MOLECULAR PHYLOGENETICS AND EVOLUTION Molecular Phylogenetics and Evolution 31 (2004) 943–960 www.elsevier.com/locate/ympev Phylogeny and evolution of the Australo-Papuan honeyeaters (Passeriformes, Meliphagidae) Amy C. Driskella,* and Les Christidisb b a Division of Birds, Field Museum, 1400 S. Lake Shore Dr., Chicago, IL, 60605, USA Department of Sciences, Museum Victoria, GPO Box 666E, Melbourne, Victoria 3001, Australia Received 8 April 2003; revised 10 October 2003 Abstract We analyzed nucleotide variation at four loci for 75 species to produce a phylogenetic hypothesis for the Meliphagidae, and to examine the evolution and biogeographic history of the Meliphagidae. Both maximum parsimony and Bayesian methods of phylogenetic analysis were employed. The family was found to be monophyletic, though the genera Certhionyx, Anthochaera, and Phylidonyris were not. Four major clades were recovered and the spinebills (Acanthorhynchus) formed the sister clade to the remainder of the family in most analyses. The Australian endemic arid-adapted chats (Epthianura, Ashbyia) were found to be nested deeply within the family Meliphagidae. No evidence was found to support the hypothesis of separate New Guinean and Australian endemic radiations, nor of a close phylogenetic relationship between taxa from the New Guinea highlands and those from Australian northern rainforests. Ó 2003 Elsevier Inc. All rights reserved. Keywords: Phylogenetic analysis; Avian systematics; Biogeography; Meliphagidae 1. Introduction One of the dominant groups of birds in Australia and New Guinea, both numerically and ecologically, is the passerine family Meliphagidae, the honeyeaters. In certain habitats more than 12 species of honeyeater can co-occur seasonally (Keast, 1985). Although the family has its centers of diversity in Australia and New Guinea, meliphagids are also important endemic elements of the biota of many of the islands in the south Pacific. Honeyeaters are diverse in size, morphology, and diet. They can be either nectarivorous, insectivorous, frugivorous, or more commonly, a combination of nectar- and insect-eating. Although many species have long, narrow, decurved bills, presumably adapted for nectar-feeding, these species are often insectivorous * Corresponding author. Present address: Center for Population Biology, 1 Shields Ave., University of California, Davis, CA 95616, USA. Fax: 1-530-752-1449. E-mail address: acdriskell@ucdavis.edu (A.C. Driskell). 1055-7903/$ - see front matter Ó 2003 Elsevier Inc. All rights reserved. doi:10.1016/j.ympev.2003.10.017 during certain seasons of the year (Lea and Gray, 1935; Rand and Gilliard, 1968). In Australia, honeyeaters are major pollinators of many endemic plant groups, including Banksia, Dryandra, Melaleuca, Hakea, and Eucalyptus (Paton and Ford, 1977; Recher, 1981). The Meliphagidae are an ecologically and evolutionarily significant element of the Australo-Papuan fauna and yet phylogenetic relationships within this large family are almost entirely unknown. Traditionally the Meliphagidae were linked with the Nectariinidae (sunbirds) and other nectarivorous birds (Cracraft, 1981; Wetmore, 1960). DNA–DNA hybridization studies (Sibley and Ahlquist, 1985, 1990) and allozyme evidence (Christidis, 1991; Christidis and Schodde, 1991) demonstrated that the honeyeaters belong to a clade originating in the Australo-Papuan region and composed of the Meliphagidae, the Pardalotidae–Acanthizidae (Australasian warblers and allies), and Maluridae (Australasian fairy-wrens and grasswrens). Mitochondrial and nuclear sequence data (Barker et al., 2002; Cracraft and Feinstein, 2000; Ericson et al., 2002a,b) confirm close affinities between the Meliphagidae, Pardalotidae–Acanthizidae, and Maluridae. 944 A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 Relationships within the Meliphagidae are poorly understood. There have been no phylogenetic studies of the entire family, but a few studies have examined relationships among some genera (Christidis and Schodde, 1993; Christidis et al., 1993). Published taxonomies for the Meliphagidae lack classification levels between family and genus: no subfamilies or tribes have been proposed (Christidis and Boles, 1994; Schodde, 1975; Schodde and Mason, 1999; Sibley and Monroe, 1990). The number of monotypic genera and the morphological distinctness of most genera have been cited as impediments to determining interrelationships within the family (Schodde, 1975; Schodde and Mason, 1999). In the last decade, molecular and biochemical studies have modified the traditional composition of the family Meliphagidae. Sibley and Ahlquist (1990) demonstrated that the South African sugarbirds Promerops and New Guinean longbills Oedistoma and Toxorhamphus were not honeyeaters but more closely related to the Nectariniidae (Promerops) and Melanocharitidae (Oedistoma and Toxorhamphus). Sibley and Ahlquist also showed that the Australian chats, Epthianura and Ashbyia (formerly the Epthianuridae), were honeyeaters, a result which was supported by allozyme data (Christidis et al., 1993). In addition, DNA–DNA hybridization (Sibley and Ahlquist, 1990) and DNA sequence (Slikas et al., 2000) studies established that south Pacific Cleptornis was not a honeyeater but instead was closely related to the white-eyes (Zosteropidae). The genus Apalopteron from Bonin Island has likewise been shown to be a white-eye rather than a honeyeater, based on DNA sequence data (Springer et al., 1995). Another genus, Macgregoria, traditionally classified as a bird-ofparadise, was shown to be a honeyeater based on combined analyses of DNA sequence and morphological data (Cracraft and Feinstein, 2000). Modifying the taxonomic review of the Meliphagidae by Sibley and Monroe (1990) with the recent changes described above, the family now comprises 182 species in 42 genera. Two of these genera (Moho and Chaetoptila) are extinct on Hawaii (Pratt et al., 1987). Australia has over 70 species of honeyeaters (Christidis and Boles, 1994), and New Guinea over 60 species (Beehler et al., 1986). A few genera are distributed across the Lesser Sunda Islands, the Moluccas and Sulawesi. One species of honeyeater, Lichmera limbata, crosses WallaceÕs line, but occurs only as far west as Bali (Coates and Bishop, 1997). In the south Pacific, honeyeaters are distributed northwards from New Guinea to the Mariana Islands, as far south as New Zealand, and east to Hawaii (Pratt et al., 1987). The main goal of this paper is to introduce a phylogenetic hypothesis for the family Meliphagidae. More specific goals are to: (1) examine systematic relationships of some taxonomically unstable genera (e.g., Certhionyx, Phylidonyris, and Meliphaga sensu lato); (2) de- termine the phylogenetic relationship of the Australian chats (Epthianura, Ashbyia), which are remarkable among the Meliphagidae for their adaptation to arid habitats; and (3) examine the biogeographical history of the family, especially the relationships among the New Guinean and Australian honeyeater faunas. In addition, we wanted to explore the utility of three different categories of genetic loci (mitochondrial protein-coding genes, mitochondrial ribosomal DNA, and a nuclear intron) for reconstructing phylogenetic relationships in such a large, divergent, and relatively old family of passerines. 2. Methods 2.1. Taxon sampling, DNA extraction, amplification, and sequencing The sample of the family Meliphagidae consisted of 63 species, representing 32 of the 41 described meliphagid genera listed in Sibley and Monroe (1990). To test for intraspecific sequence variation, multiple individuals were sampled for 10 species. Nine species, representing five genera, of the Pardalotidae (sensu Sibley and Monroe, 1990), the apparent sister group of the Meliphagidae (Christidis and Schodde, 1991; Sibley and Ahlquist, 1985), were also sampled. To establish monophyly of the ingroup (Meliphagidae) and the sister relationship of the Meliphagidae and Pardalotidae, four species of the Maluridae, generally considered the next-most closely allied family, were also included. Specimens, collection localities, and voucher specimen locations and numbers are listed in Appendix A. DNA was extracted from ethanol-preserved tissue. Blood was used as a DNA source only for the New Zealand Tui, Prosthemadera novaeseelandiae. DNA was obtained using standard proteinase-k digestion followed by either phenol–chloroform extraction or the protein precipitation method as implemented in the Puregene kit (Gentra Systems). Standard PCR amplification and automated sequencing techniques were used to sequence part or all of one nuclear gene, b-fibrinogen intron 5 (FIB5), and three mitochondrial genes: cytochrome-b (CYTB), 12S rDNA (12S), and NADH dehydrogenase subunit 2 (ND2). Both strands of all gene regions were sequenced and portions of most were sequenced multiple times. Each resequencing essay was preceded by reamplification from the original DNA extract to provide additional insurance against lab errors. Primers used for amplification and sequencing are listed in Table 1. 2.2. Data verification and sequence alignment Although it is impossible to completely guard against the amplification of nuclear copies of mitochondrial A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 945 Table 1 Primers used for amplification and sequencing in this study Gene Primer Sequence Reference Cytochrome b L14990 L15191 H15298 L15656 H15916 H16065 CCATCCAACATCTCAGCATGATGAAA ATCTGCATCTACCTACACATCGG CCCCTCAGAATGATATTTGTCCTCA ACCTACTAGGAGACCCAGA ATGAAGGGATGTTCTACTGGTTG GGAGTCTTCAGTCTCTGGTTTACAAGAC Kocher et al. (1989); modified L14841 Lanyon and Hall (1994) Kocher et al. (1989); modified H15149 Helm-Bychowski and Cracraft (1993) Lanyon and Hall (1994) Helm-Bychowski and Cracraft (1993) NADH dehydrog. subunit 2 L5206 L5575 H5802 H6313 CTAATAAAGCTTTCGGGCCCATAC AAACTAGGACTAGTGCCATTCCA GAGAATAATGGTTATTCATCC TTCTACTTAAGGCTTTGAAGGC Kirchman et al. (2001); L5208 S.J. Hackett, unpublished Kirchman et al. (2001); H5804 Kirchman et al. (2001); H6315 12S rDNA L1276 L1746 H1811 H2512 CACTGAAGATGCCAAGATGG GCTTCAAACTGGGATTAGATACC CCTTAGAGTTTAAGCGTTTGTGC GCAGAGGGTGACGGGCGGTGTGT This study Kocher et al. (1989); modified L1091 This study Kocher et al. (1989); modified H1478 b-fibrinogen intron 5 FIB5 FIB6 CGCCATACAGAGTATACTGTGACA GCCATCCTGGCGATTCTGAA F.K. Barker and S.J. Hackett, unpublished F.K. Barker and S.J. Hackett, unpublished Each is listed in 50 to 30 orientation, numbered with reference to the complete mtDNA sequence of the chicken (Desjardins and Morais, 1990), and designated as H or L with respect to their location on the heavy or light strand of the mitochondrial genome (published source in parentheses, along with original primer name if different than numbered designation). gene regions (numts) and it can be impossible to identify them in a data set (reviewed by Quinn, 1997), a number of precautions were taken. Only one of the tissue samples used was blood, which previous avian studies have shown to be particularly prone to numt contamination (Quinn, 1997). Larger fragments of target genes were amplified whenever possible, except in the case of 12S, which was amplified in two segments. Each data partition was inspected for taxa with significantly different base compositions using the ‘‘pairwise base differences’’ option PAUP* v. 4.0b10 (Swofford, 2002). The sequence data for both protein coding genes for each species was aligned to published sequences, translated and checked for termination codons, insertions, and deletions. In addition, the alignment of the 12S data to a structural model (below) allowed recognition of loss of stem complementarity or conserved binding motifs, which might be expected in a non-functional nuclear copy (Houde et al., 1997). The absence of termination codons and deletions in the protein coding genes, and lack of suspicious deletions in stems or other aspects of the 12S data allows moderate confidence that amplification and sequencing of nuclear pseudogene copies of these genes was successfully avoided. Very low levels of intraspecific sequence variation (<1.0% in CYTB and ND2, and 0.0% in 12S and FIB5) were observed. Allelic variation within individuals was, however, apparent in the nuclear FIB5 data set. Differences of 1–12 bp between the two alleles of the intron were observed in 38 of the specimens sequenced (22 of these had 1 bp difference between the two alleles), and the majority of these base differences (66%) were transitions. These differences were observed as double peaks when directly sequencing PCR products. The double peaks were verified as the consequence of different alleles by cloning PCR products using a topo-TA cloning kit (Stratagene) and sequencing multiple (up to 10) clones. In addition to sequence differences, three individuals exhibited alleles varying in length by 5–12 bp (manifested as single insertion–deletion events). PCR products of these individuals were cloned. Allelic differences in nucleotide sequence were represented in the alignment used for phylogenetic analysis by ambiguity codes. However, in every instance the two alleles from an individual were more similar in sequence to each other than to those from any other taxon. For specimens whose alleles varied in length, both alleles were included in an initial phylogenetic analysis. In all instances, the two alleles from a single individual grouped together with high bootstrap support, and in later analyses only the allelic sequence without the unique insertion or deletion was used to represent the taxon. All output from the automated sequencer was checked for accuracy in base identification. Sequences of a gene from a specimen were compiled using the program Sequencher v. 3.1 (Genecodes 1991–1998) and then imported into PAUP* (Swofford, 2002) for alignment. Sequences of the b-fibrinogen intron were aligned by eye. Within this data set, and within each of the families in this study, insertions and deletions (indels) ranging in size from 1 to 47 bp in length were present. Aside from those involving only one base, most insertions in the FIB5 data set appeared to involve repeated motifs. The inserted regions were copies (sometimes multiple copies) of flanking sequence and identification and alignment 946 A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 was relatively straightforward. Two regions containing insertions and any ambiguously aligned flanking sites (amounting 81 bp of the overall alignment) were excluded from phylogenetic analyses. One of these regions (2 bp) was exclusive to two outgroup congeners. The other excluded region is the result of at least four and probably more indel events and exact homology among nucleotides was difficult to assess. As a conservative measure, this entire region was excluded. All remaining gaps in the alignment were treated as missing data for phylogenetic analysis. Twelve of these indel gaps were coded as binary characters. To create amore general avian 12S model to serve as a template for alignment of our 12S data, published 12S sequences for the chicken (Desjardins and Morais, 1990) and falcon (Mindell et al., 1997) were aligned according to a model of secondary structure published for the falcon (Mindell et al., 1997). Use of the model simplified identification of the stem regions flanking each loop, and isolated all of the length variation to loop regions. All ambiguously aligned regions, which amounted to 11 of 62 variable loops, were excluded from phylogenetic analysis. A total of 114 bp of the total 12S sequence (12.5%) was excluded in this fashion. 2.3. Molecular characterization and phylogenetic analysis Saturation was assessed graphically by plotting transition and transversion differences between pairs of taxa against their Jukes–Cantor distance (Jukes and Cantor, 1969). Graphs were produced separately for each codon position in the two protein coding genes, for stem and loop regions in 12S, and for FIB5. A data partition was considered to be saturated if one of two criteria was met (Griffiths, 1997): (1) a levelling-off or plateau of the plotted data was apparent, or (2) most ingroup comparisons were as great as outgroup comparisons. To assess the homogeneity of phylogenetic signal contained within different data partitions, incongruence length difference tests (ILD tests; Farris et al., 1995a,b), as implemented in PAUP* (Swofford, 2002), were conducted. Invariant sites were removed from the data sets before analysis as recommended by Cunningham (1997) and tests used 1000 replicates. The mitochondrial and fibrinogen data sets were subjected to both separate and combined heuristic searches under the parsimony criterion using PAUP* vers. 4.0b10 (Swofford, 2002); these searches employed 100 random addition sequences (RAS) and tree-bisection-reconnection (TBR) branch swapping. Each random addition sequence in the separate FIB5 analysis was limited to 20 min of branch swapping. The nuclear data set provided little or no resolution at very shallow nodes and excessive time was spent branch-swapping essentially unresolvable nodes. As a means of exploring signal and noise in the mitochondrial data set, analyses were performed with all sites weighted equally, and with saturated data partitions downweighted relative to other partitions through the implementation of step matrices. Bootstrap analysis with the full heuristic option, 10 RAS, and TBR branch swapping was used to evaluate nodal support. An estimate of the maximum likelihood topology of the combined data set was produced using Bayesian analysis as implemented by MRBAYES (Huelsenbeck, 2000). The program MODELTEST 3.0 (Posada and Crandall, 1998), which performs a series of likelihood ratio tests (Huelsenbeck and Crandall, 1997; Huelsenbeck and Rannala, 1997), was used to explore the bestfit maximum likelihood models for each of the four data partitions. Eight independent analyses were performed, each with 1,000,000 generations with four simultaneous chains. Trees were sampled every 1000 generations. Likelihood scores for each tree were plotted against generation and visually inspected to ascertain the point at which the chain appeared to reach stationarity. Trees from the generations preceding stationarity were discarded in each analysis. The remaining trees from each run were combined and posterior probabilities were calculated using majority rules consensus (Larget and Simon, 1999). Comparing the likelihood scores for each topology from each generation identified the most likely topology from each run. The combined data set was also subjected to a parsimony search constrained to contain a clade found in the separate FIB5 consensus topology (Fig. 2). The constraint required only that 10 taxa form a monophyletic group. The particulars of this parsimony and bootstrap analysis were the same as above. Parametric bootstrapping was conducted to test whether the difference in tree length between the unconstrained and constrained topologies was statistically significant (Goldman et al., 2000; Swofford et al., 1996). MESQUITE 1.0 (Maddison and Maddison, 2003) was used to simulate 500 data sets on each of the two equally parsimonious topologies resulting from the constrained search. Data was simulated using a GTR + I + C model with the parameters estimated by MODELTEST 3.0 (Posada and Crandall, 1998). Each of the 1000 data sets was subjected to a constrained parsimony search (employing the same constraint as above) and an unconstrained search using PAUP* with 20 RAS and TBR branch-swapping. MESQUITE was then employed to calculate the tree length differences for each test data set, construct a distribution of tree length differences, and to estimate the critical value for a significance level a ¼ 0:05. The difference in tree length from constrained and unconstrained searches of the actual data set was compared to this critical value and an assessment of the statistical significance of the difference was made. 947 A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 3. Results The total alignment of sequences (including gaps and indels) from CYTB (1046 bp), ND2 (1040 bp), 12S (910 bp), and FIB5 (547 bp) was 3843 base pairs (alignments are available from the first authorÕs website; see Acknowledgements). All sequences were deposited in GenBank under Accession Nos. AY353241, AY353242, AY488184–AY488485. The aligned nexus file and associated trees have been submitted to TreeBASE (http:// www.treebase.org/treebase) and can also be downloaded from the authorÕs website (see Acknowledgments). 3.1. Molecular characterization Total length and levels of variation for the four gene regions are reported in Table 2. The ND2 gene had the highest proportion of variable positions (62%), with 91% of these being potentially parsimony-informative. The sequences of both CYTB and ND2 included in analyses were of similar length, and nearly all third positions of both genes were variable (97 and 99%, respectively). However, considerably more first and second positions were variable in the ND2 gene. Of the 910 bp of 12S included for analysis, only 36% were variable, and loop regions were more variable than stem regions. Of the variable positions of the FIB5 intron only 55% were potentially parsimony-informative, which indicates a high proportion of autapomorphic changes in this gene region. Although the two protein coding genes had the highest proportions of potentially parsimony informative characters, nearly 80% of these characters had a consistency index (CI) lower than 0.5 (Table 2), indicating a relatively high level of homoplasy in these data. The most variable partitions of these genes, third positions, had even higher levels of homoplasy (87% of CYTB third positions and 97% of ND2 third positions had CI < 0.5). In contrast, the majority of potentially informative FIB5 characters had CI ¼ 1.00, denoting a relatively low level of homoplasy. Within the Meliphagidae, uncorrected CYTB sequence divergences ranged from 2.7 to 19% and ND2 sequence divergences ranged from 3.8 to 30%. Divergences at the 12S and FIB5 loci were much lower and similar to each other, ranging from less than 1% to about 10%. Twelve FIB5 indels were coded as binary characters (Tables 3a and 3b). Only six of these characters were included in parsimony analyses of the individual FIB5 nucleotide data and the combined data set. The remaining six were exclusive to the outgroup taxa. Inclusion of the six binary indel characters had no effect on the topology resulting from parsimony analysis, but when mapped onto a topology (Figs. 2 and 4) provide strong corroborative evidence for particular nodes (see below). Based on saturation plots (available from the first authorÕs website; see Acknowledgements), transitions at third codon positions of ND2 and CYTB and at the first codon positions of ND2 showed evidence of leveling off at higher divergences. In these three instances, many of the ingroup comparisons were as great as or greater than comparisons between the ingroup and outgroup. These two signs were taken as evidence of saturation at higher levels of divergence as per Griffiths, 1997. There is no clear evidence for saturation in any of the remaining data partitions; in no instance do transversion Table 2 Comparative variability of genes and partitions within genes and consistency indices for characters in each gene and partition Gene Partition Number of basepairs # variable characters (%) # parsimony inform. characters (%)a % with CI < 0.5 % with CI ¼ 1.00 FIB5 547 304 (56) 168 (55) 12 66 12S Stem regions Loop regions 910 457 336 331 (36) 212 (27) 118 (35) 255 (77) 85 (83) 86 (73) 48 45 52 36 40 32 CYTB 1st positions 2nd positions 3rd positions 1046 348 349 349 508 127 43 338 (49) (36) (12) (97) 447 95 24 328 (88) (75) (56) (97) 75 54 37 87 15 32 49 4 ND2 1st positions 2nd positions 3rd positions 1040 347 347 346 646 197 106 343 (62) (57) (30) (99) 585 166 78 341 (91) (84) (73) (99) 79 65 50 97 13 22 35
1 The total number of basepairs in each gene and partition, the number of variable characters (proportion of characters variable), number of parsimony informative characters (proportion of variable characters), proportion of characters with consistency indices less than 0.5, and the proportion with consistency indices of 1.0 are listed. For the calculation of consistency indices, invariant and autapomorphic characters were excluded, and characters were mapped onto the consensus topologies resulting from parsimony analysis of each gene. FIB5, b-fibrinogen intron 5; 12S, 12S ribosomal DNA; CYTB, cytochrome-b; ND2, NADH dehydrogenase subunit 2; CI, consistency index; and
1, less than 1%. a proportion of variable characters that are also parsimony informative. 948 A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 Table 3a Description of b-fibrinogen intron 5 indels coded as binary characters Char. Sequence position # b.p. Relative type Taxa (number of species) 1. 2. 42–44 110–127 3 18 Deletion Insertion 3. 4. 175 197–243 1 var. Deletion Insertion 5. 6. NI NI NI NI NI NI 247–254 280 183 309 336–343 348–349 522–525 648–664 8 1 1 1 8 2 4 17 Deletion Deletion Deletion Deletion Deletion Deletion Deletion Insertion C. niger, G. fallax, Myzomela (5 spp.), P. melanops, Ptiloprora (2 spp.) G. fallax, C. niger, Myzomela (5 spp.), P. melanops, Ptiloprora (2 spp.), Pardalotidae (7 spp.), Maluridae (4 spp.) Philemon (5 spp.) E. cyanotis, 11 bp; Foulehaio carunculata, 36 bp; Melithreptus albogularis, 47 bp; M. brevirostris, 33 bp; see Table 3b Myzomela (5 spp.) Certhionyx niger, Myzomela (5 spp.) Maluridae (4 spp.) Maluridae (4 spp.) Acanthiza (2 spp.), Sericornis (2 spp.) Maluridae (4 spp.) Malurus (2 spp.) Acanthiza (2 spp.) For each indel, the position in the sequence alignment, size, relative type, and taxa in which it appears are listed. The sequence position is numbered from the 30 end of the primer FIB5. The relative type depends on whether the insertion or deletion is most common (e.g., deletion is the relative type if fewer taxa possess it). Character numbers refer to the characters mapped onto Figs. 2 and 4. Char., character number; NI, not included for phylogenetic analysis; # b.p., indel length in number of basepairs; and var., of variable length. Table 3b Elaboration of indel character 4 from Table 3a Taxon Indel character 4 sequence Glycichaera fallax Epthianura aurifrons Entomyzon cyanotis Foulehaio carunculata Melipthreptus albifrons Melipthreptus brevirostris GCTCACACTT-----------------------------------------------AAATA GTTCACACTT-----------------------------------------------AAGTA GCTCACACTTAAGTACTACTT------------------------------------AAGTA GCTCACACTTAAGTACTACTTAAGTACTACTTAAGTACTAC-----------TACTTAAGTA GCTCACACTTAAGTACTACTTAAGTACTACTTAAGTACTACTTAAGTACTACTACTTAAGTA GCTCACACTTAAGTACTACTTAAGTACTACTTAAGTACTACTT--------------AAGTA Epthianura aurifrons and Glycichaera fallax are included to illustrate the common (uninserted) condition. The other four taxa possess inserted sequence of varying length. differences appear to saturate. Based on these plots, only transitions at the third positions of the protein coding genes and at first position in ND2 were considered saturated and weighting schemes were applied only to these partitions. 3.2. Phylogenetic analyses None of the ILD tests between mitochondrial genes produced significant results. The test between the FIB5 data and the combined mitochondrial genes resulted in a p ¼ 0:054. This would be considered borderline significant using a traditional a level of 0.05 (but see Cunningham, 1997), so potential incongruence between the mitochondrial and nuclear partitions was investigated by topological comparisons. There were no conflicting nodes in the 75% bootstrap trees resulting from separate phylogenetic analysis of the nuclear and mitochondrial data sets and this was interpreted as a lack of strongly supported conflict between the data sets. Unweighted parsimony analysis of the mitochondrial data set resulted in one most parsimonious tree (Fig. 1). The Meliphagidae form a strongly supported monophyletic group and all genera, with the exception of Anthochaera, Certhionyx, and Phylidonyris, are strongly supported as monophyletic. Although many of the deeper nodes in the tree have little or no bootstrap support, a number of novel relationships among genera are well-supported. Downweighting transitions in the three saturated partitions (CYTB and ND2 3rd positions and ND2 1st positions) resulted in no serious change in topology (results not shown): two unsupported nodes collapsed. However, downweighting did increase bootstrap (BS) support for a number of relatively deep nodes within the family. This increase indicates that homoplasy in the saturated partitions is compromising resolution at these nodes and downweighting homoplasious partitions can boost the signal in the remaining partitions. However, as no significant differences in topology obtained with downweighting, the combined analyses were unweighted. Parsimony analysis of the FIB5 data set resulted in 8290 equally parsimonious trees. The strict consensus of these trees (Fig. 2) has a well-supported monophyletic Meliphagidae and all genera, with the same exceptions as the mitochondrial data, are well-supported and monophyletic. Resolution at the tips of the tree is poorer with the FIB5 data than the mitochondrial data; this A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 949 Fig. 1. The most parsimonious tree resulting from unweighted analysis of the three mitochondrial loci (CYTB, ND2 and 12S). Bootstrap support values greater than 70% are shown. Letters are used to label nodes for comparison with the nuclear consensus topology (Fig. 2). result is not unexpected given the low levels of divergence in the nuclear data set. Although both data sets resolve many of the same groups of genera (labelled in Figs. 1 and 2), deeper nodes in the tree, corresponding to relationships among these well-supported groups of genera, are quite different. However, due to the poor resolution in the middle levels of the tree, the differences in resolution cannot be considered serious conflict. Analysis of the combined nuclear and mitochondrial data sets under the parsimony criterion produced two equally parsimonious trees (Fig. 3). The two topologies differed in the relative positions of the three major 950 A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 Fig. 2. A strict consensus of the 8290 equally parsimonious trees resulting from unweighted parsimony analysis of the nuclear FIB5 data set. Bootstrap support values greater than 70% are shown. Letters are used to label nodes for comparison with the mitochondrial topology (Fig. 1). Shaded boxes correspond to the most parsimonious mapping of the indel characters listed in Tables 3a and 3b. pardalotid lineages and not in any ingroup relationships. The topologies resulting from the combined analysis and the separate mitochondrial analysis are nearly identical and there are no well-supported differences. Therefore, the primary result of the addition of the FIB5 data to the mitochondrial data set was a general increase in many bootstrap support values, particularly those at deeper nodes within the Meliphagidae. For Bayesian analysis a model with six substitution classes, unequal base frequencies and rate heterogeneity was employed. The results of the MODELTEST analysis indicated that each of the partitions required a substantially different C shape parameter. Therefore, instead of modeling rate variation with a single parameter for all four partitions, site specific rates were calculated for each data partition separately. Based on A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 951 Fig. 3. The topology equivalent to the maximum a posteriori estimate resulting from Bayesian analysis and the consensus topology (of two equally parsimonious tres) resulting from parsimony analysis of the combined mitochondrial and nuclear data set. Bayesian posterior probabilities greater than 95% are shown below the nodes and bootstrap support values greater than 70% are shown above the nodes. The two nodes marked by asterisks are unresolved in the parsimony consensus topology and the three numbered nodes are referred to in the text. visual examination, all eight MCMC runs appeared to reach stationarity at approximately 30,000 generations. As a conservative measure, the trees from the first 50,000 generations were excluded from the calculation of posterior probabilities. The maximum a posteriori (MAP) estimates of the topology, which are point estimates of the maximum likelihood topology (Huelsenbeck and Bollback, 2001), were identical from all eight runs. Posterior probabilities (PP) of all nodes save two were >95% and the majority were 100% (Fig. 3). 952 A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 Fig. 4. A strict consensus of the two equally parsimonious trees resulting from a constrained parsimony search. The node numbered ‘‘3’’ was the constrained node. Although 14 steps longer, his topology was not found to be significantly different from the topology in Fig. 3 using parametric bootstrapping. Bootstrap support values greater than 70% are shown. Shaded boxes correspond to the most parsimonious mapping of the indel characters listed in Tables 3a and 3b. Numbered clades are referred to in the text. The MAP topology resulting from Bayesian analysis was nearly identical to the consensus topology resulting from the unweighted parsimony analysis. Where parsimony analysis was unable to resolve the phylogenetic positions of the three pardalotid clades, in the Bayesian topology these lineages were resolved as successive outgroups to the Meliphagidae, with high PP. The only instance of conflict between the parsimony consensus topology and the MAP topology was in the relative positions of three meliphagid lineages (labelled 1, 2, and A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 3 in Fig. 3). In the parsimony topology, the trichotomy was resolved as (1,3)2 with BS support of 62%. The MAP topology had another resolution of these clades: (1,2)3 with a PP of 61%. As neither method of analysis furnished a well-supported resolution of these three clades, relationships among them should be considered equivocal and unresolved. One strongly supported node in the FIB5 consensus tree unites Myzomela, Certhionyx niger, Glycichaera, Ptiloprora, and Phylidonyris melanops (clade G + H in Fig. 2, and hereafter referred to as the ‘‘nuclear’’ arrangement of the taxa) and is not present in either the separate mitochondrial tree (Fig. 1) or in the trees resulting from analysis of the combined data set (Fig. 3). Instead, in these other topologies Myzomela + C. niger are associated with a large clade of Philemon and its relatives, while Ptiloprora, Glycichaera, and Phylidonyris albifrons form a clade elsewhere in the tree (henceforth referred to as the ‘‘mitochondrial’’ arrangement of these taxa). The nuclear arrangement of these taxa is further bolstered by the presence of two indel characters: a 3-bp deletion shared by C. niger, P. melanops, Ptiloprora, Glycichaera, and Myzomela, and an 18-bp insertion shared by these taxa and both outgroup families (Tables 3a and 3b and Figs. 2 and 4). The constrained search of the combined data set produced two equally parsimonious trees (Fig. 4) which were 14 steps longer than the trees resulting from the unconstrained search. Parametric bootstrapping showed that the null hypothesis of the constrained topology could not be rejected by the data (the critical value was 29 steps). Therefore the constrained and unconstrained topologies are not significantly different. 4. Discussion 4.1. Comparative information content of the four loci We believe the proportion of parsimony informative characters is an insufficient measure for comparing the phylogenetic utility of different data partitions, or for determining whether a partition should be downweighted for analysis (contra Allard et al., 1999; Sennblad and Bremer, 2000). Among the four loci we sampled, the more ‘‘parsimony informative’’ characters contained in a partition, the lower the average CI of characters in that partition (Table 2). An extreme example of this is ND2 third positions, of which 99% are parsimony informative and 97% have CI less than 0.5. Thus, in these data, characters classed as parsimony informative are actually very homoplasious. In our data, FIB5 has a higher proportion of variable characters, but 12S has more phylogenetically informative characters (Table 2). Although FIB5 has more autapomorphic changes, the level of homoplasy in this 953 partition is much lower. In addition, a large amount of variation in the 12S data occurs in the form of indels in the loop regions of the molecule. These indels make many loop regions difficult to align unequivocally, and force the exclusion of these regions from analyses. In this process, a potential source of phylogenetic information is lost (Lutzoni et al., 2000). Despite the fact that much of the molecular evolution of the FIB5 intron also takes the form of indels, these indels are generally not as difficult to align as the indels in 12S. Furthermore, variation in the FIB5 indel regions is quite low, and many of indels themselves can be profitably coded and included in phylogenetic analysis, making the nuclear intron a more fruitful source of phylogenetic information than 12S. Although of a similar length, ND2 provides considerably more variable characters than CYTB for phylogenetic analysis (Table 2). This is a consequence of the higher variability of the first and second codon positions in ND2. Levels of homoplasy in the two protein coding genes appear similar, and therefore ND2 appears to have greater phylogenetic utility than CYTB. 4.2. Molecular systematics of the Meliphagidae The family Meliphagidae as constituted here is strongly supported as monophyletic. The genus Pardalotus may be more closely related to the honeyeaters than to other genera in the family Pardalotidae, but these relationships are not robustly resolved in our analyses. Certainly, phylogenetic relationships among pardalotid taxa require further study. Within the Meliphagidae, all genera for which we sampled more than one species, with the exceptions of Anthochaera, Phylidonyris, and Certhionyx, are monophyletic. The Regent Honeyeater, Xanthomyza phrygia, is nested within the wattlebird genus Anthochaera. The large-bodied wattlebirds (Anthochaera carunculata and Anthochaera paradoxa) are more closely related to X. phrygia than they are to the smaller-bodied wattlebirds (Anthochaera lunulata and Anthochaera chrysoptera). While this result is unexpected, monotypic Xanthomyza is behaviorally similar to the large wattlebirds (A. Keast, pers. comm.), while Schodde and McKean (1976) noted that the eggs of Xanthomyza have a similar color and pattern to those of Anthochaera. In addition, Veerman (1992) reported some of the vocalisations of Xanthomyza resembled those of Anthochaera. Although this was interpreted as evidence of vocal mimicry by Xanthomyza, it may in fact reflect the shared phylogenetic history of the two genera. Species composition of Phylidonyris and Certhionyx has traditionally been in flux and a number of authors have questioned the validity of the genera as currently described (reviewed in Christidis and Boles, 1994). In our phylogeny, both genera are polyphyletic. The three 954 A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 species comprising Certhionyx are not closely related to one another and belong to three different clades. The five species of Phylidonyris included in this study form three separate lineages. Phylidonyris novaehollandiae, Phylidonyris nigra, and Phylidonyris pyrrhoptera form a monophyletic group that is disassociated from P. melanops and P. albifrons, both of which appear in different clades. These five Phylidonyris species are restricted to Australia and the genus has no members in northern Australia or New Guinea (Sibley and Monroe, 1990). The remaining two species are restricted a few south Pacific islands: Phylidonyris notabilis to Vanuatu and southern Melanesia, and Phylidonyris undulata to New Caledonia. The unusual geographic distribution of Phylidonyris—present only in Australia and some south Pacific islands, but absent from New Guinea—is unique among passerines (Sibley and Monroe, 1990). As the five Australian species do not form a monophyletic group, it seems unlikely the two south Pacific species will prove to be closely related to their current congeners. Schodde (1975) split the large genus Meliphaga into three genera: Lichenostomus, Xanthotis, and Meliphaga, and this division was supported by allozyme data (Christidis and Schodde, 1993). Our results show that these three genera do not form a monophyletic group. Xanthotis, distributed in rainforests in northernmost Australia and in New Guinea, is more closely related to the friarbirds (Philemon) and their allies and not at all closely related to either Lichenostomus or Meliphaga. The latter two genera are two of the most species-rich honeyeater genera and are contained within the same clade, but not as each otherÕs closest relatives. However, many species in these two genera remain to be sampled and phylogenetic relationships in this clade may change with the addition of taxa. Therefore a definitive statement on the relatedness of Lichenostomus and Meliphaga cannot be made at this time. In the constrained parsimony consensus topology, (Fig. 4) all meliphagid taxa except the spinebills (Acanthorhynchus) are contained in one of four large clades. Each of these four clades is comprised of New Guinean and Australian endemics as well as more widely ranging taxa. The lack of bootstrap support for the three unconstrained clades arises from a paucity of characters supporting this part of the tree, and therefore we do not perceive phylogenetic relationships within the family as fully resolved at this level. However, each clade includes one or more strongly supported subclades. Clade 1. The first clade (labelled 1 in Fig. 4) is primarily comprised of a large well-supported subclade containing the wattlebirds (Anthochaera), Regent Honeyeater (Xanthomyza phrygia), miners (Manorina), and White-fronted honeyeater (P. albifrons) from Australia, the New Guinean montane endemic Melidectes, and the genera Lichenostomus and Meliphaga which occur in both Australia and New Guinea. One unexpected result is the sister relationship between Manorina and Melidectes. Their close relationship has interesting implications for the evolution of cooperative breeding in the family. The miners (Manorina), arguably the only truly obligate cooperative breeders in the family Meliphagidae (Clarke, 1995), are not most closely related to other cooperatively breeding honeyeater genera such as Melithreptus and Lichenostomus, but instead to Melidectes, in which this complex behavioural trait has not been observed. Also part of clade 1 is a subclade composed of the New Zealand endemic P. novaeseelandiae, the New Guinean endemic Pycnopygius, and the Australian endemic Certhionyx variegatus Biogeographically, this clade is a hodge–podge assemblage and although these taxa appear together in all analyses, relationships among them are not robustly resolved. Furthermore, a number of monotypic south Pacific and Indonesian endemic honeyeater genera, including another New Zealand taxon (Anthornis), were not sampled in the current study. With the addition of these enigmatic taxa, which may be potential relatives of members of this subclade, a more coherent biogeographical pattern may emerge. Clade 2. The second major clade is a poorly resolved trichotomy involving the Australian endemic chats (Epthianura and Ashbyia), the primarily Australian Conopophila and Ramsayornis, and the New Guinean endemics Melilestes, Melipotes, and Timeliopsis. Both Conopophila and Ramsayornis range into southern-most New Guinea, but neither genus has an endemic New Guinean species. Therefore, if the endemic chats (Ashbyia + Epthianura) and Conopophila + Ramsayornis are in fact sister groups, this clade would represent a substantial Australian radiation with a few ancient New Guinean lineages. The Australian chats (Epthianura and Ashbyia) reliably resolve as members of the second clade. At one time classified in their own family Epthianuridae (e.g., Schodde, 1975), the chats are remarkable for their adaptation to arid habitats (Schodde and Mason, 1999). Our result supports those of Sibley and Ahlquist (1990) and Christidis et al. (1993) and establishes the chats not only as members of the Meliphagidae, but in a position nested well within the family. Clade 3. This clade is strongly supported by the nuclear data, and further strengthened by two unequivocal indel characters. While it is possible that the indel characters are homoplastic, we consider it highly unlikely that both characters would show the same pattern of homoplasy and we take these two indel characters as firm evidence for this clade. One component of clade 3 is a subclade consisting of the New Guinean endemic Ptiloprora, the Australian endemic P. melanops, and Glycichaera fallax. Monotypic Glycichaera has a very limited distribution in Australia, but is much more widespread in New Guinea (Sibley and Monroe, 1990). A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 Based on the structure of its bill and skull, Schodde and Mason (1999) re-classified G. fallax in the New Guinean genus Timeliopsis, but our results show that these two genera are not closely related. Both Glycichaera and Timeliopsis primarily feed by gleaning foliage for insects (Beehler et al., 1986), which is a rare specialty among honeyeaters (Blakers et al., 1984). Both genera are also characterized by a moderately long, straight bill, which is also rare among meliphagids. It is possible that the straight bills of Timeliopsis and Glycichaera, adapted for foliage gleaning, appear more similar than they truly are, when compared with the longer, more curved bills of the majority of honeyeaters. In the other subclade, our results show the Australian endemic Black Honeyeater, C. niger, is sister to the five representative species of the genus Myzomela. Although C. niger was originally classified in the genus Myzomela (Gould, 1838), it is more likely that C. niger is the sister taxon to Myzomela, rather than nested within it, for two reasons. First, although the sample of Myzomela species included in the present study is small (5 of 30 species), it includes taxa from Australia, New Guinea, and islands in the south Pacific. Yet, C. niger is, on average across all four genes, 12.0% divergent from the Myzomela species (range of divergences within sampled Myzomela: 5.2–10.5%). Also, although C. niger shares a unique deletion in FIB5 with Myzomela, it also lacks a FIB5 deletion unique to the 5 species of Myzomela. Given the indel data, it is difficult to postulate a phylogenetic position for C. niger within Myzomela, and these data support instead a sister position. Clade 4. The fourth primary clade, contains three well-supported subclades, one of which is substantiated by a complex indel character. Included in clade 4 are the species-rich and wide-ranging genera Philemon and Lichmera, the primarily New Guinean Xanthotis, the primarily Australian Melithreptus and Entomyzon, the Australian endemics Certhionyx pectoralis, Grantiella, Plectorhyncha, Trichodere, P. novaehollandiae, P. nigra, and P. pyrrhoptera, and the south Polynesian endemic Foulehaio. This clade contains the broadest geographic coverage of the three, with representatives from Indonesia through to the south Pacific. Within one subclade, the friarbirds (Philemon) group with Grantiella, Plectorhyncha, and Xanthotis. Within the friarbirds, species group predictably with regard to morphological similarity: the three species with knobbed bills and bare black skin on the head (Philemon argenticeps, Philemon corniculatus, and Philemon buceroides) form one clade, while the two straight-billed, more fully feathered species (Philemon citreogularis and Philemon meyeri) form a second. An unanticipated pairing is the sister relationship between the Painted Honeyeater (Grantiella picta) and the Striped Honeyeater (Plectorhyncha lanceolata). Both of these species are morphologically disparate and a close relationship between 955 them has never been suspected. However, both Grantiella and Plectorhyncha build nests with similar structure (N. W. Longmore, pers. comm.). Another well-supported subclade contains three species of Phylidonyris, the monotypic White-streaked honeyeater (Trichodere cockerelli), Lichmera, and C. pectoralis. Trichodere is sister to the Phylidonyris clade and these four taxa do share bright yellow patches of color on the wing with some members of their sister genus Lichmera. However, presence of a yellow wing patch is a poor defining character for the group, as this character is widespread among other honeyeaters and strikingly similar in some unrelated taxa (e.g., Grantiella, Xanthomyza, and P. albifrons). In the third subclade, the Blue-faced Honeyeater (Entomyzon cyanotis) is sister to Melithreptus (two species examined). Entomyzon was classified within the genus Melithreptus without explanation by Storr (1977, 1984), but most authors (e.g., Schodde, 1975) have considered it to be more closely related to the largerbodied miners (Manorina) and wattlebirds (Anthochaera). The plumage of the Blue-faced Honeyeater is remarkably similar to that of Melithreptus, but it is possible that the great size difference between the two genera has obscured the significance of this fact. A definitive statement on whether Entomyzon should be included within Melithreptus, as advocated by Storr, waits on the sampling of the remaining four species of Melithreptus. Sister to Melithreptus + Entomyzon is the south Polynesian endemic Foulehaio. This grouping is supported by the presence of a large insertion in the FIB5 intron in these taxa relative to the remaining Meliphagidae (Tables 3a and 3b). The inserted sequence is a different length in all four species, but nucleotide sequence of these four inserted regions are identical and differ from one another by deletions within the insertion. The most parsimonious reconstruction of the acquisition of this insertion would be if it occurred in the common ancestor of Foulehaio, Melithreptus, and Entomyzon. The spinebills, Acanthorhynchus, which are perhaps the most obviously specialized nectarivores in the family and inhabit dry heathland and woodlands (Longmore, 1991), have no close relatives and appear as sister to all remaining Meliphagidae in all but the nuclear consensus topology. Heathlands were widespread in Australia during the Tertiary and became fragmented as rainfall became more seasonal during the Quaternary (Specht, 1979). Although the exact age of the Meliphagidae is unknown, the order Passeriformes has recently been posited to be Cretaceous in origin (Cooper and Penny, 1997; Cracraft, 2001; Hedges et al., 1996; Paton et al., 2002), which could place the origin of the Meliphagidae within the widespread-heathland period of the midTertiary. Our finding that heathland-adapted spinebills were one of the earliest established lineages of the 956 A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 Meliphagidae habitat corresponds well to a mid-Tertiary origin of the family. 4.3. Comparison with previous studies The DNA–DNA hybridization study of Sibley and Ahlquist (1985, 1990) included considerably fewer honeyeater taxa than the present study (20 genera, compared to 31). Some of the same species are present in both studies, but a number of the species used in the DNA–DNA hybridization study are not identified. This species identification is particularly important for the genus Certhionyx, which occupies three different positions in our results. Although Sibley and AhlquistÕs methods and analysis have been criticized from many standpoints (Cracraft, 1987; Harshman, 1994; Houde, 1987; Mindell, 1992), it is the only previously published phylogenetic hypothesis for the entire family Meliphagidae. Our topology shares two aspects with Sibley and AhlquistÕs topology: the chats (Epthianura and Ashbyia) are nested within the Meliphagidae and are related to the genus Ramsayornis; and Entomyzon is sister to the genus Melithreptus. Nearly all other taxa are placed in different positions in the two topologies. Other than the DNA–DNA hybridization study, only two studies grouped some of the honeyeater taxa based on morphological (Schodde, 1975) and biochemical (Sibley, 1970) characteristics. SibleyÕs study contained very few honeyeater taxa and has only one group in common with the current study: the grouping of Lichmera with Meliornis (P. novaehollandiae and P. nigra in the present study). Schodde (1975) grouped many Australian and New Guinean genera into two ‘‘lines’’ based on morphological and plumage characteristics, but there are as many differences as similarities between SchoddeÕs groups and our tree. However, the majority of the genera in SchoddeÕs ‘‘line 1’’ also belong to a monophyletic group in our topology: Melidectes, Anthochaera, Acanthagenys, Xanthomyza, Manorina, Meliphaga, and Lichenostomus. found in four different clades, and are not necessarily each otherÕs closest relatives even within a clade. Indeed, it seems unlikely that the New Guinean genera arose at the same time. The New Guinean endemic Melidectes is more closely related to its sister taxon, the Australian endemic Manorina (average uncorrected genetic distance across all four genes 7.9%), than are the three New Guinean genera Melilestes, Melipotes, and Timeliopsis to each other (average genetic distance 12.4%). In fact, Melidectes is more closely related to Manorina than the two species of Timeliopsis are to each other (9.2%). This suggests that Melidectes is more recently derived than these other New Guinean endemics. Of course, alternative explanations for this phenomenon exist: the rate of molecular evolution in Melidectes may be slower than in the other New Guinean endemics, or Melilestes and Melipotes may be the extant remnants of much larger clades and hence their level of genetic divergence appears to be greater. There is very little evidence in the topology for the ‘‘Tumbunan’’ hypothesis, which predicts a close relationship between taxa from the New Guinea highlands and those from the Australian rainforests (Schodde and Calaby, 1972). Species in the New Guinean genus Melidectes are all found in montane areas above 1200 m elevation. Yet their closest relatives are the Australian miners (Manorina), which inhabit woodlands and forests, but not rainforests (Blakers et al., 1984). Other New Guinean highland taxa are found in clade 2 (Fig. 1), but their sister relationships are not resolved. Nonetheless, the remaining taxa in these two clades are not rainforest inhabitants, although Ramsayornis and Conopophila inhabit swamps and mangroves in northern Australia (Blakers et al., 1984). The present study does not support the highlands-southern rainforests link proposed by Schodde and Calaby (1972), but this link might be upheld by studies at a lower phylogenetic level. An examination of the phylogenetic relationships within the large genera Meliphaga and Lichenostomus, which have member species in both New Guinea and Australia, might better test this hypothesis. 4.4. Biogeography of the meliphagidae Based on current taxonomy, the honeyeater faunas of New Guinea and Australia are each composed of a number of endemic genera, with additional genera and species shared between the two (Sibley and Monroe, 1990). A reasonable hypothesis for the generation of this pattern of distribution is that the endemic taxa arose at a point in time when Australia and New Guinea were relatively isolated from one another. This would produce a phylogenetic tree with largely monophyletic Australian and New Guinean endemic radiations. However, our topology does not support this hypothesis. Certainly there is very little evidence for a New Guinean radiation. Endemic New Guinean genera are 5. Conclusions The family Meliphagidae, as constituted here, is monophyletic, although the genera Anthochaera, Certhionyx, and Phylidonyris are not. Four major clades are recovered, and the overwhelming majority of honeyeater taxa belong to one of these four clades. The exception is the genus Acanthorhynchus (spinebills) which are sister to the remaining meliphagids and have no close relatives. The arid-adapted chats (Epthianura, Ashbyia) are nested deeply within the family, although their sister group is not identified. Each of the four major clades contains a mix of New Guinean and Australian A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 endemics, along with more wide-ranging taxa. There is no evidence for the occurrence of separate Australian and New Guinean endemic radiations. There is also little evidence for the ‘‘Tumbunan’’ hypothesis of a close phylogenetic relationship between New Guinea highland and eastern Australia rainforest taxa. Acknowledgments The authors thank the following individuals and institutions for their assistance in obtaining samples: Museum Victoria (R. OÕBrien, B. Gilles, and J. Norman), Australian National Wildlife Collection (R. Schodde, J. Wombey, I. Mason, and B. Gill), Western Australia Museum (R. Johnstone), Museum of New Zealand (A. Tennyson), R. Fleischer, S. Pruett-Jones, 957 and D. Armstrong. Funding for this project (to A.C.D.) was provided by the University of Chicago Hinds Fund, the American OrnithologistsÕ Union, the American Museum of Natural History Chapman Fund and the National Science Foundation (DEB-9623526). A.C.D. was supported by a Field Museum Lester Armour Fellowship and an American Association of University Women Dissertation Fellowship. All molecular work was completed in the Field MuseumÕs Pritzker Laboratory for Molecular Systematics and Evolution. The aligned data matrix and complete set of saturation plots are available on the web at http://ginger.ucdavis.edu/ driskell.html or from TreeBASE http://www.treebase. org. A.C.D. would like to thank S. Hackett, J. Voight, B. Chernoff, M. McMahon, and G. Burleigh for beneficial discussion and advice. This manuscript has benefited from critiques by two anonymous reviewers. Appendix A List of specimens sequenced in the present study Family Species Meliphagidae Acanthagenys rufogularis Acanthorhynchus superciliosus Acanthorhynchus tenuirostris Anthochaera carunculata Anthochaera chrysoptera Anthochaera lunulata Anthochaera paradoxa Ashbyia lovensis Certhionyx niger Certhionyx pectoralis Certhionyx variegatus Conopophila albogularis Conopophila rufogularis Entomyzon cyanotis Epthianura albifrons Epthianura aurifrons Epthianura crocea Epthianura tricolor Foulehaio carunculata Glycichaera fallax Grantiella picta Lichenostomus flavescens Lichmera alboauricularis Lichmera indistincta Manorina flavigula Melidectes ochromelas Manorina melanophrys Melidectes belfordi Meliphaga albonotata Meliphaga gracilis Melipotes fumigatus Spec. No. Voucher Collection locality MV1122 MV248 B873 C257 B792 MV175 B736 D173 C954 C912 W036 MV1216 MV1300 F274 D328 D156 D175 D229 2077 E663 MV2673 D029 E629 C271 42856 E360 42737 E168 E471 C753 E332 MV MV ANWC ANWC ANWC MV ANWC ANWC ANWC ANWC SAM MV MV ANWC ANWC ANWC ANWC ANWC RF UV ANWC MV ANWC ANWC ANWC ANWC ANWC ANWC ANWC ANWC ANWC ANWC Stockyard HS, NT Albany, WA Tenterfield, NSW StuartÕs Point, NSW Upper Blessington, TAS Esperance, WA Upper Blessington, TAS Koonchara Dune, SA Winton, QLD Musgrave, QLD Mabel Creek, SA Gunn Point, NT Cape Crawford, NT Chillagoe, QLD Eyre Peninsula, SA Innaminka, SA Koonchara Dune, SA William Creek, SA Unknown Veimari River, PNG Killawarra State Forest, VIC Timber Creek, NT Port Moresby, PNG StuartÕs Point, NSW Charters Tower, QLD Tetebedi, PNG Kempsey, NSW Tetebedi, PNG Tetebedi, PNG McIlwraith Range, QLD Tetebedi, PNG 958 A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 Appendix A (continued) Family Species Spec. No. Voucher Collection locality JC100 MV371 2494 MV1198 C531 E240 C402 JCW095 C863 D008 C720 E683 D361 D451 MV198 B685 B615 C379 11/1996 E173 C173 C057 C035 MV1230 C900 E233 E714 42941 F724 E594 ANWC MV RF UV MV ANWC ANWC ANWC ANWC ANWC ANWC ANWC ANWC ANWC ANWC MV ANWC ANWC ANWC MNZ ANWC ANWC ANWC ANWC MV ANWC ANWC ANWC ANWC ANWC ANWC Burra Range, QLD Ceduna, SA Unknown Gunn Point, NT Cathu, QLD Tetebedi, PNG Agnes Waters, QLD Coen region, QLD Silver Plains, QLD Pine Creek, NT Musgrave, QLD Veimari River, PNG SinclairÕs Gap, SA Big Desert, VIC Raventhorpe, WA Launceston, TAS Mt. Lofty Ranges, SA Agnes Waters, QLD New Zealand Tetebedi, PNG Efogi, PNG Efogi, PNG Efogi, PNG Finniss River, NT Silver Plains, QLD Tetebedi, PNG Veimari River, PNG Silver Plains, QLD Sutton, NSW Kokoda, PNG Pardalotidae Acanthiza apicalis Acanthiza chrysorrhoa Dasyornis broadbenti Gerygone chrysogaster Gerygone chloronotus Pardalotus punctatus Pardalotus striatus Sericornis frontalis Sericornis perspicillatus MV158 MV116 MV2172 E670 E122 B479 B471 MV228 E313 MV MV MV ANWC ANWC ANWC ANWC MV ANWC Norseman, WA Port Augusta, SA Aireys Inlet, VIC Veimari River, PNG Tetebedi, PNG Sutton, NSW Sutton, NSW Albany, WA Tetebedi, PNG Maluridae Amytornis striatus Malurus lamberti Malurus splendens Stipiturus mallee SGW1 VW104 SW683 MEW1 SPJ UV BCP BCP SPJ UV Hattah-Kulkyne, Brookfield Cons. Brookfield Cons. Hattah-Kulkyne, Melipthreptus albogularis Melipthreptus brevirostris Myzomela cardinalis Myzomela eythrocephala Myzomela obscura Myzomela rosenbergii Myzomela sanguinolenta Philemon argenticeps Philemon buceroides Philemon citreogularis Philemon corniculatus Philemon meyeri Phylidonyris albifrons Phylidonyris melanops Phylidonyris nigra Phylidonyris novaehollandiae Phylidonyris pyrrhoptera Plectorhyncha lanceolata Prosthemadera novaeseelandiae Ptiloprora guisei Ptiloprora plumbea Pycnopygius cinereus Pycnopygius stictocephalus Ramsayornis fasciatus Ramsayornis modestus Timeliopsis fulvigula Timeliopsis griseigula Trichodere cockerelli Xanthomyza phrygia Xanthotis flaviventer N.P., VIC Park, SA Park, SA N.P., VIC Species are organized by family. Specimen number (Spec. No.) is the collectorÕs field number, in most instances, or the tissue or Accession number. Voucher is the location of the voucher specimen corresponding to the tissue specimen. Collection locality is the nearest named location to the specimenÕs collection locality. ANWC ¼ Australian National Wildlife Collection, C.S.I.R.O., Canberra, Australia; BCP ¼ Brookfield Conservation Park, SA; MNZ ¼ Museum of New Zealand Te Papa Tongerewa, Wellington, New Zealand; MU ¼ Massey University, Palmerston North, New Zealand; MV ¼ Museum Victoria, Melbourne, VIC; RF ¼ Rob Fleischer, Smithsonian Institution, Washington, DC; SPJ ¼ Steve Pruett-Jones, University of Chicago, Chicago, IL; WAM ¼ Western Australia Museum, Perth, WA; NT ¼ Northern Territory; QLD ¼ Queensland; SA ¼ South Australia; SAM ¼ South Australian Museum, Adelaide, SA; TAS ¼ Tasmania; PNG ¼ Papua New Guinea; VIC ¼ Victoria; and UV ¼ unvouchered. A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 References Allard, M.W., Farris, J.S., Carpenter, J.M., 1999. Congruence among mammalian mitochondrial genes. Cladistics 15, 75–84. Barker, K.F., Barrowclough, G.F., Groth, G.F., 2002. A phylogenetic hypothesis for passerine birds: taxonomic and biogeographic implications of an analysis of nuclear DNA sequence data. Proc. R. Soc. Lond. B 269, 295–308. Beehler, B.M., Pratt, T.K., Zimmerman, D.A., 1986. Birds of New Guinea. Princeton University Press, Princeton. Blakers, M., Davies, S.J.J.F., Reilly, R.N., 1984. The Atlas of Australian Birds. Melbourne University Press, Melbourne. Christidis, L., 1991. Molecular and biochemical evidence for the origins and evolutionary radiations of the Australasian avifauna. In: Bell, B.D., Cossee, R.O., Flux, J.E.C., Heather, B.D., Hitchmough, R.A., Robertson, C.J.R., Williams, M.J. (Eds.), Proceedings of the twentieth International Ornithological Congress. New Zealand Ornithological Trust Board, Wellington, pp. 392–397. Christidis, L., Boles, W.E., 1994. The Taxonomy and Species of Birds of Australia and its Territories. Australasian Ornithological Monographs No.1. Royal Australasian Ornithologists Union, Melbourne. Christidis, L., Schodde, R., 1991. Relationships of Australo-Papuan songbirds (Aves: Passeriformes): protein evidence. Ibis 133, 277– 285. Christidis, L., Schodde, R., 1993. Relationships and radiations in the meliphagine honeyeaters, Meliphaga, Lichenostomus, and Xanthotis (Aves: Meliphagidae): protein evidence and its integration with morphology and ecogeography. Austral. J. Zool. 41, 293–316. Christidis, L., Schodde, R., Robinson, N.A., 1993. Affinities of the aberrant Australo-Papuan honeyeaters, Toxorhamphus, Oedistoma, Timeliopsis, and Epthianura: protein evidence. Austral. J. Zool. 41, 423–432. Clarke, M.F., 1995. Co-operative breeding in Australasian birds: a review of hypotheses and evidence. Corella 19, 73–90. Coates, B.J., Bishop, K.D., 1997. A Guide to the Birds of Wallacea. Dove Publications, Alderley. Cooper, A., Penny, D., 1997. Mass survival of birds across the Cretaceous-Tertiary boundary: molecular evidence. Science 275, 1109–1113. Cracraft, J., 1981. Toward a phylogenetic classification of the recent birds of the world (Class Aves). Auk 98, 681–714. Cracraft, J., 1987. DNA hybridization and avian phylogenetics. Evol. Biol. 21, 47–96. Cracraft, J., 2001. Avian evolution, Gondwana biogeography and the Cretaceous-Tertiary mass extinction event. Proc. R. Soc. Lond. B 268, 459–469. Cracraft, J., Feinstein, J., 2000. What is not a bird of paradise? Molecular and morphological evidence places Macgregoria in the Meliphagidae and the Cnemophilinae near the base of the corvoid tree. Proc. R. Soc. Lond. B 267, 233–241. Cunningham, C.W., 1997. Can three incongruence tests predict when data should be combined? Mol. Biol. Evol. 14, 733–740. Desjardins, P., Morais, R., 1990. Sequence and gene organization of the chicken mitochondrial genome: a novel gene order in higher vertebrates. J. Mol. Biol. 212, 599–634. Ericson, P.G., Christidis, L., Cooper, A., Irestedt, M., Jackson, J., Johansson, U.S., Norman, J.A., 2002a. A Gondwanan origin of passerine birds supported by DNA sequences of the endemic New Zealand wrens. Proc. R. Soc. Lond. B 269, 235–241. Ericson, P.G., Christidis, L., Irestedt, M., Norman, J.A., 2002b. Systematic affinities of the lyrebirds (Passeriformes: Menura), with a novel classification of the major groups of passerine birds. Mol. Phylogenet. Evol. 25, 53–62. Farris, J.S., Kallersj€ o, M., Kluge, A.G., Bult, C., 1995a. Constructing a significance test for incongruence. Syst. Biol. 44, 570–572. 959 Farris, J.S., Kallersj€ o, M., Kluge, A.G., Bult, C., 1995b. Testing significance of incongruence. Cladistics 10, 315–319. Goldman, N., Anderson, J.P., Rodrigo, A.G., 2000. Likelihood-based tests of topologies in phylogenetics. Syst. Biol. 49, 652–670. Gould, J., 1838. The Birds of Australia and the Adjacent Islands. John Gould, London. Griffiths, C.S., 1997. Correlation of functional domains and rates of nucleotide substitution in cytochrome b. Mol. Biol. Evol. 7, 352– 365. Harshman, J., 1994. Reweaving the tapestry: what can we learn from Sibley and Ahlquist (1990)? Auk 111, 377–388. Hedges, S.B., Parker, S.H., Sibley, C.G., Kumar, S., 1996. Continental breakup and the ordinal diversification of birds and mammals. Nature 381, 226–229. Helm-Bychowski, K., Cracraft, J., 1993. Recovering phylogenetic signal from DNA sequences: relationships within the Corvine assemblage (Class Aves) as inferred from complete sequences of the mitochondrial DNA cytochrome-b gene. Mol. Biol. Evol 10, 1196–1214. Houde, P., 1987. Critical evaluation of DNA hybridization studies in avian systematics. Auk 104, 17–32. Houde, P., Cooper, A., Leslie, E., Strand, A.E., Montano, G.A., 1997. Phylogeny and evolution of 12S rDNA in Gruiformes (Aves). In: Mindell, D.P. (Ed.), Avian Molecular Evolution and Systematics. Academic Press, New York, pp. 121–158. Huelsenbeck, J.P., 2000. MRBAYES: Bayesian inference of phylogeny. Distributed by the author, Department of Biology, University of Rochester. Huelsenbeck, J.P., Bollback, J.P., 2001. Application of the likelihood function in phylogenetics. In: Balding, D.J., Bishop, M., Cannings, C. (Eds.), Handbook of Statistical Genetics. Wiley, Chichester, pp. 415–443. Huelsenbeck, J.P., Crandall, K.A., 1997. Phylogeny estimation and hypothesis testing using maximum likelihood. Ann. Rev. Ecol. Syst. 28, 437–466. Huelsenbeck, J.P., Rannala, B., 1997. Maximum likelihood estimation of phylogeny using stratigraphic data. Paleobiology 23, 174–180. Jukes, T.H., Cantor, C.R., 1969. Evolution of protein molecules. In: Munro, H.N. (Ed.), Mammalian Protein Metabolism. Academic Press, New York, pp. 21–132. Keast, A., 1985. An introductory ecological biogeography of the Australo-Pacific Meliphagidae. NZ J. Zool. 12, 605–622. Kirchman, J.J., Hackett, S.J., Goodman, S.M., Bates, J.M., 2001. Phylogeny and systematics of ground rollers (Brachypteraciidae) of Madagascar. Auk 118, 849–863. Kocher, T.D., Thomas, W.K., Meyer, A., Edwards, S.V., P€a€abo, S., Villablanca, F.X., Wilson, A.C., 1989. Dynamics of mitochondrial DNA evolution in animals: amplification and sequencing with conserved primers. Proc. Natl. Acad. Sci. USA 86, 6196–6200. Lanyon, S.M., Hall, J.G., 1994. A reconsideration of barbet monophyly using mtDNA sequence data. Auk 111, 389–397. Larget, B., Simon, D.L., 1999. Markov chain Monte Carlo algorithms for the Bayesian analysis of phylogenetic trees. Mol. Biol. Evol. 16, 750–759. Lea, A.H., Gray, J.T., 1935. The food of Australian birds. An analysis of the stomach contents. Emu 35, 251–280. Longmore, W., 1991. Honeyeaters and their allies of Australia. Angus and Robertson, North Ryde. Lutzoni, F.P., Wagner, P., Reeb, V., Zoller, S., 2000. Integrating ambiguously aligned regions of DNA sequences in phylogenetic analyses without violating positional homology. Syst. Biol. 49, 628–651. Maddison, W.P., Maddison, D.R., 2003. M E S Q U I T E : a modular system for evolutionary analysis. Version 1.0. http://mesquiteproject.org.. Mindell, D.P., 1992. DNA–DNA hybridization and avian phylogeny. Syst. Biol. 41, 126–134. 960 A.C. Driskell, L. Christidis / Molecular Phylogenetics and Evolution 31 (2004) 943–960 Mindell, D.P., Sorenson, M.D., Huddleston, C.J., Miranda Jr, H.C., Knight, A., Sawchuk, S.J., Yuri, T., 1997. Phylogenetic relationships among and within select avian orders based on mitochondrial DNA. In: Mindell, D.P. (Ed.), Avian Molecular Evolution and Systematics. Academic Press, New York, pp. 213–247. Paton, D.C., Ford, H.A., 1977. Pollination by birds of native plants in South Australia. Emu 77, 73–85. Paton, T., Haddrath, O., Baker, A.J., 2002. Complete mitochondrial DNA genome sequences show that modern birds are not descended from transitional shorebirds. Proc. R. Soc. Lond. B 269, 839–846. Posada, D., Crandall, K.A., 1998. MO D E L T E S T : testing the model of DNA substitution. Bioinformatics 14, 817–818. Pratt, H.D., Bruner, P.L., Berrett, D.G., 1987. A Field Guide to the Birds of Hawaii and the Tropical Pacific. Princeton University Press, Princeton. Quinn, T.W., 1997. Molecular evolution of the mitochondrial genome. In: Mindell, D.P. (Ed.), Avian Molecular Evolution and Systematics. Academic Press, New York, pp. 4–29. Rand, A.L., Gilliard, E.T., 1968. Handbook of New Guinea Birds. Weidenfold and Nicolson, London. Recher, H.F., 1981. Nectar-feeding and its evolution among Australian vertebrates. In: Keast, A. (Ed.), Ecological Biogeography of Australia. Junk, The Hague, pp. 1637–1648. Schodde, R., 1975. Interim list of Australian Songbirds. Passerines. Royal Australasian Ornithologists Union, Melbourne. Schodde, R., Calaby, J.H., 1972. The biogeography of the Australopapuan bird and mammal faunas in relation to Torres Strait. In: Walker, D. (Ed.), Bridge and Barrier: The Natural and Cultural History of Torres Strait. Australian National University Press, Canberra, pp. 257–300. Schodde, R., Mason, I.J., 1999. The Directory of Australian Birds: Passerines. CSIRO Publishing, Collingwood. Schodde, R., McKean, J.L., 1976. The relations of some monotypic genera of Australian oscines. In: Frith, H.J., Calaby, J.H. (Eds.), Proceedings of the XVI International Ornithological Congress. Australian Academy of Science, Canberra, pp. 530–541. Sennblad, B., Bremer, B., 2000. Is there a justification for differential a priori weighting in coding sequences? A case study from rbcL and the Apocynaceae s.l.. Syst. Biol. 49, 101–113. Sibley, C.G., 1970. A comparative study of the egg-white proteins of passerine birds. Bull. Peabody Mus. Nat. Hist. 32, 1–131. Sibley, C.G., Ahlquist, J.E., 1985. The phylogeny and classification of the Australo-Papuan passerine birds. Emu 85, 1–14. Sibley, C.G., Ahlquist, J.E., 1990. Phylogeny and Classification of Birds. A Study in Molecular Evolution. Yale University Press, New Haven. Sibley, C.G., Monroe, B.L., 1990. Distribution and Taxonomy of Birds of the World. Yale University Press, New Haven. Slikas, B., Jones, I.B., Derrickson, S.R., Fleischer, R.C., 2000. Phylogenetic relationships of Micronesian white-eyes based on mitochondrial sequence data. Auk 117, 355–365. Specht, R.L., 1979. The sclerophyllous (heath) vegetation of Australia: the eastern and central states. In: Specht, R.L. (Ed.), Heathlands and Related Shrublands. Elsevier, Amsterdam, pp. 125–210. Springer, M.S., Hiroyoshi, H., Keisuke, U., Jason, M., Sibley, C.G., 1995. Molecular evidence that the Bonin Islands ‘‘Honeyeater’’ is a white-eye. J. Yamashina Inst. Ornith. 27, 66–77. Storr, G.M., 1977. Birds of the Northern Territory. Western Australian Museum Special Publication No. 7. Storr, G.M., 1984. Revised List of Queensland birds. Records of the Western Australian Museum Supplement No. 19. Swofford, D.L., 2002. P A U P *. Phylogenetic Analysis using Parsimony (*and other methods). Sinauer Associates, Sunderland. Swofford, D.l., Olsen, G.J., Waddell, P.J., Hillis, D.M., 1996. Phylogenetic inference. In: Hillis, D.M., Moritz, C., Mable, B.K. (Eds.), Molecular Systematics. Sinauer, Sunderland, MA, pp. 407– 514. Veerman, P.A., 1992. Vocal mimicry of larger honeyeaters by the Regent Honeyeater Xanthomyza phrygia. Austral. Bird Watcher 14, 180–189. Wetmore, A., 1960. A classification for the birds of the world. Smithsonian Misc. Coll. 139, 1–37.