcare). Primers used for this study are summarized in Table 2, and thermocycle parameters include an initial 3 min at 94 °C, followed by 35 cycles of 20 s at 94 °C, 15 s at 53 °C, and 60 s at 72 °C, followed by a 7 min final extension at 72 °C and 4 °C soak. This protocol was modified to incorporate an annealing touch down of eight cycles at 60 °C, eight cycles at 57 °C, and 25 cycles at 55 °C for CR reactions and both nuDNA markers. ND2, ND3, and HMGN2 were sequenced for both fresh and ancient DNA samples whereas the remaining markers were only sequenced for fresh samples. All PCR products were visualized on a 1% agarose gel stained with ethidium bromide and amplicons were subsequently cleaned of unincorporated DNTPs and primers with ExoSaP-IT purification (USB Corp.) Purified PCR products were cycle sequenced with ABI Prism BigDye v3.1 terminator chemistry under manufacture’s

31

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44 Table 2 Summary of primers used in this study. Gene

Primer name

Sequence 50 –30

Reference

ND2

L5216 H6313 ND2-1H ND2-2L ND2-2H ND2-3L ND2-3H ND2-4L ND2-4H ND2-5L

GGCCCATACCCCGRAAATG CTCTTATTTAAGGCTTTGAAGGC CATTGTCCTGTGGTTCAAGC CGAGCCATYGAAGCAACAATC TGTTGATAGGAGGAGGGCGG AGCACCATTYCACTTCTGATTCC TTAGRAGYGTGAGYTTGGGGGC TAGCCTTCTCCTCCATCTCCCACC TGTGGGGGTTATTCTTGCTTGG CAAGAAATAACCCCCACAGC

Johnson and Sorenson (1998) Johnson and Sorenson (1998) This study This study This study This study This study This study This study This study

ND3

L10755 H11151

GACTTCCAATCTTTAAAATCTGG GATTTGTTGAGCCGAAATCAAC

Chesser (1999) Chesser (1999)

ATP6-8/COIII

tRNA-Lys_L COIII_RH

CAGCACTAGCCTTTTAAGCT ATTATTCCGTATCGNAGNCCYTTTTG

Sorenson et al. (1999) Sorenson et al. (1999)

Control region

tThr L16087 tPro H16137 CRINH CRINL

TGGTCTTGTAARCCAAARANYGAAG ARAATRYCAGCTTTGGGAGYTGG GTTGCTGATTTCACGTGAGG ACTTGCTCTTTTGCGCCTCTGG

Johnson and Sorenson (1998) Johnson and Sorenson (1998) This study This study

ß-fibrinogen intron 7

FIB-BI7L FIB-BI7U

TCCCCAGTAGTATCTGCCATTAGGGTT GGAGAAAACAGGACAATGACAATTCAC

Prychitko and Moore (1997) Prychitko and Moore (1997)

HMGN2

HMG17.2F HMG17.4R HMG-1R HMG-3F HMG-2R HMG-4F

GCTGAAGGAGATACCAARGGCGA CTTTGGAGCTGCCTTTTTAGG CTCAGTTCAAAGGAGTAAAATCCCAG TGGGATAGTTTCCTGCTTCTT AAGAACCACACAACAAGGC CAACGGAGGTCAGCGAGGTTATCTG

Kimball et al. (2009) Kimball et al. (2009) This study This study This study This study

thermocycling protocols using the initial PCR primers. Two internal sequencing primers (L3, H2, Table 2) were also used for ND2 in addition to the standard external primers to ensure accurate reads were obtained at the 50 –30 ends of the gene. Cycle sequencing products were cleaned of excess terminator dyes with Sephadex purification, and analyzed on an ABI Prism 3100 Genetic Analyzer (Applied Biosystems). Recent high-yield DNA samples of key geographic forms within C. flavescens were not available in current US museum holdings, and prior to its rediscovery in 2006, the only known specimen of C. obrieni was the holotype, collected in 1926. As such, we employed ancient DNA sequencing techniques to obtain complete ND2, ND3, and partial HMGN2 sequence data from toepads of museum skins in C. flavescens (seven specimens), the C. obrieni holotype, and the widely allopatric Celeus flavus subflavus (one specimen). The later was sequenced only for ND2 and ND3. Genomic extractions of ancient DNA samples (collected from 1923 to 1926) were performed outside of the main KUNHM molecular facility in a lab free of PCR and genomic procedures. Prior to each extraction, the workstation and all equipment was cleaned with a 5% bleach solution, and to further protect against contamination, filtered pipette tips were used throughout ancient DNA sequencing procedures, as were multiple negative controls during extraction and amplification to enable detection of possible contamination. Samples were extracted using a DNeasy Tissue Kit (Qiagen), extending the tissue lysis step overnight with the addition of 10 ll 1 M dithiothreitol to facilitate complete lysis of the sample. A suite of internal primers were designed to amplify 250–300 bp fragments, with each amplicon overlapping a minimum of 10 bp excluding primers (Table 2). Subsequent purification, cycle sequencing, and analysis of ancient DNA samples follow standard protocols detailed for fresh DNA samples. 2.3. Phylogenetic analysis Chromatograms of complimentary strands were compiled in SEQUENCHER 4.1 (Gencodes) and all sequence alignments were per-

formed in CLUSTAL X (Thompson et al., 1997) using default settings. Alignment of nuDNA markers was straightforward, as sequence variation was minimal in both loci. Gaps resulting from indels in DI and DII of control region sequences and the few indels present within nuDNA loci were corrected by eye in MESQUITE v. 2.72 (Maddison and Maddison, 2009). Heterozygous sites in nuDNA markers were inferred by the presence of equal intensity double peaks in chromatograms of both strands, and were assigned the respective IUPAC ambiguity codes. Best-fit models of evolution for each gene and individually concatenated mtDNA and nuDNA data sets were determined under the Akaike Information Criteria (AIC) implemented in jMODELTEST 0.1.1 (Posada and Crandall, 2001; Posada, 2008; Guindon and Gascuel, 2003). Potential conflict in phylogenetic signal among and within nuDNA and mtDNA data sets was assessed by comparing topology and node support across individual gene tree analyses with discordance between the respective genomic data sets inferred by nodes that were in disagreement at a 70% or higher maximum likelihood (ML) bootstrap support value. We examined evolutionary rate heterogeneity across lineages using a likelihood ratio test (LRT) to determine the difference in likelihood scores for a ML topology with and without a molecular clock enforced. Twice the difference in log likelihood value was compared to a Chi-square distribution with n  2 degrees of freedom, where n = number of taxa. Base homogeneity was also tested in PAUP⁄ v.4.0b10 (Swofford, 2002) using the v2 test of homogeneity to further examine possible sources of discordance in phylogenetic signal among loci. Maximum likelihood analyses were conducted in GARLI v0.951 (Zwickl, 2006), which is an evolutionary computing application that uses a genetic algorithm to simultaneously estimate model parameters and tree topology thereby yielding significant advances in computational efficacy for large data sets. ML trees were estimated for individual genes as well as the individually concatenated mtDNA and nuDNA data sets. A total of thirty runs under default parameters were conducted for each data set to ensure the optimal ln L solution had been reached. Topologies were selected after 10,000 generations with no significant improvement in ln L

32

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

imported into RAVEN v1.3 (Cornell Lab of Ornithology, Charif et al., 2008) and visualized as spectrograms to assess song attributes, which were characterized by note number, frequency (max, min, delta, central, and 95% power), note duration, and mean inter-note duration (Table 6). Measurements were performed on the fundamental harmonic in all taxa, and measurements of the second note complex are means of all notes within the complex unless stated otherwise. Representative spectrograms were exported to Adobe Photoshop CS3 as image files and enhanced by manually removing background noise to maximize visual clarity for qualitative comparison in Fig. 5. Plumage traits including pectoral band and barring of ventral, dorsal and tail plumage were scored as multi-state characters by examining museum study skins and primary literature. These traits were then mapped onto the mtDNA consensus topology to evaluate the evolutionary history of several Celeus plumage characters that have influenced previous taxonomic arrangements. Limited intraspecific taxon sampling in the present study fails to capture the full complexity and variation of Celeus phenotypic traits within most lineages; consequently, our qualitative assessment of plumage evolution is limited to traits that exhibit minimal intraspecific variation and thus represent broader trends within the genus.

(improvement values set at 0.01 with a total improvement lower than 0.05 compared to the last topology recovered). Node support was assessed using 1000 non-parametric bootstrap replicates under the same parameters as above. We conducted Bayesian analyses in MRBAYES 3.1 (Ronquist and Huelsenbeck, 2003), with a flat default prior distribution implemented for parameter estimation. The concatenated mtDNA data set was partitioned by gene and analyzed independently of nuDNA sequences given significant conflict in phylogenetic signal among these genomes. Independent Bayesian analyses were conducted on each nuDNA genes, which were also analyzed together and partitioned by gene. Data partitions were permitted to vary independently in order to optimize model specificity by unlinking all parameters except topology and branch length. Four Markov chains were used to sample the parameter and tree space with all analyses run for 2  107 generations and sampled every 100 generations, resulting in a total of 2  105 samples. Multiple analyses were run to avoid convergence on local optima, and stationarity of each run was determined by monitoring average standard deviation of split frequencies, plotting ln L against generation time, assessing model parameter posterior probability densities in TRACER v. 1.5 (Rambaut and Drummond, 2007), and examining clade posterior probabilities across runs using the compare and slide functions in AWTY (Nylander et al., 2008). All trees sampled prior to the analyses reaching stationarity were discarded as a burnin. Considerable debate has centered around the implications of missing data on phylogenetic inference, however data simulations appear to indicate the quantity of missing data may be less critical than previously thought, with quality of existing data likely of greater importance in phylogenetic reconstruction (Wiens, 2003). In order to assess the effect of missing data on topology and node support we conducted additional Bayesian analyses on a ND2–ND3 data set, and a third run excluding partial sequence taxa from the full mtDNA data matrix.

3. Results 3.1. Sequence attributes The 6-gene combined mtDNA sequence alignment contained 3415 characters for 34 Celeus samples and two picid outgroup taxa yielding 991 (29%) variable sites, of which 753 (22%) were parsimony informative (Table 3). ATP synthase subunit 8 exhibited the highest substitution rates (39.8%), followed by ND2 (31.3%), COXIII (29.6%) ATP6 (28.9%), ND3 (28.2%), and CR (25.3%), however when analyzed by domain, the higher rates characteristic of the CR domain 1 (33.7%) were evident. All sequences appeared to be of mitochondrial origin rather than nuclear copies, as no stop codons were present in open reading frames, overlapping fragments did not conflict, base composition was homogeneous across taxa, and codon substitution rates were consistent with other picid studies that used these genes (Fleischer et al., 2006; Fuchs et al., 2008). The aligned control region sequences contained nine single basepair indels that were restricted to DI and DII, while the conserved central domain exhibited little differentiation across taxa. A 44 bp thiamine motif (four replicates of TTTTTTTTTCA) within the middle of DII precluded obtaining reliable sequence at the 3’ end of DII, thus approximately 300 bp from the highly variable 3’ end of DII were excluded from the analysis. All ND3 samples contained a single cytosine insertion at position 174; however, this extra base is not translated in birds and therefore was removed from our analyses (Mindell et al., 1998).

2.4. Analysis of vocalizations and plumage Vocal recordings of all Celeus species were acquired from multiple sources including the Macaulay Library (http://macaulaylibrary.org/index.do), Xeno-canto America (http://www.xenocanto.org/america/), commercially published material (Table 4), and personal recording collections (authors and others, see acknowledgments) to conduct an initial qualitative analysis of Celeus song and assess the utility of these behavioral traits in phylogenetic inference. Approximately 85 audio samples were examined to identify the primary vocalization for each taxon; however, the lack of multiple samples per individual, recording context, sound quality, and regional sampling precluded detailed quantitative characterizations. Audio samples for each taxon were

Table 3 Attributes of sequence variation in eight genes across Celeus woodpeckers. Gene

ND2 ND3 ATP-6 ATP-8 COIII CR mtDNA ß-FIBI7 HMGN2 nuDNA

Total sites

1041 351 684 168 192 957 3415 911 693 1604

Informative sites (%)

244 (23.4) 80 (22.7) 156 (22.8) 49 (29.2) 40 (20.8) 184 (18.6) 753 (22.0) 21 (2.3) 26 (3.7) 47 (2.9)

Variable sites by codon (Informative)

Nucleotide frequencies

1st (%)

2nd (%)

3rd (%)

%A

%C

%G

%T

41 (11.8) 17 (14.5) 28 (12.5) 8 (14.3) 11 (17.2) NA NA NA NA NA

22 (6.3) 6 (5.1) 6 (2.6) 9 (16.1) 1 (1.56) NA NA NA NA NA

181 58 122 32 28 NA NA NA NA NA

29.6 24.6 25.4 32.4 25.6 24.2 26.8 30.9 25.7 28.7

38.9 37.4 38.5 37.7 34.9 27.5 35.3 17.8 16.5 17.2

9.0 12.0 11.0 6.1 15.8 16.2 11.9 18.0 27.3 22.1

22.5 26.0 25.1 23.8 23.7 32.1 26.0 33.3 30.5 32.0

(52.2) (49.6) (54.5) (57.1) (43.8)

Best-fit model (AIC)

ln L

GTR + I + C TrN + C GTR + I + C TrN + I HKY + C GTR + I + C GTR + I + C HKY TIM1 + C TIM1 + C

4069.9526 1333.9293 2526.5578 685.0652 700.4243 3865.2117 13618.07 1600.44 1318.64 2954.7

0.011 0.083 0.082 0.003 0.082 0.081 0.092 0.095 0.1 0.094 0.011 0.089 0.093 0.098 0.092 0.046 0.047 0.091 0.094 0.104 0.1 0.007 0.047 0.048 0.089 0.093 0.103 0.099 0 0.013 0.013 0.015 0.045 0.045 0.086 0.086 0.083 0.086 0.085 0.083 0.082 0.085 0.083 0.087 0.099 0.1 0.097 0.105 0.002 0.03 0.001 0.001 0.014 0.014 0.016 0.047 0.047 0.087 0.087 0.084 0.087 0.086 0.084 0.083 0.086 0.084 0.089 0.1 0.101 0.098 0.106

0.032 0.003 0.003 0.016 0.016 0.018 0.049 0.049 0.09 0.09 0.086 0.09 0.088 0.086 0.085 0.089 0.086 0.091 0.102 0.103 0.1 0.109

0.029 0.029 0.025 0.025 0.027 0.053 0.053 0.086 0.084 0.083 0.088 0.087 0.086 0.085 0.089 0.091 0.096 0.093 0.094 0.1 0.104

0.013 0.013 0.015 0.045 0.045 0.086 0.086 0.083 0.086 0.085 0.083 0.082 0.085 0.083 0.087 0.099 0.1 0.097 0.105

0 0.006 0.045 0.045 0.086 0.086 0.082 0.088 0.087 0.083 0.082 0.085 0.08 0.085 0.096 0.097 0.099 0.102

0.006 0.045 0.045 0.086 0.086 0.082 0.088 0.087 0.083 0.082 0.085 0.08 0.085 0.096 0.097 0.099 0.102

0.045 0.045 0.086 0.084 0.083 0.088 0.087 0.085 0.084 0.085 0.083 0.087 0.096 0.097 0.099 0.102

0 0.092 0.092 0.088 0.094 0.093 0.092 0.093 0.092 0.092 0.097 0.101 0.102 0.116 0.117

0.092 0.092 0.088 0.094 0.093 0.092 0.093 0.092 0.092 0.097 0.101 0.102 0.116 0.117

0.002 0.003 0.044 0.043 0.085 0.082 0.083 0.08 0.082 0.094 0.095 0.1 0.097

0.005 0.044 0.043 0.085 0.082 0.083 0.081 0.082 0.094 0.095 0.1 0.097

0.045 0.044 0.084 0.081 0.082 0.079 0.081 0.093 0.094 0.099 0.095

0.003 0.092 0.091 0.095 0.088 0.091 0.104 0.105 0.098 0.097

0.094 0.092 0.093 0.086 0.09 0.105 0.106 0.097 0.096

0.007 0.01 0.048 0.049 0.093 0.096 0.104 0.1

19 18 14 13 12 11 10 9 8 7 6 5 4 1

2

3

elegans (14) elegans (9) elegans (17) lugubris (34) lugubris (35) flavescens (18) flavescens (23) flavescens (24) flavescens (20) flavescens (19) undatus (43) grammicus (28) grammicus (27) castaneous (1) castaneous (2) flavus (7) flavus (3) flavus (8) obrieni (36) spectabilis (38) loricatus (31) loricatus (32) torquatus (41) torquatus (39)

Taxon

C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Table 4 ND2 pairwise genetic distances (corrected with GTR model) across Celeus woodpeckers. Numbers in parentheses correspond to taxon id’s in Table 1.

15

16

17

20

21

22

23

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

33

Four nuclear markers were initially screened for this study, but given the low yield of informative variation recovered in the initial screening the final nuDNA data set was limited to ß-FIBI7 and HMGN2 gene regions. Sequences were obtained for both markers from at least two samples per taxon with the exception of Celeus grammicus in which only one sample was sequenced for HMGN2, and Celeus spectabilis, for which sequence was obtained of only one specimen from both markers. Among the 911 bp ß-FIBI7 alignment, 21 bp (2.3%) were parsimony informative, whereas 26 bp (3.7%) of the 693 bp HMGN2 alignment were informative (Table 3). Few insertions or deletions were present in either locus, with informative indels limited to a 3 bp and 16 bp deletion in the HMGN2 locus shared by all members of clade C. 3.2. Ancient DNA Complete ND2 and ND3 sequences were obtained from ancient DNA samples of nine additional ingroup specimens via alignment of five ND2 amplicons per specimen, whereas complete ND3 sequences were obtained with the standard external primers (Table 2). Overlap between ND2 fragments ranged from 51 to 82 bp, with no conflicts among amplicons, each of which exhibited sufficient synapomorphic substitutions to confirm the correct target sequence had been obtained. Two cytosine–thymine (C–T) double peaks were observed when reviewing chromatograms of ancient DNA samples, indicating a potential deamination of cytosine to thymine on the L strand, a phenomenon that has been previously reported as one of the more common errors encountered when working with degraded ancient DNA (Hofreiter et al., 2001; Sefc et al., 2006). These samples were re-amplified and sequenced to clarify the correct base calls for both taxa. Additional sequence anomalies of potential concern include seven autapomorphic substitutions observed in ancient DNA samples at sites that were otherwise invariable in fresh tissue samples. Of these, three were third-position adenine–guanine (A–G) transitions, not unexpected given the mutational bias of third position codons. Two second-position transitions (G–A, C–T) and a first position A–G transition were also observed and may represent sequence errors due to lesions in the ancient DNA templates, considering the relative conservatism of these sites. Sequence anomalies appeared to be randomly distributed across ancient DNA samples and amplicons. Cloning techniques were not used to confirm the nature of these anomalies as these sites had no appreciable impact on the overall phylogenetic signal of the respective sequence; however, they serve to highlight the potential shortcomings of working with degraded low yield DNA samples, especially in the realm of population genetics where accurate genotyping may be critical. All new sequences used in this study have been deposited in GenBank under the accession series JF433088 to JF433371. 3.3. mtDNA phylogenetic analysis Aikake Information Criteria (AIC) values generated in jMODELTEST 0.1.1 (Posada, 2008) indicated a general time reversible model of evolution (GTR + I + C) was most appropriate for ND2, ATP-6, and CR, whereas the TrN + C, TrN + I and HKY + C models were selected for ND3, ATP-8, and COXIII respectively. Individual gene analyses all converged on the same overall topology albeit with minor, non-significant variation in node support, thus we present results from the combined mtDNA data set. Bayesian and ML analyses based on the concatenated mtDNA alignment produced near identical topologies, supported by significant posterior probability/ bootstrap values (.95/85) for all but two nodes at the species level. Independent Bayesian analyses conducted on an ND2–ND3 data set with no missing data indicated the impact of partial sequences

34

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

Table 5 Sources of primary vocalizations analyzed in Raven, with all but C. ochraceus depicted in spectrographs. Species

Locality

Source

Voucher #

Recordist

Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus Celeus

Belize: 28 km NW of Middlesex Peru: Loreto; 20 km NE of Iquito Venezuela: Monagas; Cano Colorado Brazil/Argentina: Iquazu Bolivia: La Paz Ecuador: Esmeraldas Paraguay: Ypacaraí; Near Asunción Ecuador: Yasuni NP Guyana: Demerara–Mahaica Venezuela: Bolivar Brazil: Tocantins; Recursolândia Brazil: Pará; Monte Alegre

Macaulay Library Macaulay Library Private collection Private collection Macaulay Library Jahn et al. (2002) Private collection Moore (1997) Macaulay Library Private collection Xeno-Canto: Americas Xeno-Canto: Americas

6789 34210 – – 101881 – – – 84992 – XC20900 XC32215

I. Davis T. Parker C. Marantz D. Finch B. Hennessey J. Moore D. Finch R. Ridgely D. Finch D. Finch M. Barbosa S. Dantas

castaneus flavus elegans flavescens grammicus loricatus lugubris spectabilis torquatus undatus obrieni ochraceus

on topology and node support within the full mtDNA data set was negligible. Consequently, all nine ancient DNA samples were included in the final concatenated mtDNA results presented herein. Monophyly of Celeus was strongly supported in all analyses, with four primary clades consistently recovered (Fig. 1). A basal cis–trans Andean lineage (clade D) comprising Amazonian Celeus torquatus and Mesoamerican/Choco Celeus loricatus was recovered as sister to the remainder of Celeus, and exhibited the greatest pairwise sequence divergence at 8.3% (all distances hereafter are ND2 GTR corrected; Table 4) for any taxon pair. The Amazonian ‘elegans’ complex encompassing C. elegans, C. lugubris, and C. flavescens (clade A) was recovered as sister to a broadly distributed suite of taxa arranged in subclades B and C comprised of (C. castaneus (C. grammicus, Celeus undatus)) and (C. flavus (C. spectabilis, C. obrieni)) respectively. Broad geographic sampling within clade A revealed mitochondrial paraphyly in C. flavescens and C. elegans, with the distinctive C. f. ochraceus (hereafter Celeus ochraceus) sister to the four remaining taxa within the clade (Figs. 1 and 2). The long-crested forms of C. elegans from the Guianan shield are sister to an Amazonian clade comprised of nominate C. flavescens and the sister pair C. elegans ‘jumana’ and C. lugubris, however this three taxon arrangement is not strongly supported in either analysis. Less than 0.5% average pairwise sequence divergence separates ‘jumana’ forms from the phenotypically and vocally distinct C. lugubris, with several individuals in fact sharing identical ND3 haplotypes. By contrast, C. ochraceus exhibits more substantial genetic differentiation (4.5– 5.3%) from other members of the clade, followed by moderate to low divergences among the respective sister lineages of nominate C. elegans and C. flavescens (2.5–2.9% and 1.3–1.8%). Within clade B, the Meosamerican C. castaneus exhibited little conspecific sequence divergence (0.3%) across Mexico–Panama sample localities, whereas a more substantial split (4.3–4.4%) separates it from its sister lineage grammicus—undatus, which comprise the smallest members of the genus and whose minimal morphological, behavioral, and genetic differentiation (0.2–0.3%) draw into question their status as distinct species. Clade C consists of three Amazonian taxa including the phenotypically variable C. flavus, which showed little genetic differentiation across its broad geographic distribution. The allopatric Atlantic forest taxon C. f. subflavus exhibited 0.7–1.0% pairwise sequence divergence from Ecuador and Guyana C. flavus populations. Our data corroborate the close sister relationship between the enigmatic C. obrieni and the distantly allopatric C. spectabilis (1.1%), both of which are range restricted taxa, the former containing no geographic variation and the latter showing little differentiation. With the exception of two nodes in clade A, this topology was well supported with strong concordance among BA and ML analyses. These phylogenetic results differ considerably from the current taxonomic arrangement of Celeus diversity, and are treated in

greater detail within the discussion, as is the issue of mtDNA paraphyly recovered in C. elegans and C. flavescens. Overall, genetic distances among ingroup taxa were relatively shallow ranging from 0.3% between grammicus and undatus to 11.7% between torquatus and flavescens (Table 5), which is consistent with the emerging picture of woodpecker phylogeny (Webb and Moore, 2005; Benz et al., 2006; Moore et al., 2006; Fuchs et al., 2008).

3.4. nuDNA and combined-data phylogenetic analyses Based on AIC values recovered from model selection analyses, the HKY model was selected for ß-FIBI7, whereas TIM1 + C was most appropriate for the HMGN2 locus (Table 3). Phylogenetic analysis of ß-FIBI7 recovered three of the four clades inferred in the mtDNA data set, with clades B and C lumped into an unresolved polytomy that included significant support for the C. grammicus–undatus sister relationship (Fig. 3). The monophyly of C. elegans was recovered with a posterior probability of 1.0 and moderate ML bootstrap support (72%), however the relationships within clade A were not well resolved. Similarly, analysis of the HMGN2 locus recovered all four mtDNA clades with moderate to strong support, but relationships within clades were not resolved due to an absence of informative sequence variation at this scale. This topology differed from ß-FIBI7 in that C. lugubris was nested within C. elegans, but with non-significant node support. Lastly, nominate C. flavescens were recovered as a separate clade distinct from C. ochraceus (Fig. 3). As differences in topology among loci were not significantly supported, a concatenated nuDNA alignment was analyzed to improve phylogenetic resolution through combining potentially concordant phylogenetic signal among loci. The combined analysis resulted in stronger node support across the topology, with moderate support for a sister relationship between clades B and C and greater resolution within B, confirming C. castaneus as sister to the grammicus–undatus lineage (Fig. 4). We conducted a Bayesian analysis on the combined nuclear (ß-FIBI7 + HMGN2) and mitochondrial (ND2 + ND3) data sets with introgressed C. e. jumana samples removed from the alignment to explore further the phylogenetic relationships among members of the elegans–flavescens complex. The partitioned Bayesian analysis recovered a well-resolved topology identical to that of the mtDNA results, with the exception of relationships within clade A, in which nominate C. elegans was placed as sister to C. lugubris. These taxa were in turn sister to nominate C. flavescens, with C. ochraceus occupying the basal position within the clade. This discordance is not surprising given the weak node support recovered for C. elegans and C. flavescens in the mtDNA analysis, the taxonomic implications of which are discussed below.

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

35

Inter-note duration 95% power frequency

2.5840 1.9810 2.5840 2.5840 3.1008 2.3256 1.3125 1.2000 1.5504 1.0336 1.8949 1.2059 1.7227 1.0336 1.7227 1.2920 1.7801 2.0672 1.6537 2.5840 2.0672 2.8854 2.4117 2.4117 2.4117 2.2394 1.6655 2.4117 2.4117 2.7562 1.9810 1.1250 1.1250 1.2059 0.8613 1.3781 1.0336 1.5504 0.7752 1.5504 1.0623 1.3781 1.7227 1.3781 2.4117 1.8949 2.7131 2.0672 2.2395 2.2395

Frequency bandwidth Center frequency

2.4117 1.8088 1.4615 1.4434 2.2335 1.4389 1.0478 0.8282 1.2082 0.6904 1.1693 0.8114 1.4099 0.6934 1.3626 0.7267 1.3089 1.6633 1.1189 1.7850 1.3440 0.5543 1.7270 1.7160 1.8806 2.6873 2.0714 2.8436 2.7562 3.1570 2.4375 1.4969 1.3122 1.6613 1.1219 2.1239 1.4318 2.0802 1.1094 1.8829 1.3543 2.0068 2.2850 1.7641 2.7090 2.1840 2.9893 2.5214 2.5623 2.6563

Maximum frequency Minimum frequency

0.4479 0.4059 1.3821 1.3128 0.9235 0.9986 0.4491 0.8400 0.4531 0.4315 0.9545 0.6205 0.6703 0.4160 0.5203 0.6276 0.6978 0.6217 0.6452 0.9240 0.8400 2.4350 0.7944 0.8463 0.7757 0.2020 0.1565 0.2590 0.1795 0.1100 0.0540 0.3050 0.0528 0.2490 0.3250 0.2880 0.3730 0.2460 0.0515 0.1780 0.0378 0.0346 0.2090 0.0642 0.2490 0.1640 0.0972 0.0366 0.2630 0.1850

Note duration Note # (range)

1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2a 2b 1 2 1 1 (ave) 1 (ave) 2 1 1 (ave)

1 2 (2–15) 1 5 1 2 (2–3) 1 4 (2–5) 1 1 1 1 1 2 (2–10) 1 5 (5–11) 5 (5–11) 1 5 4 (4–15) NA 4 (2–4) 2 (1–2) 4 (3–8) NA

Note complex

C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C. C.

flavescens flavescens ochraceus ochraceus lugubris lugubris elegans elegans grammicus grammicus undatus undatus castaneus castaneus spectabilis spectabilis spectabilis obrieni obrieni flavus flavus loricatus loricatus torquatus torquatus

Taxon

Table 6 Summary of nine acoustic variables measured across Celeus vocalizations.

0.265 0.280 0.200 0.263 0.232 0.224 0.249 0.259 0.171 NA 0.170 NA 0.142 0.133 0.091 0.066 0.012 0.071 0.082 0.185 0.172 0.104 0.118 0.162 0.190

3.5. Song structure and diversity As in all Picinae, Celeus woodpeckers use bill drumming for inter and intraspecific communication in addition to primary and secondary vocalizations. Most taxa have multiple secondary vocalizations, typically strident and short in duration, that appear to be used for communicating with mates as well as inter-territorial contexts. Our qualitative assessment of Celeus song focuses solely on the primary vocalization that presumably is used in the context of soliciting a mate and territory maintenance. To facilitate interpreting spectrographs and song descriptions herein, we refer the reader to two on-line sound recording collections, where vocalizations of all taxa can be consulted: Xeno-canto Americas (http:// www.xeno-canto.org/america/) and the Macaulay Library (http:// macaulaylibrary.org/index.do). Source, collection number, and locality information of vocal samples detailed in the following accounts are given in Table 5. We use phonetic descriptions of these primary vocalizations from works where they have been accurately described (Stiles and Skutch, 1989; Ridgely and Greenfield, 2001; and Hilty, 2003) to compliment acoustic measurements performed in Raven (Table 6). Primary vocalizations across Celeus are highly simplistic in terms of song structure and note composition, consisting of one or two distinct note complexes with little to no frequency modulation of individual notes. The first complex is characterized by a single note that is slightly higher in amplitude and frequency and longer in duration compared to the notes that comprise the second complex, with the exception of C. loricatus and C. grammicus–undatus. Strong harmonics were present for each note across all taxa; however, the fundamental harmonic comprised the majority of the signal power, with the exception of the second note complex in C. grammicus–undatus, C. castaneus, and C. flavescens, all of which exhibited the greatest signal power in the second or third harmonic (Fig. 5). Three distinct song types are present in clade A. Although the primary vocalization of C. ochraceus was omitted from Fig. 5 given the limited comparative material available, vocalizations of C. ochraceus are highly similar to C. flavescens (pers. comm. Kevin Zimmer). Song structure of C. flavescens consists of two complexes, the first of which is a single explosive pure tone note (2.5 kHz) followed by a series of two to four notes that are .6 kHz lower in frequency and weakly modulated from 0.4 kHz to 2.0 kHz producing a faint raspy quality (Kree, reek, reek). Call structure and note composition are similar in C. lugubris, but with an explosive first complex note that is higher in pitch (3.1 kHz) followed by a series of raspy notes in the second complex that are shorter in duration and faster in tempo (Kree!, rac rac). Unique within Celeus, all forms of elegans seem to give the primary vocalization less frequently and at lower amplitudes (pers. comm. Curtis Marantz). As such, many experienced field workers are only familiar with the raspy scolding and contact calls of this species. The primary vocalization is composed of a modulated single note in the first complex (1.3 kHz) that is notably lower in frequency compared to other species in the clade. A series of short, lower frequency (1.2 kHz) modulated notes form the second complex, also having a raspy laughing quality (Wakrrik, wahk-wahk-wahk-wahk-wahk). Limited vocal material of the primary vocalization of this taxon precluded identification of potential differences among the ‘elegans’ and ‘jumana’ forms. Within clade B, grammicus and undatus share similar vocalizations that are unique from all other Celeus. Both produce a loud, two note call with the first slightly ascending in frequency and the second descending (kuwee? Kuuu). The first and second notes in grammicus exhibit the lowest center frequency (1.2 and 0.861 kHz) within Celeus and undatus is only marginally higher in pitch. Given the near lack of variation between these primary vocalizations (Fig. 5), most field workers consider them ‘‘identical’’

36

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

Fig. 1. Phylogenetic relationships within Celeus woodpeckers as inferred from a maximum likelihood analysis of the concatenated mtDNA data set. Nodal support is indicated by non-parametric bootstrap values above and Bayesian posterior probabilities below. Nodes with blackened circles indicate ML bootstrap and posterior probability values of 90/.95 or higher whereas an  indicates the node was unresolved in the Bayesian analysis.

(Ridgely and Greenfield, 2001; Hilty, 2003). The first complex of C. castaneus is higher in frequency (1.7 kHz) and descending in pitch compared to the former sister pair. The second complex is composed of 2–10 sharp notes given with little delay in between, producing a laughing quality (Khee, kew-kew). Two song types are present in clade C, with primary vocalizations given by C. spectabilis and C. obrieni exhibiting a high degree of similarity, including an initial ascending squeal (1.7–2.0 kHz) comprising the first complex followed by a series of shorter modulated clucking notes that in spectabilis contain two syllables with

the second slighter higher in frequency, resulting in a faster tempo of the second complex (squeeah! kluh-kluh-kluh-kluh-kluh). By contrast C. flavus gives a series of four to six pure tone (2.5–2.0 kHz) notes that typically decrease in frequency and duration in each successive note. The far carrying, ringing quality and single complex structure of this call is most similar to that of C. torquatus in clade D, whose call structure is identical, but with notes somewhat lower in frequency (2.2 kHz), typically held at the same pitch from note to note, and longer in duration yielding a faster tempo. In the trans-Andean sister taxon C. loricatus, the pure tone notes are 

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

37

Fig. 2. Approximate distributions, sampling localities, and associated mtDNA and nuDNA maximum likelihood and Bayesian estimates of phylogeny within the ‘elegans’ complex. Nodes with blackened circles indicate ML bootstrap and posterior probabilities of 90/.95 or higher in the mtDNA topology.

Fig. 3. Phylogenetic relationships of Celeus inferred from individual gene trees of ß-FIBI7 and HMGN2 loci, with ML bootstrap/posterior probability support values indicated above and below each node respectively, whereas an  indicates the node was unresolved.

half the duration of those in torquatus and higher in frequency (2.7 kHz). A second note complex comprised of 2–4 shorter, weakly

modulated notes that successively descend in frequency and amplitude further distinguish C. loricatus from its sister taxon.

38

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

Fig. 4. Maximum likelihood and Bayesian estimates of phylogeny inferred from the nuDNA (left) and Bayesian topology of combined nuDNA + mtDNA (ND2 and ND3) data sets partitioned by gene. Non-parametric bootstrap support values are indicated above nodes and Bayesian posterior probabilities are given below. Significant node support (ML bootstrap and posterior probabilities of 90/.95) is indicated by solid black circles. All terminal nodes in the combined nuDNA + mtDNA Bayesian topology received significant support. Clade labels A–D correspond to those recovered in mtDNA analyses (Fig 1), and colored taxon labels indicate sample localities depicted in Fig. 2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

4. Discussion 4.1. Discordance among mtDNA and nuDNA topologies Although both nuclear loci exhibited little genetic variation, contributing to individual gene trees with marginal node support and lack of resolution within clades, relationships among clades were for the most part in agreement with the well-resolved mtDNA topology, as evidenced by the same four primary clades that were recovered from each genome. A notable exception of this congruence was manifest by mitochondrial paraphyly in C. elegans, which exhibited substantial genetic divergence among the nominate elegans forms of the Guianan shield and Amazonian jumana forms, the latter differing less than 0.5% average pairwise sequence divergence (ND2) from C. lugubris. This arrangement may be attributed to several possible phenomena including introgression of mtDNA from lugubris into elegans ‘jumana’ forms, incomplete lineage sorting processes including deep coalescence, or simply the presence of cryptic species diversity masked by labile plumage evolution. The latter can be rejected based on a well-defined suite of morpho-

logical and vocal characters shared by nominate ‘elegans’ and ‘jumana’ forms; moreover, reciprocal monophyly of C. elegans was recovered from ß-FIBI7 and combined nuDNA analyses. The monophyly of elegans within the ß-FIBI7 gene tree indicates coalescence of this locus occurred post speciation, thus if mtDNA paraphyly was the result of incomplete lineage sorting, the coalescence time of mtDNA would have to exceed that of ß-FIBI7, an unlikely scenario given the fourfold reduction in effective population size inherit to the mitochondrial genome. Although this does not reject the possibility of incomplete lineage sorting altogether, and the lack of substantial genetic variation among nuDNA loci warrant caution in their interpretation, hybridization between C. lububris and C. e. jumana appears to be the most parsimonious explanation to account for this discordance given the partial sympatry of these taxa and several purported hybrids (AMNH 34294, 127134, and NHMW 57531) described by Short (1972). Widespread introgression of lugubris mtDNA throughout the Amazonian ‘jumana’ forms suggests this process was not initiated recently. At the same time, shared ND3 haplotypes among these taxa illustrates the lack of significant mtDNA differentiation across a broad geographic expanse indicat-

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

39

Fig. 5. Representative spectrograms of vocalizations across Celeus woodpeckers, with letters A–D corresponding to mtDNA/nuDNA clades. Measurements of individual notes are given in Table 6. Units are in seconds (x axis) and kHz (y axis).

ing the initial hybridization was not a deep history event and may be ongoing given the above mentioned intermediate specimens. The nature of mitochondrial paraphyly recovered in C. flavescens is less clear, as the position of nominate C. f. flavescens was not strongly supported within the mtDNA topology. Nonetheless, the

presence of a deep mitochondrial genetic split between C. ochraceus and nominate C. flavescens, coupled with genetic variation at the HMGN2 locus (Fig. 4) and substantial morphological variation both in body size and plumage traits (Fig. 6), clearly indicates two distinct species are involved.

40

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

4.2. Phylogeny and taxonomic implications Reliance upon phenotypic traits in previous taxonomic treatments has confounded several aspects of Celeus phylogeny, which we address within our molecular phylogenetic framework presented herein (Peters, 1948; Short, 1972, 1982; Winkler and Christie, 2002). Traditional linear arrangements have treated C. loricatus as the basal taxon within the genus due to shared similarities with the purported Old World congener Micropternus [Celeus] brachyurus, whereas C. torquatus occupied the opposite end of this arrangement, loosely associated with C. spectabilis (Peters, 1948; Short, 1982; Winkler and Christie, 2002). Conversely, our analyses recovered a well-supported loricatus–torquatus basal dichotomy that is sister to the remainder of Celeus. Broad geographic sampling within C. torquatus revealed minimal genetic differentiation (1.1%) among the nominate Guianan form and the Amazonian C. t. occidentalis, while just 0.3% sequence divergence separates nominate loricatus sampled in Ecuador from the central Panamanian C. l. mentalis. Although both species encompass highly distinctive geographic forms, their subspecific allocation is appropriate until further phylogeographic investigation of these forms is undertaken. In his detailed comparative analysis of Celeus morphometrics and plumage traits, Short (1972) recognized castaneus shares a number of traits with the grammicus–undatus lineage including broad red facial markings, barred ventral and dorsal surfaces, rufous base of the tail, and small size. However, he concluded that the presence of a curved culmen, proportionally larger bill, longer crest, and unmarked throat aligned castaneus more closely with the elegans–flavescens complex. Contrary to this phenetic arrangement, our results indicate unequivocally a close sister relationship between castaneus and grammicus–undatus, a clade which is not directly related to the other Mesoamerican congener, C. loricatus. The close sister relationship of grammicus–undatus has never been disputed; however, given the subtle differences in phenotype, behavior, and voice, combined with extremely low genetic differ-

entiation, the case could be made to treat these forms as a single species. From a plumage standpoint, levels of intraspecific variability are equivalent to those between grammicus and undatus. Near the contact zone in the Delta Amacuro region of eastern Venezuela, birds assigned to undatus amacurensis closely approach grammicus with the exception of lightly barred rectrices, opening the question to the correct allocation of this taxon. Given that mtDNA divergences within grammicus exceed that among grammicus–undatus, several informative substitutions within the nuDNA loci supporting a well resolved clade B was unexpected, and opens the possibility of additional mtDNA introgression among these taxa. Consequently, we defer suggesting a taxonomic change pending denser population-level sampling with additional independent markers to test for evidence of gene flow, especially within western Amazonian populations of grammicus, and the eastern contact zone mentioned above, to fully resolve the status of these taxa. Due to its unique blonde plumage, Short (1982) suggested that C. flavus had no close relatives, but was most likely distantly related to the elegans–flavescens complex. Molecular data reveal this broadly distributed Amazonian species is sister to a taxon pair comprised of C. obrieni and the western Amazonian C. spectablis. Although C. flavus varies considerably in plumage ranging from bright blonde to dull buff with dusky-olive dorsal and ventral markings present in some populations, samples across its distribution reveal minimal genetic divergence, including the isolated Atlantic forest C. f. subflavus. Prior to its rediscovery in 2006 (A.D. do Prado, http://www.birdlife.org/news/news/2006/12/caatinga_woodpecker_redisc.html), C. obrieni was known from only the holotype collected in 1926 at Urussuhy (presently Uruçuí; 07°140 S, 44°330 W), on the Rio Parnáiba in northeast Brazil, state of Piauí. Charles O’Brien initially brought the specimen to Short’s attention having concluded the odd plumage most closely resembled C. spectabilis. Short (1973) later described this specimen as a subspecies of C. spectabilis; however, the approximately 2400 km that separate these taxa combined with obrieni’s puzzling suite of plumage characters left

Fig. 6. (A) Two C. f. ochraceus specimens sampled for this study (AMNH 278666, left; AMNH 242703, right) illustrating considerable plumage variation within this taxon as well as dramatic phenotypic differences from nominate C. flavescens. (B) Study skins of nominate C. flavescens (AMNH 242688, left) and (AMNH 314390, right) also sequenced for this study. (C) C. obrieni holotype (AMNH 242687). See Table 1. for localities.

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

41

Fig. 7. Mitochondrial consensus tree of Celeus woodpeckers with plumage character states mapped on the topology. The presence of barring on ventral, dorsal, and tail plumage is indicated by barred boxes, whereas concolor plumage is designated by a solid black box. The presence of a pectoral band is indicated by a blackened box.

open the possibility that the specimen was simply an aberrant individual or hybrid between C. torquatus and C. flavus or C. ochraceus. Complete ND2 and ND3 gene sequences confirm the close maternal genetic relationship of C. obrieni and C. spectabilis, and 675 bp from HMGN2 revealed several informative substitutions as well as two indels that further support the placement of C. obrieni within clade C. With multiple pairs of C. obrieni recently discovered across several distant sites spanning the states of Maranhão, Tocantins, and Goiás (the latter site based on two historical specimens collected by J. Hidasi and E. Tomazzeti in 1967 and only recently identified in the ITS-UCG collection), the range of this little known species

is now estimated at 280,000 km2 (Pinheiro and Dornas, 2008; Birdlife International, 2010). The ecological differences between its sister taxon have also become apparent with C. spectabilis specializing in bamboo foraging within primary rain forest (Kratter, 1997), whereas C. obrieni appears to be restricted to cerrado and gallery forest, but also specializing on patches of bamboo vegetation within these environments, including stands of Guadua paniculata (Pinheiro and Dornas, 2008). In considering the genetic, vocal, and ecological differences of these taxa, we treat obrieni at the species level, and concur with the previously suggested English name Kaempfer’s woodpecker, after its collector Emil Kaempfer. Nonetheless, additional sampling for both taxa is

42

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

needed to test for reciprocal monophyly and confirm the maintenance of distinct evolutionary lineages now that the potential distribution of C. obrieni has come to light. Six subspecies are currently recognized within the broadly distributed elegans complex, which Short (1973, 1982) arranged in two groups (treated as species by Peters (1948)) based largely on crest length and crown color. The long-crested, pale-crowned nominate Guianan shield forms include: elegans from the Brazilian state of Roraima south to the Amazon and east to Amapa and French Guiana; hellmayri from eastern Venezuela, Guyana, and Suriname; deltanus restricted to the Delta Amacuro, Venezuela; and leotaudi endemic to Trinidad. The short-crested, dark-crowned Amazonian group includes the dark plumaged citreopygius of eastern Ecuador and Peru, and the widely distributed jumana from eastern Colombia, southern Venezuela, east to the Rio Negro, and south to Mato Grosso and northern Bolivia. South of the Amazon, jumana occurs east to Maranhão. Our limited sampling shows little genetic variation within the Amazonian and Guianan shield forms; however, the lack of introgressed mtDNA among our three Guianan sample sites indicates the maintenance of a distinct lineage within this well defined area of endemism. For reasons discussed above, the jumana and elegans groups appear to be correctly allocated as subspecies within C. elegans, however further genetic and vocal sampling is required on the western rim of the Guyana shield to understand the distributional limits of these lineages and the nature of their interactions at this contact zone. The variable forms of C. ochraceus (Fig. 6) were historically treated as a distinctive subspecies within flavescens whose range it abuts in eastern Brazil. Our mtDNA and combined data analyses indicate this distinctive taxon is in fact the ancestral lineage within the ‘elegans’ complex, exhibiting the highest genetic differentiation within the clade. Although both nuclear loci contained few informative substitutions within the elegans–flavescens complex, the presence of HMGN2 genetic variation distinguishing nominate C. flavescens from C. ochraceus is consistent with reciprocal monophyly recovered in mtDNA and combined data analyses, indicating these lineages are maintaining unique evolutionary trajectories. The presence of HMGN2 genetic variation among these lineages provides strong evidence in diagnosing these taxa as distinct species, considering all members of Clade C shared the same HMGN2 haplotype. Moreover, substantial morphological differences distinguish C. flavescens from C. ochraceus, including smaller size, a conservative trait that may correspond to strategies of foraging specialization distinct from those of the larger C. flavescens and other sympatric Celeus. Consequently, C. ochraceus clearly merits species recognition under the evolutionary and phylogenetic species concepts and we suggest Ochre-backed woodpecker as an appropriate English name. Although some populations of C. ochraceus closely resemble Celeus flavescens intercedens where they meet in southern Tocantins, northern Goiás, and western Bahia (pers. comm. V. Piacentini), the taxonomic status of C. f. intercedens remains unclear and will require dense geographic sampling to clarify the distribution and phylogeographic history of the this highly variable taxon. 4.3. Evolution of vocal and plumage traits in Celeus Vocalizations within woodpeckers and allies are presumed to be genetically ‘‘hardwired’’ and potentially useful traits in understanding species limits and phylogenetic relationships, as song learning is largely limited to the Trochilidae, Psittacidae, and oscine Passeriformes (Baptista and Kroodsma, 2001; Podos et al., 2004). Avian vocalizations typically encompass complex signals composed of multiple discrete traits, each with potentially independent evolutionary trajectories that may be governed by one or more selective mechanisms concomitantly (Lorenz, 1950; Podos

et al., 2004). Although our limited sampling has prevented explicit quantification of these traits within Celeus, several generalizations can be drawn from our qualitative comparisons. Song structure and note composition vary considerably within clades and are largely incongruous with the molecular framework presented herein, with the exception of grammicus–undatus and spectabilis–obrieni both of which represent close sister pair relationships with highly similar vocalizations. By contrast, sister taxa in each respective clade exhibit distinct song types both in terms of complex structure and note composition. The lack of phylogenetic signal within these traits may point to a brief speciational history in which Celeus recently filled its specialized ant and termite foraging ecological niche. Population level quantitative analysis of variation in these vocal attributes, examined within a detailed phylogeographic and ecologically informed context, may shed light on the evolutionary history of Celeus vocalizations. The presence of just two monotypic taxa (castaneus and obrieni) within Celeus highlights the significant phenotypic variation present within members of the genus. This is exemplified within loricatus where the dorso-ventrally barred trans-Andean nominate subspecies dramatically contrasts to the plain concolored innotatus of northern Colombia. Similar patterns of contrast are present in each Celeus clade defined herein. Although the dorsal concolor plumage within clade A is consistent with a monophyletic C. elegans, Celeus elegans hellmayri exhibits extremely light spotting on the dorsum representing an ambiguous intermediate state. With the exception of body size, pectoral band, and strong barring in clade B, few phenotypic characters employed in previous taxonomic assessments unambiguously track the evolutionary history of Celeus woodpeckers (Fig 7). This disparity from previous work serves to highlight the limitations of phenotypic characters in reconstructing the evolutionary history of woodpeckers, adding to a growing body of evidence indicating labile and or convergent plumage evolution across the family (Weibel and Moore, 2002; Benz et al., 2006; Moore et al., 2006). 4.4. Biogeography Celeus woodpeckers are broadly distributed from southern Veracruz, Mexico to northern Argentina, inhabiting a diversity of environmental conditions and showing some similarity in distributional limits with other co-occurring picid radiations (e.g. Picumnus, Piculus; Veniliornis). Resolving the geographic origin and underlying factors that have shaped patterns of community assembly through time are fundamental questions in advancing biogeographic understanding. The robust phylogenetic framework for Celeus presented in this contribution provides several salient points of discussion towards clarifying the biogeographic history of the clade. Although the geographic origin of Celeus remains ambiguous as a consequence of the basal dichotomy among South American C. torquatus and the intercontinental C. loricatus, the phylogenetic position of the Mesoamerican castaneus within an otherwise South American assemblage, coupled with an Amazonian center of Celeus diversity argues against a Central American origin of the genus. Moreover, few intercontinentally distributed avian radiations have dispersed back across the Isthmus of Panama to their continent of origin once trans-isthmus dispersal has taken place (Smith and Klicka, 2010). While moderate genetic divergence separating loricatus–torquatus from the rest of Celeus is consistent with late Pliocene continental interchange corresponding to closure of the Panamanian Isthmus 3.1–4 mya (Coates and Obando, 1996; Kirby et al., 2008), more recent Quaternary phenomena are likely involved in shaping the shallow genetic structure among the remaining Celeus sister taxa. Broadly distributed Amazonian species are present in each of the four clades, and most exhibit a lack of significant genetic differentiation across river systems that

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

define prominent areas of endemism, suggesting ongoing or recent gene flow throughout the region. Further indication of Celeus dispersal capacity is evidenced by the widely allopatric Atlantic forest populations of Celeus torquatus tinnunculus and C. f. subflavus as well as the close sister relationship of spectabilis and obrieni, separated by 2400 km. While the distinctive C. t. tinnunculus remains to be sampled, genetic distances among flavus–subflavus populations and the spectabilis–obrieni split are similar, potentially a consequence of a shared evolutionary history. Three additional woodpecker taxa exhibit similar disjunct geographic distributions with peripherally isolated populations in Atlantic forest including Veniliornis affinis, Piculus flavigula, and Piculus chrysochloros; however, these taxa have yet to be examined within a phylogeographic framework and it remains unclear whether genetic structure is indicative of a common biogeographic history. Paleodistributional modeling of the Atlantic forest biome during the last glacial maximum (21 ka years before present) indicates a possible westward expansion of similar environmental conditions, that may have formed a patchy corridor extending towards the base of the Andes (Carnaval and Moritz, 2008). Although these models are in disagreement with regional pollen records that suggest dry cerrado forest and savannah habitats likely dominated much of the area directly adjacent to Atlantic forest during periods of Pleistocene cooling, the paleodistributional history of these picid taxa merit investigation within a comparative phylogeographic framework to examine the potential role of Quaternary climate change in shaping patterns of lineage origin in these groups, as well as other avian lineages that share this pattern. 4.5. Nomenclatural summary Additional nuclear loci and population-level sampling are needed within several taxa before definitive taxonomic resolution can be reached. The contact zones between lugubris–elegans and grammicus–undatus require in depth analysis to rule out incomplete lineage sorting in the former and clarify the species status and potential introgression in the latter taxon pair. Several distinctive subspecies including C. t. tinnunculus, Celeus loricatus innotatus, Celeus elegans leotaudi, Celeus elegans deltanus, and C. f. intercedens have yet to be sampled. Broader geographic sampling and behavioral data are required within the various forms of C. ochraceus to understand the nature of this variability, and more extensive sampling within the wide latitudinal distribution of C. castaneus is needed to clarify its regional phylogeographic history. Lastly, the taxonomic position of Dryocopus galeatus requires molecular phylogenetic investigation given its peculiar mix of phenotypic and behavioral traits shared by several Celeus taxa (Short, 1982). We recommend the following species treatments based on morphology, vocalizations, and genetic data presented herein. Cinnamon woodpecker (C. loricatus). Nicaragua to southwestern Ecuador. Ringed woodpecker (C. torquatus). Amazonia and southeastern Brazil. Rufous-headed woodpecker (C. spectabilis). Western Amazonia. Kaempfer’s woodpecker (C. obrieni). East-central Brazil. Cream-colored woodpecker ( C. flavus). Amazonia and southeastern Brazil. Chestnut-colored woodpecker (C. castaneus). Mexico to western Panama. Scaly-breasted woodpecker (C. grammicus). Western Amazonia. Waved woodpecker (C. undatus). Guianan Shield and western Amazonia. Ochre-backed woodpecker (C. ochraceus). Eastern Brazil. Blond-crested woodpecker (C. flavescens). Eastern Brazil, eastern Paraguay, and northeastern Argentina.

43

Pale-crested woodpecker (C. lugubris). Southern Brazil, eastern Bolivia, Paraguay, and northeastern Argentina. Chestnut woodpecker ( C. elegans). Guianan Shield and Amazonia.

Acknowledgments Tissue samples were kindly provided by the Academy of Natural Sciences Philadelphia (ANSP), the American Museum of Natural History (AMNH), the Field Museum of Natural History (FMNH), the National Museum of Natural History, Smithsonian Institution (USNM), Louisiana State University Museum of Natural Science (LSUMNS), the Museo de Zoologia, Facultad de Ciencias, Universidad Nacional Autonoma de Mexico (UNAM), and the University of Kansas Natural History Museum (KUMNH). To the many field collectors that made this research possible, we are most appreciative of your efforts. We are grateful to Paul Sweet and curators of the American Museum of Natural History as well as Dave Willard and curators of the Field Museum of Natural History for providing ancient DNA samples from study skins of several key taxa. We thank Greg Budney and Martha Fischer of the Macaulay Library for facilitating access to recordings. Curtis Marantz, Davis Finch, Kevin Zimmer, and Steve Hilty kindly provided recordings and significant insight into Celeus vocalizations. We are grateful to A.T. Peterson for his generous support of this research through a University of Kansas General Research Fund. This article benefited from constructive comments by Vítor de Q. Piacentini, Robert G. Moyle, and one anonymous reviewer. References Baptista, L.F., Kroodsma, D.E., 2001. Avian bioacoustics, a tribute to Luis Baptista. In: Del Hoyo, J., Elliott, A., Sargatal, J. (Eds.), Handbook of the Birds of the World, vol. 6. Lynx Editions, Barcelona, pp. 11–52. Benz, B.W., Robbins, M.B., Peterson, A.T., 2006. Evolutionary history of woodpeckers and allies (Aves: Picidae): placing key taxa on the phylogenetic tree. Mol. Phylogenet. Evol. 40, 389–399. BirdLife International, 2010. Species Factsheet: Celeus obrieni. . Carnaval, A.C., Moritz, C., 2008. Historical climate modeling predicts patterns of current biodiversity in the Brazilian Atlantic forest. J. Biogeogr. 35, 1187–1201. Charif, R.A., Waack, A.M., Strickman, L.M., 2008. Raven Pro 1.3 User’s Manual. Cornell Laboratory of Ornithology, Ithaca, NY. Chesser, R.T., 1999. Molecular systematics of the rhinocryptid genus Pteroptochos. Condor 101, 439–446. Coates, A.G., Obando, J.A., 1996. Geological evolution of the Central American Isthmus. In: Jackson, J.B.C., Coates, A.G., Budd, A. (Eds.), Evolution and Environment in Tropical America. University of Chicago Press, Chicago, IL, pp. 21–56. Fleischer, R.C., Kirchman, J.J., Dumbacher, J.P., Bevier, L., Dove, C., Rotzel, N.C., Edwards, S.V., Lammertink, M., Miglia, K.J., Moore, W.S., 2006. Mid-Pleistocene divergence of Cuban and North American ivory-billed woodpeckers. Biol. Lett. 2, 466–469. Fuchs, J., Ohlson, J.I., Ericson, P.G.P., Pasquet, E., 2007. Synchronous intercontinental splits between assemblages of woodpeckers suggested by molecular data. Zool. Scrip. 36, 11–25. Fuchs, J., Pons, J.M., Ericson, P.G.P., Bonillo, C., Couloux, A., Pasquet, E., 2008. Molecular support for a rapid cladogenesis of the woodpecker clade Malarpicini, with further insights into the genus Picus (Piciformes: Picinae). Mol. Phylogenet. Evol. 48, 34–46. Guindon, S., Gascuel, O., 2003. A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst. Biol. 52, 696–704. Hilty, S.L., 2003. Birds of Venezuela, second ed. Princeton Univ. Press, Princeton, New Jersey. Hofreiter, M., Jaenicke, V., Serre, D., Haeseler, A.V., Paabo, S., 2001. DNA sequences from multiple amplifications reveal artifacts induced by cytosine deamination in ancient DNA. Nucleic Acids Res. 29, 4793–4799. Jahn, O., Moore, J.V., Valenzuela, P.M., Krabbe, N., Coopmans, P., Lysinger, M., Ridgely, R.S., 2002. The Birds of Northwest Ecuador. The Lowlands and Lower Foothills, vol. 2. J.V. Moore Nature Recordings. Johnson, K.J., Sorenson, M.D., 1998. Comparing molecular evolution in two mitochondrial protein coding genes (Cytochrome b and ND2) in the dabbling ducks (Tribe: Anatini). Mol. Phylogenet. Evol. 10, 82–94.

44

B.W. Benz, M.B. Robbins / Molecular Phylogenetics and Evolution 61 (2011) 29–44

Kimball, R.T., Braun, E.L., Barker, F.K., Bowie, R.C.K., Braun, M.J., Chojnowski, J.L., Hackett, S.J., Han, K.-L., Harshman, J., Heimer-Torres, V., Holznagel, W., Huddleston, C.J., Marks, B.D., Miglia, K.J., Moore, W.S., Reddy, S., Sheldon, F.H., Smith, J.V., Witt, C.C., Yuri, T., 2009. A well-tested set of primers to amplify regions spread across the avian genome. Mol. Phylogenet. Evol. 50, 654–660. Kirby, M.X., Jones, D.S., MacFadden, B.J., 2008. Lower Miocene stratigraphy along the Panama Canal and its bearing on the Central American Peninsula. PloS ONE 3, e2791. Kratter, A.W., 1997. Bamboo specialization by Amazonian birds. Biotropica 29, 100– 110. Lorenz, K.Z., 1950. The comparative method in studying innate behavior patterns. Symp. Soc. Exp. Biol. 4, 221–268. Maddison, W.P., Maddison, D.R., 2009. Mesquite: a Modular System for Evolutionary Analysis. Version 2.72. . Mindell, D.P., Sorenson, M.D., Dimcheff, D.E., 1998. An extra nucleotide is not translated in mitochondrial ND3 of some birds and turtles. Mol. Bio. Evol. 15, 1568–1571. Moore, J.V., 1997. Ecuador. More Bird Vocalizations from the Lowland Rainforest, vol. 3. J.V. Moore, Nature Recordings, San Jose, California. Moore, W., Weibel, A.C., Agius, A., 2006. Mitochondrial DNA phylogeny of the woodpecker genus Veniliornis (Picidae, Picinae) and related genera implies convergent evolution of plumage patterns. Biol. J. Linn. Soc. 87, 611–624. Nylander, J.A.A., Wilgenbusch, J.C., Warren, D.L., Swofford, D.L., 2008. AWTY (are we there yet?): a system for graphical exploration of MCMC convergence in Bayesian phylogenetics. Bioinformatics 24, 581–583. Peters, J.L., 1948. Check-list of Birds of the World. Museum of Comparative Zoology, vol. 6. Harvard University Press, Cambridge, Massachusetts. Pinheiro, R.T., Dornas, T., 2008. New records and distribution of Kaempfer’s woodpecker Celeus obrieni. Revista Brasileira de Ornitologia. 16, 167–169. Podos, J., Huber, S.K., Taft, B., 2004. Bird song: the interface of evolution and mechanism. Annu. Rev. Ecol. Evol. Syst. 35, 55–87. Posada, D., 2008. jModelTest: phylogenetic model averaging. Mol. Biol. Evol. 25, 1253–1256. Posada, D., Crandall, K.A., 2001. Selecting the best-fit model of nucleotide substitution. Syst. Biol. 50, 580–601. Prychitko, T.M., Moore, W.S., 1997. The utility of DNA sequences of an intron from the b-fibrinogen gene in phylogenetic analysis of woodpeckers (Aves: Picidae). Mol. Phylogenet. Evol. 8, 193–204. Rambaut, A., Drummond, A.J., 2007. Tracer v1.4. .

Ridgely, R.S., Greenfield, P.J., 2001. The Birds of Ecuador. Status, Distribution, and Taxonomy. Cornell University Press, Ithaca, New York. Ronquist, F., Huelsenbeck, J.P., 2003. MrBayes3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19, 1572–1574. Sefc, K.M., Payne, R.B., Sorenson, M.D., 2006. Single base errors in PCR products from avian museum specimens and their effect on estimates of historical genetic diversity. Conserv. Genet. 8 (4), 879–884. Short, L.L., 1972. Relationships among the four species of the superspecies Celeus elegans (Aves, Picidae). Amer. Mus. Novitates 2487, 1–26. Short, L.L., 1973. A new race of Celeus spectabilis from eastern Brazil. Wilson Bull. 85, 465–467. Short, L.L., 1982. Woodpeckers of the World. Delaware Museum of Natural History. Greenville, Delaware. Smith, B.T., Klicka, J., 2010. The profound influence of the late Pliocene Panamanian uplift on the exchange, diversification, and distribution of New World birds. Ecography 33, 333–342. Sorenson, M.D., Ast, J.C., Dimcheff, D.E., Yuri, T., Mindell, D.P., 1999. Primers for a PCR-based approach to mitochondrial genome sequencing in birds and other vertebrates. Mol. Phylogenet. Evol. 12, 105–114. Stiles, F.G., Skutch, A.F., 1989. A Guide to the Birds of Costa Rica. Cornell University Press, Ithaca, New York. Swofford, D.L., 2002. PAUP⁄: Phylogenetic Analysis Using Parsimony ( and other methods). Ver 4.0b10. Sinauer Associates, Sunderland, Massachusetts. Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins, D.G., 1997. The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 25, 4876–4882. Webb, D.M., Moore, W.S., 2005. A phylogenetic analysis of woodpeckers and their allies using 12S, Cyt b, and COI nucleotide sequences. Mol. Phylogenet. Evol. 36, 233–248. Weibel, A.C., Moore, W.S., 2002. Molecular phylogeny of a cosmopolitan group of woodpeckers (genus Picoides) based on COI and cyt b mitochondrial gene sequences. Mol. Phylogenet. Evol. 22, 65–75. Wiens, J.J., 2003. Missing data, incomplete taxa, and phylogenetic accuracy. Syst. Biol. 52 (4), 528–538. Winkler, H., Christie, D.A., 2002. Family Picidae (woodpeckers). In: del Hoyo, J., Elliot, A., Sargatal, J. (Eds.), Handbook of the Birds of the World, Jacamars to Woodpeckers, vol. 7. Lynx Editions, Barcelona. Zwickl, D.J., 2006. Genetic Algorithm Approaches for the Phylogenetic Analysis of Large Biological Sequence Datasets under the Maximum Likelihood Criterion. Ph.D. Dissertation, The University of Texas at Austin.