Abstract
The systematics of the babblers (Timaliidae) and related members of the Old World insectivorous passerines have been particularly difficult. To clarify our understanding of this group, phylogenetic relationships were constructed using sequences of three mitochondrial genes (cytochrome b, rRNA 12S and 16S). The results indicated that several species traditionally placed among babblers, the shrike babblers (Pteruthius) and the Gray-chested Thrush Babbler (Kakamega poliothorax), are not related to the Timaliidae, but belong to other passerine groups. Furthermore, the phylogenetic hypotheses inferred from molecular data suggest that the babblers assemblage includes two other oscine taxa traditionally considered to be distantly related, Sylvia (Sylviidae) and Zosterops (Zosteropidae). The polyphyly of several babbler genera is discussed, with particular attention to the laughingthrushes (genera Garrulax and Babax) for which the phylogeny is compared to previous hypotheses of relationships. Results from different tests under the maximum-parsimony and maximum-likelihood criteria indicate the rejection of the hypothesis of monophyly for the laughingthrushes group. Thus, the molecular phylogeny challenges the traditional classification of the Timaliidae.
Résumé
La systématique des timalies (Timaliidae), ainsi que celle des groupes d'insectivores de l'Ancien Monde qui leur sont apparentés, est depuis toujours un point difficile de la classification des passereaux. Au cours de cette étude, nous avons utilisé les séquences d'ADN de trois gènes mitochondriaux (le cytochrome b et les ARNr 12S et 16S) afin de reconstruire les relations phylogénétiques de ce groupe d'oiseaux. Les résultats montrent que plusieurs espèces classées traditionnellement parmi les timalies, les Allotries (genre Pteruthius) et l'Akalat à poitrine grise (Kakamega poliothorax), ne sont pas apparentées aux Timaliidae et appartiennent en réalité à d'autres groupes de passereaux. De plus, les hypothèses phylogénétiques déduites de l'analyse des données moléculaires suggèrent que deux autres taxa, qui sont traditionnellement placés dans des familles différentes, à savoir les Sylviidae pour le genre Sylvia, et les Zosteropidae pour le genre Zosterops, appartiennent en réalité aux Timaliidae. La polyphylie de plusieurs genres de timalies est discutée, particulièrement en ce qui concerne les grives bruyantes ou Garrulaxes (genres Garrulax et Babax) pour lesquelles la phylogénie obtenue est comparée aux hypothèses proposées précédemment. Différents tests conduits sous le critère du maximum de vraisemblance permettent de rejeter l'hypothèse de la monophylie du groupe des grives bruyantes. Ainsi, la phylogénie moléculaire obtenue remet en question de nombreux points de la classification actuelle des Timaliidae.
The family Timaliidae, the babblers, comprises more than 200 species, primarily forest birds, distributed for the most part in Indo Malayan Asia and Africa. Babblers have often been placed in the “Old World insectivorous” group, the Muscicapidae sensu lato, which includes various other passerines like the Turdidae (thrushes), Muscicapidae (flycatchers), and Sylviidae (warblers) (Hartert 1910, Mayr and Amadon 1951, Deignan 1964, Morony et al. 1975). Morphologically, babblers differ from thrushes and flycatchers by the lack of distinct juvenal plumage, and they differ from warblers by heavier size and shape, as well as being nonmigratory (McClure 1974). Babblers display great diversity in size, bill shape, and plumage coloration, and perhaps the most reliable character that unites babblers is their high sociability. They often gather in parties and flocks, and some species cooperate in nesting activities and communal defense (Lack 1968, Grimes 1976, Zahavi 1976, Gaston 1977). Most babbler species clump together when perched during the day and while roosting at night, and mutual preening is observed in many species (Simmons 1963).
Delacour (1946, 1950) conducted the main systematic review of the group. He defined 252 species in 47 genera, distributed into six tribes: Pellorneini, Pomatorhinini, Timaliini, Chamaeini, Turdoidini, and Picathartini (Table 1). Delacour's taxonomic list was proposed to follow a sequence from “primitive” (Pellorneini) to “derived” taxa (Picathartini), based mostly on his observation of plumage coloration (duller, more primitive) and bill shape. According to Delacour, the Old World warblers (Sylviidae) were the sister group of the Timaliidae. Since Delacour's work, the systematics of babblers has been the subject of limited revisions, dealing mostly with generic limits and definitions. The jungle babblers, genus Trichastoma and related genera, have been revised on the basis of phenetic analysis of morphological and ecological characters (Ripley and Beehler 1985). Other systematic works included considerations of morphological similarities, but no taxonomic decisions were based on discontinuous variation (Irwin 1983; Harrison 1986a, b; Van de Weghe 1988). Thus, the diagnosis and taxonomic sequence of the babblers given by Delacour has been largely followed by subsequent authors (including Peter's Checklist, Deignan 1964), until Sibley and Ahlquist's attempt to clarify the phylogeny of the Timaliidae using DNA hybridization (Sibley and Ahlquist 1982, 1990). They proposed important reevaluations of the systematics of several taxa traditionally placed among babblers. Firstly, they showed that the Australian and New Guinean genera (Garritornis and Pomatostomus) are in fact members of the Corvoidea, despite their morphological and ecological similarities with the Asian scimitar babblers (Pomatorhinus). Secondly, they suggested that the African rock-fowls (Picathartes), whose taxonomic position has long been debated (Lowe 1938, Webb 1949, Delacour 1950, Delacour and Amadon 1951, Mayr and Amadon 1951), are not closely related to the babblers, although their analysis failed in finding the sister group of that genus (Barker et al. [2002] confirmed with nuclear DNA sequence data that Picathartes does not belong to the core Sylvioidea). The phylogenetic tree obtained by Sibley and Ahlquist is presented in Figure 1. The main results are that (1) the laughingthrushes (Garrulax) are sister taxa to all other babblers; (2) the warbler genus Sylvia belongs to the babbler group, and is sister taxon to the Wrentit (Chamaea fasciata); and (3) all other African and Asian babblers studied are grouped together. However, those results should still be considered as provisional because of the low number of taxa studied (11 species for 9 genera), and methodological problems affecting their analysis of hybridization distances (Lanyon 1992, Mindell 1992). A previous cytochrome b study including a small number of babblers also placed Sylvia among babblers (Fjeldså et al. 1999).
Timaliidae tribes, according to Delacour (1946, 1950). The taxa included in this study are indicated in boldface.
Timaliidae tribes, according to Delacour (1946, 1950). The taxa included in this study are indicated in boldface.
Relationships among babblers as proposed by Sibley and Ahlquist (1990).
The main objective of this study is to clarify the relationships of the babblers by answering the following questions: (1) What are the limits of the babbler family? (2) What are the phylo-genetic relationships among babblers, and is Delacour's classification of babblers consistent with the phylogeny of the family? (3) Are hypotheses about babblers derived from DNA hybridization distances correct, and especially, is Sylvia more closely related to babblers than to warblers? This article is a contribution to a long-term study on the systematics and evolution of babblers (Cibois 2000). Previous publications dealt, first, with the “babbler-like” endemics of the island of Madagascar and their relationships to the core of Asian and African babblers (Cibois et al. 1999, 2001), and second, with a subset of Asian babblers in the genera Yuhina and Stachyris (Cibois et al. 2002). The present study extends the taxon sampling of that group for Asian, African, and American taxa.
Materials and Methods
Taxa sampling and source of DNA
Sixty-two species of babblers representing 31 genera were examined, along with 3 genera belonging to the Sylviidae (warblers); one representative of each of the families Turdidae, Muscicapidae, Zosteropidae, Laniidae, Corvidae, Monarchidae, Platysteiridae; and one suboscine as the most distant outgroup, the Tropical Kingbird (Tyrannus melancholicus, Tyrannidae). Tissue samples origin and GenBank accession numbers are presented in the Appendix. For most of the samples, total genomic DNA was extracted from frozen or alcohol-preserved tissues (muscle, liver, blood) or small pieces (0.5–1 cm2) of museum skins (labeled MNHN CG), with CTAB buffer containing proteinase K (0.1 mg mL−1) (Sambrook et al. 1989, Winnepenninckx et al. 1993). Alternatively DNA was extracted by placing <20 mg in 250 μ L of 5% Chelex (Bio-Rad) and heating at 100°C for 15 min. In the case of museum skins, DNA was extracted using CTAB buffer and the time of protein digestion was increased from 2 to 12 hours.
DNA amplification and sequencing
Specific fragments of the cytochrome b, 16S, and 12S genes were amplified using the polymerase chain reaction (PCR). The standard protocols and the primers are described in previous papers (Cibois et al. 1999, 2002). Amplification products were sequenced with the same primers used for PCR amplification and also internal primers. The same set of primers was used for the DNA extracted from museum skins with no particular difficulty for amplification. Sequencing was performed manually for most of the samples with the Thermo Sequenase Cycle Sequencing kit (Amersham Pharmacia Biotech, Buckinghamshire, United Kingdom), based on the Sanger method (Sanger et al. 1977), and sequences were read manually and compiled with the MUST package (Philippe 1993). For samples extracted with the Chelex method, PCR were performed using the same protocols, and PCR products were purified using GeneClean (Bio101) kits. Those products were resuspended in 12 μL of water, and then sequenced in a ABI 9600 thermocycler in both directions in 7 μL total volume reactions containing 2.5 μ L of PCR products, 3 μL of Terminator Mix (dRhodamine, Applied Biosystems) and 1.5 μL of primer (10 μM). Sequenced reactions were cleaned of excess nucleotides by ethanol precipitation, using 74 μL of a solution containing 10 mlL of Ethanol (70%) and 10 μL of Magnesium chlorate (0.5 M), dried and resuspended in 1.8 μ L formamide loading dye. Reactions were then electrophoresed on an Applied Biosystems 377 automated sequencer. Contig alignments were created using SEQUENCHER (Genecodes, Ann Arbor, Michigan). Accuracy of the DNA sequencing was verified with sequencing both heavy and light strands of most PCR fragments, and when possible using overlapping fragments and sequencing several individuals and types of tissues. Alignments of 12S and 16S sequences were accomplished both by hand and by using CLUSTAL W (Thompson et al. 1994).
Data analysis
Phylogenetic trees were estimated using maximum-parsimony and maximum-likeli-hood analyses, all performed using PAUP* 4.0 (Swofford 1999). Prior to all analyses, heterogeneity in base composition was investigated for each gene, by codon position for cytochrome b and by partition of loops and stems for 12S and 16S. For rRNA sequences, secondary structure was determined using the 16S model (see Gutell et al. inMaidak et al. 1994) for the chicken (Gallus gallus) and the 12S model presented by Mindell et al. (1997) for the Peregrine Falcon (Falco peregrinus). Saturation among sequences was examined graphically, by category of substitution (transversion and transition), codon position, or secondary structure partitions. Under the maximum-parsimony criterion, heuristic searches were conducted with 10 replicates of the random taxon sequence addition option. Searches were undertaken with gaps treated as missing data. Data were treated with equal weight or with weighting schemes. Weights were applied to the data using the step matrix option in PAUP*. The robustness of the clades was assessed by bootstrap analysis with 100 iterative resamplings using heuristic searches (Felsenstein 1985) and by calculating Bremer's indices (Bremer 1988, 1994) using AUTODECAY 2.9.9. (Eriksson 1997). The partition homogeneity test in PAUP* was used to test for homogeneity among the three data sets (Farris et al. 1995). Phylogenetic analyses were performed with and without the most distant outgroup (Tyrannus) to test the effect of divergent rooting (Smith 1994). Maximum-likelihood analysis was performed only on the combined data set, after using the program MODELTEST (Posada and Crandall 1998) to choose the model of DNA evolution that fit the data the best. Bootstrap replicates are computationally intensive under the maximum-likelihood criterion and therefore was not performed with our large data set (73 taxa).
Several tests can be used to compare alternative topologies. The Kishino-Hasegawa test (Kishino and Hasegawa 1989) and Templeton's test (Templeton 1983) have been used extensively to compare topologies obtained using different data sets (often different genes, or morphological versus molecular characters) or with the same data set but using different reconstruction methods (see for instance Zardoya et al. 1998, Goto and Kimura 2001). It has been argued, however, that those tests cannot be validly applied when comparing an optimal tree, obtained a posteriori, to an alternative a priori topology, nor to compare two trees obtained a posteriori (Goldman and Whelan 2000, Goldman et al. 2000, Buckley et al. 2001). The topology dependent test (t-PTP; Faith 1991) has been criticized for problems dealing with the formulation of the null hypothesis, and the fact that the test is sensitive to the number of taxa and characters (the parameters that determine the number of possible trees) (see Harshman 2001 for references on critics and replies from Faith and collaborators). Despite those critics, this test has been used in different studies to compare the topology obtained to a prioiri hypotheses (see for instance Alvarez et al. 2000, Bernardi et al. 2000). Alternative tests have been proposed to compare both a priori and a posteriori hypotheses: a nonparametric test, the Shimodaira-Hasegawa test (Shimodaira and Hasegawa 1999, Buckley et al. 2001), and two parametric tests, the parametric bootstrap test (Huelsenbeck et al. 1996) and the SOWH test (Goldman et al. 2000). All of those procedures were performed with this data set to test for the monophyly of one group of babblers, the laughingthrushes (Garrulax and Babax). With >50 described species, the laughingthrushes represent the most important group among Timaliidae. Contrary to most babbler genera for which nothing has been proposed in terms of interspecific relationships, the systematics of the laughingthrushes has been reviewed in detail by Berlioz (1930). He proposed several hypotheses of relationships among species and gave a taxonomic sequence that has been followed in the main part by Delacour (1946). Despite the limited taxa sampling of laughingthrushes available for this study (10 species), it seemed important to compare the results of the molecular phylogeny to Berlioz's hypotheses, and especially to perform several tests for the monophyly of that group because all preliminary phylogenetic analyses conducted with this data set suggested that that group was not monophyletic. The Shimodaira-Hasegawa test is implemented in PAUP*, and was used to compare two topologies obtained with or without constraining the monophyly of the laugh-ingthrushes. The protocol suggested by Anderson, Goldman, and Rodrigo (see Acknowledgments) was followed to perform the SOWH test. First, the program SEQ-GEN 1.2.5. (Rambaut and Grassly 1997) was used to generate 100 simulated data sets via parametric bootstrapping (general time reversible model, option REV). For each simulated data set, the likelihood scores of the null and maximum-likeli-hood topologies were calculated with PAUP* (the null hypothesis to be tested is the monophyly of the laughingthrushes). The difference between those like-lihood scores was used as the null distribution for the test statistic in the SOWH test (Goldman et al. 2000). The SOWH test was performed with partial optimization, using a reduced data set of 40 taxa because the initial data matrix was too large to perform this test. In that reduced data set (see Appendix), the phylogenetic tree was similar to the topology found using the complete matrix, a result that allowed the use of this reduced data set in the analysis. The maximum-likeli-hood searches on simulated data sets were performed using SPR branch swapping, with the starting tree obtained by neighbor joining, two options also used for the analysis of the complete matrix. The parametric bootstrap test compares the length of trees obtained under different hypotheses with the distribution of difference generated by simulations. Once again, the simulated data sets were obtained using SEQ-GEN (general time reversible model, option REV), with parameters estimated on an initial maximum-likelihood topology (with the reduced data set of 40 taxa, see Appendix). Those simulated data sets were then analyzed with PAUP*, under the maximum-parsimony criterion, with or without constraining the monophyly of laughingthrushes. The distribution of simulated length differences was compared to observed difference to test for rejection of the null hypothesis.
Results
Sequence variation and saturation analysis
Fragments of 477, 414, and 543 bp of the cytochrome b, 12S, and 16S, respectively, were obtained for all taxa (length including indels for the RNA genes with Clustal alignments, submitted to European Molecular Biology Laboratory, accession number ALIGN_000464 and ALIGN_000465). As expected for a protein coding gene, cytochrome-b sequences align without gaps. Sequences were converted to amino acids using MACCLADE 4.0 (Maddison and Maddison 1992), and no stop codons were found. Within the region examined, 243 (51%) of sites are variable and 219 (46%) are parsimony informative. The distribution of the parsimony-informative sites is highly dependant on codon position: 23.7% in first, 6.8% in second, and 69.5% in third codon position. Most substitutions are synonymous, and translation of sequences to amino acids leads to a matrix with only 35 parsimony-informative sites (32.7% variable sites, among which 67.3% are parsimony informative). The average base composition of sequences is skewed, with little bias at first codon position, a deficiency of adenine (19.1%) and guanine (8.9%) and an overabun-dance of thymine (40.9%) for second position, and a strong bias in third position: deficiency of guanine (3.5%) and thymine (12.4%), and abundance of adenine (38.1%) and cytosine (46%). That bias in base composition does not differ significantly across taxa when the whole cytochrome-b fragment is considered (chi-squared test implemented in PAUP*, P > 0.05). The same result is obtained for first and second codon positions analyzed separately, but not for third codon position, where the test is significant at the 0.5% level (P = 0.019). Because heterogeneity in base composition across taxa is known to affect phylogenetic reconstruction (Lockart et al. 1994, Galtier and Gouy 1998), it was possible that bias in third codon position, which contains most of the sequence variation, may influence phylogenetic reconstruction. To consider that hypothesis, the most divergent taxa in GC content (more than twice the standard deviation) were identified for the sequences of third bases of cytochrome b (GC content average for all taxa = 49.5%, SD = 4.1): Corvus corone (40.8%), Alcippe chrysotis (57.8%), Chrysomma sinense (60.3%), Minla cyanouroptera (40.6%), and Lanius collaris (38.9%). Because none of those taxa were found to be sister taxa in the analysis, one can assume that the heterogeneity of base composition among taxa in the cytochrome-b sequence is not affecting the analysis. Saturation in sequences was also assessed by plotting uncorrected total sequence divergence versus divergence on the basis of transition and transversion for each codon position (Fig. 2). Those curves indicated that third positions experienced multiple transition substitution, shown by plateauing of the curve, whereas third position transversions did not show evidence of saturation. Those results indicated that it may be appropriate to test different positional or substitutional weighting schemes to the cytochrome-b sequence data.
For the RNA genes, alignment performed manually had a slightly different number of indels, but preliminary phylogenetic analyses performed with both alignments gave similar results (maximum-parsimony analyses, not shown). Therefore, the Clustal alignments were used for the rest of the study. Total percentage of sites that included gaps ranges from 8.9% for 12S to 9.5% for 16S. The proportion of sites with gaps is slightly higher for variable sites (15.5 and 22.5%) and parsimony-informative sites (16.5 and 26%, respectively). Both genes were analyzed with respect to secondary structure (see above for structural models), and results show that secondary structure has a major influence on sequence variation across the molecule, both on the pattern of base compositional bias and on the distribution of informative sites. The base composition bias is most visible in loops, with an increase of adenine (42.9 and 38% for 12S and 16S, respectively). It is less important in stems, with a slight increase of cytosine and guanine (30.4 and 27.3% for 12S; 27 and 31.7% for 16S). This result is typically found in RNA sequence analyses (Dixon and Hillis 1993, Morrison and Ellis 1997). This bias in base composition does not differ significantly across taxa for each gene, even when stems and loops are tested separately (chi-squared test implemented in PAUP*, P > 0.05). For both genes, most of the sequence variation is included in the loops. Among the 414 sites analyzed for 12S, 180 (43.5%) are variable and 133 (32%) are parsimony-informative, among which 88 (66%) are located in regions recognized as loops. Among the 543 sites analyzed for 16S, 204 (37.6%) are variable and 146 (26.9%) are parsimony-informative, among which 120 (82.2%) are located in loops. For both genes, most indels found in informative sites are located in loops (20 gapped sites in loops for 2 in stems in 12S; 34 gapped sites in loops for 4 in stems in 16S). Maximum-parsimony search was performed with the exclusion of gapped sites (89 sites in total): overall topology of the tree was not modified (tree not shown), but bootstrap support was slightly decreasing, as expected with the loss of informative characters. Treating gaps as fifth characters leads also to the same supported topology (maximum-parsimony analysis, tree not shown). Saturation curves (Fig. 2) indicate that for both genes and secondary structure elements, transversions increase linearly with overall distance. The same pattern is observed for transitions, except for pairwise distances including the most distant outgroup Tyrannus. Phylogenetic analysis was performed with and without this taxon to see if distant outgroup influences the topology. In all cases, the 16S and 12S sequences were analyzed with equal weights for all characters.
Saturation analyses. Pairwise transition (ti) and transversion (tv) sequence distance (uncorrected P distance) plotted against total sequence divergence for first (diamonds), second (squares), and third (triangles) codon positions (cytochrome b), for loops (diamonds) and stems (squares) (12S and 16S).
Phylogenetic analyses
The partition homogeneity test (ILD test), performed in PAUP* after removing invariant characters, detected no significant incongruence between cytochrome b and 12S data sets, as well as between the cytochrome b and 16S data sets (P > 0.05), but the test is significant at the 5% level for the comparison between 16S and 12S data sets (P = 0.02). However, Cunningham (1997) suggested that the critical value for rejection of congruence should be between 0.01 and 0.001 for the ILD test (but see Yoder et al. 2001). Therefore, I assumed that combining the three data sets should improve phylogenetic accuracy, and all subsequent analyses were performed using all data available. Maximum-parsimony analyses were conducted with different a priori weighting schemes for the cytochrome b partition: weighting of transversions over transitions of 2:1 and 5:1 for the whole cytochrome-b or only for the third codon position, elimination of transitions for the third codon position (−3ti), compared to unweighted analysis. The trees obtained differed slightly in topology, but always for weakly supported nodes. Nodes that appeared in every analyses are indicated in Figure 3 by thicker branches. The maximum-parsimony analysis without the more distant outgroup Tyrannus led to the same supported topology.
Maximum-parsimony combined analysis. Strict consensus of the 72 equal maximum-parsimonious trees (3,545 steps, CI = 0.2330, RI = 0.4478) obtained with PAUP* performed with the elimination of third position transitions for cytochrome b. Numbers represent percentage of bootstrap replicates and Bremer indices. Nodes present in every maximum-parsimony analyses (unweighted and weighted) are indicated by thicker branches.
Maximum-likelihood analysis was performed using first the program MODELTEST (see above). The selected model was the GRT + I + G model (general time-reversible), which allows a different probability for each of the six different substitution types, an estimation of the nucleotide frequency, an estimation of proportion of invariant sites (I), and a gamma-distributed rate variation across sites with the estimation of the shape parameter (α). The parameters used are the following: probabilities for the six substitution types Rmatrix = 1.2702, 3.3479, 1.2149, 0.1872, 7.8304, 1.0, proportion of invariable sites I = 0.5054, shape parameter α = 0.6803. The maximum-likelihood tree obtained using these parameters in PAUP* (Fig. 4) conflicts with the maximum-parsimony consensus tree for a few nodes, all of which are weakly supported by bootstrap values. Likelihood of the maximum-likelihood tree was compared with and without a molecular clock imposed, using a likelihood-ratio test LRT (Swofford et al. 1996), that assumes that the test statistic (2ΔL, where ΔL is the difference in log-likelihood between the clock and nonclock trees) follows a chi-squared distribution with N − 2 degrees of freedom when N is the number of taxa (Felsenstein 1988). The likelihood of the tree with a molecular clock imposed was significantly worse than the tree without a molecular clock (P < 0.001), suggesting rate variation among taxa.
Maximum-likelihood combined analysis. Tree obtained with a heuristic search implemented in PAUP* (−ln(L) = 22497.96937), with six substitution types (Rmatrix = 1.2702, 3.3479, 1.2149, 0.1872, 7.8304, 1.0), proportion of invariable sites I = 0. 5054, and gamma-distributed rate variation across sites α = 0. 6803.
Several nodes in Figures 3 and 4 have been numbered from 1 to 8 for reference in the following discussion. In both maximum-parsimony and maximum-likelihood topologies, all babblers are grouped together (node 1) with the exclusion of two taxa. The first one is a shrike babbler genus (Pteruthius). The two species stud-ied are sister taxa and are placed among the cor-void outgroups in all analyses. The second one is the Gray-chested Thrush Babbler (Kakamega poliothorax), which is grouped with thrushes and flycatchers according to the maximum-like-lihood topology (this node is unresolved in the maximum-parsimony tree). Inside the timaliids, two main clades are found in every analyses, nodes labeled 2 and 3. The first clade (node 2) includes the Sardinian Warbler (Sylvia melanocephala, Sylviidae), Wrentit (Chamaea fasciata), African Hill Babbler (Pseudoalcippe abyssinica), and seven Asian taxa belonging to the genera Alcippe (fulvettas), Paradoxornis (parrotbills), and Chrysomma sinense (Yellow-eyed Babbler). All other Asian and African species studied belong to the other main clade (node 3), which includes also the Japanese White-eye (Zosterops japonica, Zosteropidae), which is grouped within the yuhinas in all analyses with a good support (node 4). Average uncorrected sequence divergence between Zosterops japonica and the yuhinas is 7.55% (8.46% with Y. diademata and 6.64% with Y. gularis), which is the same level of divergence as among yuhinas themselves (7.13% between diademata and gularis).
Even though not all parts of the topology are well supported, results indicate that several currently recognized babbler genera may not be monophyletic. The data are particularly informative for two large babbler genera. The first one is the genus Alcippe (5 species studied out of 16), which are divided between two well-supported clades (nodes 2 and 3). The second genus is Garrulax, the laughingthrushes (9 species studied out of 51). The polyphyly of laugh-ingthrushes as defined by Berlioz (1930), that is, the genera Garrulax and Babax, is obtained in both maximum-likelihood and maximum-parsimony topologies, but with low bootstrap value and Bremer support. However, all the different tests indicated rejection of the null hypothesis (monophyly of the laughingthrushes), thus giving support for the polyphyly of that group (Table 2).
The different tests performed to test the null hypothesis (H0) of the laughingthrushes' monophyly.
The different tests performed to test the null hypothesis (H0) of the laughingthrushes' monophyly.
Discussion
Monophyly of the Timaliidae
The babbler group in its current definition (Sibley and Monroe 1990) forms a large group of very diverse taxa, and monophyly of the family was dubious. Analyses conducted here, as well as previous molecular studies, now give a more accurate view of the limits of the family. In a previous study using mitochondrial sequence data, we showed that the Malagasy “babblers” as well as the Asian White-bellied Yuhina (Yuhina zantholeuca) are not related to true Asian and African babblers (Cibois et al. 1999, 2001, 2002). Here, there is evidence that two taxa previously classified as babblers—one from Asia, the other from Africa—are not closely related to the core of the Timaliidae. The Asian “nonbabbler” group is the shrike babblers, genus Pteruthius (two of five species studied), which are placed among the outgroups in all analyses. Shrike babblers were originally described as shrikes (Lanius), because of their hooked bill (Vigors 1830–1831), but have been subsequently placed among babblers. Unlike typical babbler species, Pteruthius are sexually dimorphic. The study of the position of Pteruthius is beyond the scope of this article, because it requires a large number of corvoid representatives and will be presented elsewhere (S. Reddy et al. unpubl. data). The second nonbabbler taxon revealed in this analysis is the African species Kakamega. The systematic position of this taxon has been long debated, first being described among thrushes in the genus Alethe (Reichenow 1900). Some authors have recognized the morphological similarities of this species with thrushes (Sclater 1930, Delacour 1946), but absence of a spotted juvenal plumage led most ornithologists to put this bird among babblers, within Malacocincla (Chapin 1953), or Trichastoma (Deignan 1964, Hall and Moreau 1970, Morony et al. 1975), or in its own genus on the basis of song and morphological characters (Mann et al. 1978, Ripley and Beehler 1985). My data indicate that Kakamega poliothorax may be closely related to thrushes and flycatchers, even though it lacks the spotted juvenal plumage characteristic of most muscicapids (Pasquet et al. [1999] have shown that two other birds, Neocossyphus and Stizorhina— closely related to typical thrushes—do not possess this character either). Patterns of syringeal muscles have been used to infer relationships among putative turdine birds (Ames 1971, 1975). Information on the syrinx of Kakamega is not currently available, but it may provide more evidence in the future concerning placement of this species among muscicapoids.
All other babbler species studied here form a monophyletic group (node 1), with the inclusion of two taxa originally classified within other songbird families, Sylvia (Sylviidae) and Zosterops (Zosteropidae). Positions of those taxa in the phylogeny are discussed further below (nodes 2 and 4).
Node 2: Sylvia and related taxa
The position of Sylvia among babblers confirms results obtained by Sibley and Ahlquist (1990) using DNA hybridization data and is consistent with a previous cytochrome-b study including a small number of babblers (Fjeldså et al. 1999). That result indicates polyphyly of the Sylviidae, which has already been suggested by other studies using a larger sample of warblers (Sheldon and Gill 1996, Sibley and Ahlquist 1990, Sturmbauer et al. 1998, Cibois et al. 1999). In light of these independent results, a reevaluation of the phylogeny of warblers is required to clarify the systematics of that group. With the present data set, Pseudoalcippe abyssinica is the sister taxon of Sylvia, which was also the relationship obtained by Fjeldså et al. (1999) with a smaller taxon sampling. The taxonomy of Pseudoalcippe abyssinica, as well as its related species P. atriceps, has been long debated; those two taxa have been included, on the basis of morphological characters, either (1) in the African jungle babblers Illadopsis (Ripley and Beehler 1985, Howard and Moore 1994); (2) among the Asian genus Alcippe (Delacour 1946, Deignan 1964); or (3) in their own genus Pseudoalcippe (Bannerman 1923, Serle 1950, Dowsett and Forbes-Watson 1993). The phylogeny obtained here supports hypothesis (3), showing that Pseudoalcippe abyssinica is not closely related to either Illadopsis or Alcippe.
With 5 of 16 species in the genus Alcippe being studied, the fulvettas are strongly suspected to be polyphyletic, with two species (chrysotis and cinereiceps) in one clade (node 2), and three species (morrrisonia, poioicephala, and castaneceps) in another (node 3). The nonmonophyly of the fulvetta group is not surprising because of the morphological diversity of its members. A more detailed study of the systematics and phylogeny of the fulvettas will be presented in a forthcoming publication (E. Pasquet et al. unpubl. data).
Chamaea fasciata occurs in chaparral and other dense and bushy vegetation in western Oregon, California, and northwestern Baja California. On the basis of its superficial similarities with titmice, this species was originally placed in the Paridae (Hellmar 1903, Reichenow 1914). However, numerous characters link this bird to babblers: morphological characters like the round shape of wings, loose plumage, strong tarsus and bill, as well as ethological characters (clump together when perched, mutual preening) (Erickson 1938, Brown 1959). Several authors placed this species among the Timaliidae (Delacour 1946, Mayr and Amadon 1951, Mayr and Greenway 1956, Deignan 1964, Wolters 1975–1982, Howard and Moore 1994), and others in its own monotypic family (Rigway 1904, Wetmore 1960, Voous 1977). Sibley and Ahlquist's (1982) DNA hybridization results indicated a close affinity between Chamaea and other babblers, especially with Sylvia (Fig. 1). Analyses conducted here support placement of Chamaea among Timaliidae, in the same clade as Sylvia, but the topology inside this clade is poorly resolved (Figs. 3 and 4). The most strongly supported relationship indicates a position of Chamaea among a clade of Asian babblers, including several species of Alcippe, Paradoxornis, and Chrysomma sinense. This result is consistent with Delacour's Chamaeini tribe, which includes Paradoxornis, Chamaea, and Chrysomma (Table 1).
The genus Paradoxornis is not monophyletic in the maximum-likelihood topology, whereas the maximum-parsimony topology is unresolved. However polyphyly of the parrotbills seems dubious because of their very homogeneous bill morphology (short and laterally compressed). Further studies including a larger sample of parrotbills should resolve relationships within that group.
Node 4: Yuhinas and white-eyes
Phylogenetic relationships and systematics of the yuhinas have been studied in detail in a previous publication, which dealt also with the tree babblers of the genus Stachyris (Cibois et al. 2002). A novel result obtained with an extensive data set is the placement of Zosterops japonica among yuhinas. That result challenges all previous classifications, because a close relationship between white-eyes and babblers has never been suspected before on the basis of morphological or ecological characters. Using nuclear markers, Barker et al. (2002) found also a close relationship between Sylvia, Garrulax, and Zosterops, but a larger sampling was necessary to define phylogenetic position of Zosterops among babblers. Yuhinas occur exclusively in the Indo-Malayan region, with a large diversification in the Philippine archipelagos (species previously classified among the genus Stachyris). The genus Zosterops is widely distributed in the Old World, both in Asia (including the Philippines) and Africa, and also in cis-Wallacea regions (New Zealand, New Guinea, Australia, and Indonesian islands on the Australian side of Wallace's Line), as well as Pacific islands. Because of their adaptation to nectar feeding, white-eyes share morphological similarities with the Meliphagidae, Dicaeidae, and Nectariniidae, and those groups were placed sequentially in most classifications (Mayr and Greenway 1956, Paynter 1967, Wolters 1975–1982). However, DNA hybridization studies suggested that those similarities are due to convergence and not to close relationships (Sibley and Ahlquist 1990). In Sibley and Ahlquist's phylogeny, white-eyes are closely related to sylvioids (i.e. Old World warblers and babblers). Thus, Zosteropinae are sister taxa to the group including two warbler lineages, Acrocaphalinae and Megalurinae, along with the babblers + Sylvia assemblage (Sylviinae). The hypothesis of a close relationship between white-eyes and yuhinas suggests an Asian origin for Zosterops, followed by dispersal to Africa and the cis-Wallacea region and Pacific islands. Further work is required to clarify the biogeography of this group, and also to investigate the phylogenetic relationships of the others members of the Zosteropidae (nine genera, mainly monotypic), most of which are restricted to the cis-Wallacean region (Mees 1957, 1961, 1969). Slikas et al. (2000) performed a phylogenetic analysis for several Micronesian white-eyes, but did not investigate the position of white-eyes among sylvioids.
Node 5: Fulvettas
The polyphyly of the fulvettas has been discussed previously (node 2). The two species that form this clade (morrisonia and poioicephala) have very similar plumage patterns, and their close relationship according to molecular characters is consistent with their morphological similarity.
Node 6: Timalia and related taxa
This clade regroups some, but not all, representatives of two different tribes defined by Delacour (1946), Pomatorhinini and Timaliini (Table 1). The Pomatorhinini include two groups of Asian babblers, scimitar babblers and wren babblers. Scimitar babblers (Pomatorhinus and Xiphirhynchus) are mainly characterized by a long curved bill. Wren babblers (eight genera including Spelaeornis, Napothera, Jabouillea, and Kenopia) share a general similarity with wrens of the genus Troglodytes, particularly in having a small size, a brown cryptic plumage, that is stripped or spotted, a short tail, and the habit of living on the ground or in dense bushes. Two genera, Rimator and Jabouillea, are supposed to represent the transition between scimitar and wren babblers, because they share some characteristic of both groups. Nevertheless, all analyses performed in this study indicated that the Pomatorhinini are not monophyletic, because the scimitar babblers are not the sister group of any wren babblers, and the wren babbler group itself is polyphyletic. The monophyly of scimitar babblers (3 of 10 species studied here) is not supported (irresolution in the maximum-parsimony tree, and short branch in the maximum-likelihood topology; Figs. 3 and 4). However, both topologies indicate a close relationship between the White-browed (Pomatorhinus schisticeps) and Slender-billed (Xiphirhynchus superciliaris) scimitar babblers. Spelaeornis is the only member of the polyphyletic wren babblers included in clade 6. The polyphyly of the wren babblers indicates that the similarity of those birds is due to convergence and does not result from a close phylogenetic relationship.
Stachyris, Timalia, and Macronous are said to be members of the same tribe Timaliini, on the basis of similarities in size, shape, and habits, and consistently they are sister taxa in all topologies obtained here. However, the tree babblers (Stachyris) do not form a monophyletic group (three species studied). That result is consistent with a previous study that included a larger sampling of tree babblers, in which the genus was shown to be polyphyletic (Cibois et al. 2002).
Node 7: Jungle babblers and related taxa
This clade clusters, with low bootstrap and Bremer support (61%, 2 steps), members of three different tribes: Pellornini, Pomatorhinini, and Turdoidini (Table 1). The Pellornini include the jungle babblers, which inhabit African and Asian tropical forests. All jungle babblers stud-ied here (seven species in four genera) belong to clade 7, but few nodes are resolved within this group. The only strongly supported node reveals the monophyly of the African jungle babblers (Illadopsis). They occur in forests of west and central Africa, and were often placed among Asian babblers in the genus Trichastoma (Delacour 1946, Deignan 1964, Morony et al. 1975). The relationships between Illadopsis and the Asian jungle babblers (Trichastoma, Malacocincla, and Malacopteron) are uncertain in all analyses performed here, as well as are relationships among the Asian taxa themselves.
Representatives of the Pomatorhinini in this clade include five wren babblers placed in Napothera, Jabouillea, or Kenopia. There is no bootstrap support for their relationships inside the clade 7. The only clades with a bootstrap support superior to 50% include, on the one hand, the Streaked (Napothera brevicaudata) and Mountain (N. crassa) wren babblers (66% and 3 steps); and on the other, the Eyebrowed Wren Babbler (N. epilepidota) and Short-tailed Scimitar Babbler (Jabouillea danjoui) (61%, 2 steps). Jabouillea is a monotypic genus, restricted to central Vietnam, known from only very few specimens (stored at the British Museum, London; Muséum National d'Histoire Naturelle, Paris; and American Museum of Natural History, New York). Jabouillea danjoui was originally described among the genus Rimator, from which it differs by a larger size and a different bill shape (Robinson and Kloss 1919, Delacour 1927). No tissue sample was available for Rimator, but the phylogeny obtained here suggests that Jabouillea is closely related to another wren babbler, the genus Napothera, with which it shares the general “wren-like” morphology and terrestrial habits.
Two wren babblers species, Napothera crassa and Kenopia striata (the Striped Wren Babbler) are endemic to the island of Borneo. Napothera crassa is sister taxon with good support to N. brevicaudata, which occurs in a large portion of Asia, from northeast India to Malaysia. The position of the monotypic genus Kenopia is not supported in this analysis, but the topology obtained suggests that wren babblers from Borneo may not have a unique origin. Species of the genus Napothera have a distribution pattern similar to the one found in yuhinas, with some widespread species in the Asian mainland, and with several taxa endemic to the Greater Sundas and the Philippine archipelagos. A previous study on yuhinas has shown that the Philippine endemics are related to the Bornean endemic species, that group being sister taxon to species from the mainland (Cibois et al. 2002). Further studies will possibly show whether a similar phylogenetic pattern can be found among Napothera species, thus giving support from two independent lineages for a hypothesis of colonization of southeast Asian islands.
Two babblers found in clade 7 were classied among the Turdoidini tribe by Delacour: Rufous-winged Fulvetta (Alcippe castaneceps) and White-hooded Babbler (Gampsorhynchus rufulus). The positions of those two taxa are not consistent between the maximum-parsimony and maximum-likelihood topologies, but in both cases their placement has no bootstrap support and small Bremer indices (<50%, 2 steps). Polyphyly of the genus Alcippe has been discussed previously (see nodes 2 and 4). Gampsorhynchus, a monotypic genus, occurs in a large part of Indo Malayan Asia. It possesses a very distinctive plumage pattern (brown upperparts and white head and belly) and was sometimes placed among the shrike babblers because of its hooked bill. Phylogenetic affinities of that particular species among babblers are still unclear.
Laughingthrushes
Berlioz (1930) has conducted the most comprehensive revision of laughingthrush systematics, a group of babblers found only in Asia, with a large number of species occurring in the Himalayas. He defined three species groups on the basis of bill morphology, plumage coloration, and habitat characteristics: the Ianthocincla species group (nine species including those in the genus Babax), the Trochalopterum species group (16 species), and the Garrulax species group (20 species). The taxonomic sequence proposed by Berlioz (1930) has been mainly followed by Delacour (1946) and subsequent authors. However, the different tests performed here gave support for the polyphyly of laughingthrushes (see above), suggesting that the genera Garrulax and Babax need to be reevaluated in the near future. For the moment, only the nodes found in all analyses can be compared to Berlioz's (1930) classification. Among the nine Garrulax species studied here, four were placed in the same Trochalopterum species group (erythrocephalus, variegatus, squamatus, and subunicolor). All except the Variegated Laughingthrush (G. variegatus) possess similar dark scaling on plumage. The monophyly of that subgroup is obtained in both maximum-likelihood and maximum-parsimony topologies, but with no support (<50%, 1 step). Inside that subgroup, G. variegatus is the sister species of the Chestnut-crowned Laughingthrush (G. erythrocephalus) (97%, 8 steps), and the Blue-winged (G. squamatus) and Scaly (G. subunicolor) laughingthrushes are also sister taxa (54%, 1 step). The genus Babax comprises three species, all characterized by a brown or grey striped plumage. Their general shape is similar to Garrulax species, but Babax species possess a longer and more curved bill. The results of our analysis suggest that the Chinese Babax (B. lanceolatus) may be closely related with two Garrulax species, the White-throated (G. albogularis) and the White-browed (G. sannio) laughingthrushes (81%, 3 steps). Berlioz (1930) also placed Babax species within the Garrulax, but he considered them to be related to members of the Ianthocincla sub-group, represented here by only one species, the White-spotted Laughingthrush (G. ocellatus), whose position in the phylogeny is uncertain (<50%, 1 step). The last supported clade among laughingthrushes clusters the White-crested (G. leucocephalus) and Black-hooded (G. milleti) laughingthrushes (100%, 9 steps), which were placed by Berlioz in the same group as G. albogularis and G. sannio. But that clade has no bootstrap support and small Bremer indices in the maximum-parsimony topology (<50%, 2 steps) and is not present in the maximum-likelihood topology. Thus, groups proposed my Berlioz are poorly supported by results of our analysis, but my data set includes a limited number of laughingthrush species, which represent only parts of the diversity of that group. The phylogeny obtained in this study should be considered as a first attempt to comprehend the diversity of these babblers. However, the position of laughingthrushes inside the large clade 2 (Figs. 3 and 4) differs from Sibley and Ahlquist's (1990) results, in which Garrulax was placed as the sister group to all other babblers (Fig. 1). Furthermore, the classification proposed by Sibley and Monroe (1990) is not consistent with the phylogeny obtained here, inasmuch as my results showed no close relationship between Garrulax and Liocichla, the two genera that form the subfamily Garrulacinae in Sibley and Monroe's checklist. Berlioz (1930) also supposed that Liocichla phoenicea was embedded among laughingthrushes (in the Trochalopterum group), but the analysis performed here does not support that hypothesis.
Two other babblers, the Arrow-marked Babbler (Turdoides jardineii) and the Cutia (Cutia nipalensis), are sister taxa to laughingthrushes in the maximum-parsimony tree, but with no bootstrap support and a Bremer index of 2 steps. The genus Turdoides includes 26 species found in west Asia, the Middle East, and Africa. All species have a brown or gray plumage that is often striped or scaled, recalling Babax, but with a different bill shape. On the basis of those similarities, Delacour (1946) suggested that Turdoides, Babax, and Garrulax may be related, but the results of my study do not confirm that hypothesis. The monotypic Asian genus Cutia was sometimes thought to be related to shrike babblers (Pteruthius), on the basis of its very slightly hooked bill shape, the distinctive plumage between males and females, and the habit of living in high arboreal levels. Those similarities are in fact due to convergence, because the results here indicate that Cutia is not closely related to Pteruthius but belongs to the large core of Timaliidae. However, the phylogenetic relationships of that taxon within babblers are still uncertain, and the position of Cutia nipalensis among laughingthrushes has to be verified in future work.
Node 8: Song babblers
This clade includes with good support (97%, 9 steps) 11 babblers previously classified among the song babbler group (Turdoidini), including barwings (Actinodura), minlas (Minla), sibias (Heterophasia), liocichlas (Liocichla), mesia and leiothrix (Leiothrix). Relationships among those birds are not fully resolved, and a few nodes have good support. The two Actinodura species studied here are closely related and are sister taxa to the Blue-winged Minla (M. cyanouroptera) in both topologies with low support (59%, 1 step). Harrison (1986a) suggested placing cyanouroptera in the genus Leiothrix or in its monotypic genus Siva on the basis of its differences with the other Minla species (e.g. its blue plumage coloration and different tail shape). The phylogeny obtained here does not support a close relationship between M. cyanouroptera and either Leiothrix or the other Minla, including the type species of the genus, the Red-tailed Minla (M. ignotincta). However, the position of M. ignotincta and the Chesnut-tailed Minla (M. strigula) are not consistent in maximum-parsimony and maximum-likelihood topologies, and taxonomic modifications must await further studies on that genus. Heterophasia (two species studied among six) is sister taxon to Leiothrix with low support (62%, 2 steps). Most of those birds exhibit colorful plumage, and several species are popular cage birds. The position of Liocichla (two of three species studied) inside the song babblers clade is not consistent in all analyses, therefore no phylogenetic relationships can be inferred now for those birds.
Naming the babbler group
The close relation-ship between Sylvia and babblers leads to the nomenclatural problem of naming the babbler and warbler families. The present study and previous ones suggest that the Sylviidae are polyphyletic, first because of the position of Sylvia among babblers, and second because the other warblers do not form a single monophyletic group. The implication of those results for the systematics of both babblers and warblers has been reviewed in detail elsewhere (A. Cibois unpubl. manuscript), where the conditional suppression of the Sylviidae with respect to Timaliidae is proposed. If in the future babblers and warblers are placed by some authors altogether in the same family, then this group should be called Sylviidae as a strict application of the priority rule. However, if as it seems likely, the Timaliidae and several groups of warblers are recognized at the same family level, then the family-group name Sylviidae Leach, 1820 should be suppressed and the name Timaliidae Vigors and Horsfield, 1827 kept for the babblers and Sylvia. Other names should be used for the different warbler groups, chosen for instance among the 26 synonyms listed for Sylviinae by Bock (1994)
Acknowledgments
For providing tissue samples I thank M. Kalyakin (Zoological Museum of Moscow Lomonosov State University, Russia, and Russian-Vietnamese Tropical Research Center, Vietnam); J. Fjeldså and J. Garcia-Moreno (Institute of Zoology, Copenhagen); O. Kobkhet (Kasetsart University, Bangkok); F. Baillon (ORSTOM Yaoundé, Cameroon); F. Sheldon and R. Moyle (Museum of Natural Science, Louisiana State University); C. Cicero (Museum of Vertebrate Zoology, University of California, Berkeley); P. Sweet and J. Groth (American Museum of Natural History, New York). I am very grateful to E. Pasquet for his help during all stages of this study. J. Cracraft, P. Sweet, S. Hackett, and two anonymous reviewers provided helpful comments on this manuscript. This work was conducted at and supported by the team “Systématique et Evolution des Vertébrés Tétrapodes” (EA No. 2586, MNHN) and the Service de Systématique Moléculaire (CNRS FR 1541, MNHN). Fieldworks in Asia were supported by PPF/MNHN “Biodiversité de la Faune et de la Flore de l'Asie du Sud-Est.” A.C. was supported on a Frank M. Chapman postdoctoral fellowship at the American Museum of Natural History while finishing this work. This article is a contribution from the Monell Molecular Laboratory and the Cullman Research Facility in the Department of Ornithology, American Museum of Natural History, and has received generous support from the Lewis B. and Dorothy Cullman Program for Molecular Systematics Studies, a joint initiative of The New York Botanical Garden and The American Museum of Natural History. The protocol suggested by Anderson, Goldman, and Rodrigo on Goldman's website can be accessed at www.zoo.cam.ac.uk/zoostaff/goldman/tests.
Literature Cited
APPENDIX
APPENDIX: List of species, sample numbers, GenBank accession numbers, locality, and taxonomic family (following Peters' Checklist, Deignan 1964). AMNH: American Museum of Natural History; New York; MNHM: Museum National d'Histoire Naturelle, Paris; MVZ: Museum of Vertebrate Zoology, Berkeley; LSU-MNS: Louisiana State University Museum of Natural Science, Baton Rouge; MZC: Museum of Zoology, Copenhagen. Asterisks indicate the species selected in the reduced data set of 40 taxa.
APPENDIX: List of species, sample numbers, GenBank accession numbers, locality, and taxonomic family (following Peters' Checklist, Deignan 1964). AMNH: American Museum of Natural History; New York; MNHM: Museum National d'Histoire Naturelle, Paris; MVZ: Museum of Vertebrate Zoology, Berkeley; LSU-MNS: Louisiana State University Museum of Natural Science, Baton Rouge; MZC: Museum of Zoology, Copenhagen. Asterisks indicate the species selected in the reduced data set of 40 taxa.
