PHYLOGENETIC RELATIONSHIPS of the WRENTIT BASED on MITOCHONDRIAL CYTOCHROME B SEQUENCES

The Condor 104:740±749 ᭧ The Cooper Ornithological Society 2002

PHYLOGENETIC RELATIONSHIPS OF THE WRENTIT BASED ON MITOCHONDRIAL CYTOCHROME b SEQUENCES

DINO N. BARHOUM AND KEVIN J. BURNS1 Department of Biology, San Diego State University, San Diego, CA 92182-4614

Abstract. The phylogenetic relationship of the Wrentit (Chamaea fasciata) to other passerine birds is understood poorly. A variety of taxa have been proposed as closely related to the Wrentit, but in general this species is placed in a monotypic taxon to emphasize its morphological distinctiveness and lack of clear relationship to other groups of birds. We used the mitochondrial cytochrome b gene to infer the relationship of this monotypic genus to other avian groups including representatives of Sylvioidea, Muscicapoidea, and Passe- roidea. Results of this study corroborate in part earlier proposals based on DNA-DNA hybridization. Maximum parsimony, maximum likelihood, and Bayesian phylogenetic analyses support the Wrentit, Sylvia, and babblers (Garrulax, Illadopsis, Leiothrix, Pomatorhin- us, Stachyris, and Yuhina) as a clade, with the Wrentit sharing a more recent common ancestor with Sylvia than with babblers. A Sylvia and Wrentit association is further supported by similarities in morphology and ecology. In addition to ®ndings on Wrentit relationships, our study in general agrees with the major groups of oscine passerines identi®ed by earlier DNA hybridization studies. Key words: babbler, Chamaea fasciata, mitochondrial DNA, phylogeny, Sylvia, Wrentit.

Relaciones FilogeneÂticas de Chamaea fasciata Basadas en Secuencias del Citocromo b Mitocondrial Resumen. La relacioÂn ®logeneÂtica de Chamaea fasciata con otros paserinos es vaga- mente entendida. Una variedad de taxa ha sido propuesta como relacionada cercanamente a C. fasciata, pero en general esta especie es situada en un taxoÂn monotõÂpico para enfatizar su morfologõÂa distintiva y la falta de claridad en la relacioÂn con otros grupos de aves. Utilizamos el gen del citocromo b mitocondrial para inferir la relacioÂn de este geÂnero mo- notõÂpico con otros grupos de aves incluyendo representantes de Sylviodea, Muscicapoidea y Passeroidea. Resultados de este estudio corroboran parcialmente propuestas anteriores basadas en hibridacioÂn de ADN-ADN. AnaÂlisis ®logeneÂticos de maÂxima parsimonia, de maÂxima probabilidad y bayesianos respaldan a C. fasciata, Sylvia, Garrulax, Illadopsis, Leiothrix, Pomatorhinus, Stachyris y Yuhina como un clado, con C. fasciata compartiendo un ancestro comuÂn maÂs reciente con Sylvia. Una asociacioÂn entre Sylvia y C. fasciata es respaldada por similitudes en morfologõÂa y ecologõÂa. En adicioÂn a los descubrimientos en las relaciones de C. fasciata, nuestro estudio coincide en general con los grupos mayores de paserinos oscinos identi®cados en estudios previos mediante hibridacioÂn de ADN.

INTRODUCTION blers (Timaliidae), and Old World warblers The Wrentit (Chamaea fasciata) is a small song- (Sylviidae) have all been considered the Wren- bird that frequents brushy habitats from northern tit's closest relatives (Grinnell 1913, Sibley Baja California to Oregon. This species is un- 1970, Sibley and Ahlquist 1982, 1990). Reasons usual among North American birds in that it ap- for this uncertainty stem from the lack of de®n- parently is not closely related to any other New able similarities in behavior, ecology, and mor- World species. The evolutionary relationship of phology to other North American species. Be- the Wrentit to other birds has been debated since cause of its lack of clear relationship to other Gambel (1845) ®rst described the species. Tit- birds, the Wrentit has been assigned to a mono- mice (Paridae), wrens (Troglodytidae), bushtits typic taxon at various taxonomic levels (genus, (Psaltriparus), mockingbirds (Mimidae), bab- tribe, family). Most recent classi®cations consider the Wren- tit to be related to Old World warblers and bab- Manuscript received 5 February 2002; accepted 23 blers (traditionally de®ned as the families Sylvii- July 2002. 1 Corresponding author. E-mail: kburns@sunstroke. dae and Timaliidae, respectively; Wetmore sdsu.edu 1960, Howard and Moore 1991). For example,

[740] WRENTIT PHYLOGENETIC RELATIONSHIPS 741

Sibley (1970) compared electrophoretic mobility World warblers are found in three families of the of egg-white proteins and was unable to un- Sylvioidea, including Sylviidae. Sibley and equivocally de®ne the sister taxon, but narrowed Monroe (1990) also placed most traditional bab- the possibilities to either Timaliidae or Sylvii- blers into two of the four subfamilies of Sylvii- dae. However, this result is not very enlighten- dae (Garrulacinae and Sylviinae). One of these ing because the limits and monophyly of these subfamilies (Sylviinae) includes the majority of two families are uncertain (Sibley and Ahlquist the babblers (tribe Timaliini), the Old World 1982, 1990, Cibois et al. 1999). Babblers are a warbler genus Sylvia (tribe Sylviini) and the diverse group that primarily inhabit tropical ar- Wrentit (tribe Chamaeini). These results were eas of Asia and Africa. They do not have well- consistent with Sibley and Ahlquist's (1982) ear- recognized synapomorphies, and genera without lier DNA hybridization study in which they con- other obvious taxonomic af®nities have been included that the Wrentit is approximately equi- cluded among the babblers (Sibley and Monroe distant from babblers and from Sylvia. Thus, 1990, Howard and Moore 1991). Although bab- DNA hybridization evidence indicates that the bler monophyly has been questioned, Cibois et Wrentit is closely related to the genus Sylvia and al. (1999, 2001) demonstrated the monophyly of to a clade containing most traditional babblers several babblers including species in the genera (Timaliini). Garrulax, Illadopsis, Leiothrix, Pomatorhinus, Although the classi®cation of Sibley and Stachyris, and Yuhina. Old World warblers Monroe (1990) is becoming more widely used, (Sylviidae) are another extensive radiation of several authors (e.g., Harshman 1994) have crit- birds that occupy a variety of niches and are icized some of the methods of analysis used by abundant in Europe, Asia, and Africa. Although Sibley and Ahlquist (1990) to derive this tax- the group is not monophyletic (Sibley and Ahl- onomy. Several recent DNA sequencing studies quist 1990), recent systematic studies have iden- (Groth 1998, Ericson et al. 2000, Lovette and ti®ed several monophyletic assemblages within Bermingham 2000, Barker et al. 2001) have di- Old World warblers. These include the reed war- rectly tested predictions made from these DNA blers (Acrocephalus, Chloropeta, and Hippolais) hybridization studies. However, none of these and some leaf warblers (Phylloscopus), based on studies have addressed the phylogenetic position studies by Helbig and Seibold (1999) and Price of the Wrentit. In this study, we use mitochon- et al. (1997, 1998), respectively. drial DNA sequence data to identify the closest The only recent, comprehensive systematic living relative of the Wrentit. In addition, we study of both Timaliidae and Sylviidae that in- compare our phylogenies of oscine passerine re- cluded the Wrentit was the DNA-DNA hybrid- lationships based on DNA sequence data to the ization study of Sibley and Ahlquist (1982, DNA-DNA hybridization phylogenies produced 1990). The taxonomic classi®cation based on by Sibley and Ahlquist (1982, 1990) and more Sibley and Ahlquist (1990) was outlined in Sib- recent DNA sequencing studies. ley and Monroe (1990). Sibley and Ahlquist (1990) found the families Sylviidae and Tima- METHODS liidae as traditionally de®ned (Wetmore 1960, TAXON SAMPLING Howard and Moore 1991) to be polyphyletic. Thus, Sibley and Monroe (1990) altered the Because of the taxonomic uncertainty surround- classi®cation of these groups. Polyphyly of the ing higher-level oscine passerine relationships, a traditionally recognized Sylviidae and Timalii- broad sample of taxa was necessary to investi- dae was indicated by the inclusion of the genus gate phylogenetic relationships of the Wrentit. Pomatostomus, traditionally a babbler, in the Forty-four representatives of parvorder Passeri- parvorder Corvida and close association of the da (sensu Sibley and Monroe 1990) were sam- remaining babblers (tribe Timaliini) to Sylvia (an pled, including representatives of all three Pas- Old World warbler genus). Both Old World war- serida superfamilies (Muscicapoidea, Passeroi- blers and babblers are included in Sibley and dea, and Sylvioidea). Fourteen genera of Old Monroe's (1990) superfamily Sylvioidea (199 World warblers (Acrocephalus, Bradypterus, genera, 1195 species). Sylvioidea consists of 11 Cettia, Chloropeta, Cisticola, Hippolais, Hyper- families, one of which Sibley and Monroe gerus, Locustella, Orthotomus, Phylloscopus, (1990) designated as Sylviidae. However, Old Regulus, Scotocerca, Sylvia, and Thamnornis) 742 DINO N. BARHOUM AND KEVIN J. BURNS and seven genera of babblers (Illadopsis, Garru- isolated using Chelex௡ (5% solution, incubated lax, Leiothrix, Neomixis, Pomatorhinus, Stachy- for 20 min at 95ЊC, Walsh et al. 1991) from fresh ris and Yuhina) were chosen to represent the liver, heart, or breast muscle stored at Ϫ80ЊC. Sylvioidea. Seven genera of Muscicapoidea Cytochrome b was ampli®ed using three primer (Catharus, Copsychus, Erithacus, Ficedula, pairs (H15298/L14830, H1570/L15184, and Muscicapa, Toxostoma, and Turdus) and six H16108/L15635; Groth 1998) for three overlap- genera of Passeroidea (Anthus, Melospiza, Mo- ping fragments. Initial ampli®cation was per- tacilla, Passerella, Piranga, and Vidua) were formed in 10-␮L reactions, in capillary tubes for sampled as well. Two Wrentit exemplars from 40 cycles of 94ЊC for 3 sec, 43ЊC for 1 sec, 71ЊC opposite ends of its distribution (southern Cali- for 30 sec in a Rapid Cycler (Idaho Technology, fornia and Oregon) were included. Because of Inc., Salt Lake City, Utah). Desired PCR prod- the lack of conclusive evidence for relationships ucts were veri®ed on 2.3% agarose gels and ex- among Muscicapoidea, Passeroidea, and cised using disposable glass pipets. Reampli®- Sylvioidea, representatives of the parvorder Cor- cation of melted plugs (74ЊC for 20 min in 250 vida (Aphelocoma, Corvus, Lophorina, Poma- ␮LH2O) was also performed in capillary tubes tostomus, and Terpsiphone) were included to in a Rapid Cycler but in larger reaction volumes root relationships among these Passerida super- for longer denaturing times and at higher tem- families. Multiple exemplars were taken from peratures (40-␮L reactions, 94ЊC for 12 sec, Corvida, Muscicapoidea, Passeroidea, and 52ЊC for 4 sec, 71ЊC for 30 sec). Double-strand- Sylvioidea to break up long branches. This ex- ed product was puri®ed using a BIO 101 gene- emplar approach is appropriate because previous cleaning kit and subjected to cycle sequencing studies have narrowed the relationships among (ABI Prism௢ Dye Terminator Cycle Sequencing the groups of interest to Sylvioidea, and the in- Ready Reaction kit with AmpliTaq DNA Poly- clusion of other groups (Passeroidea, Muscica- merase; Perkin Elmer, Foster City, California). poidea, and Corvida) facilitated the rooting of Cycle sequencing was done in 10-␮L reaction Sylvioidea and the arrangement of taxa within volumes with a DNA Engine (MJ Research, this group. Previous studies with more inclusive Inc., Reno, Nevada) under the following condi- data sets were used as a guideline to choose a tions: 96ЊC for 1 min and then 28 cycles of 96ЊC pertinent subset of taxa to use in this study (i.e., for 30 sec, 50ЊC for 15 sec, and 60ЊC for 4 min. single exemplars for genera or other higher-level DNA sequencing was performed on an ABI 377 groups). For example, Cibois et al. (1999, 2001) automated sequencer. Accuracy of DNA se- evaluated relationships among some babblers quencing was ensured in two ways: sequencing (but did not include the Wrentit), Helbig and both heavy and light strands of all PCR frag- Seibold (1999) resolved relationships between ments and using overlapping fragments of the reed warbler species (Acrocephalus, Chloropeta target sequence (approximately 15% of the total and Hippolais), and Price et al. (1997, 1998) sequence). DNA sequences were edited and studied relationships among leaf warbler species linked using Sequencher௢ 3.0 (Gene Codes (Phylloscopus). Corp., Ann Arbor, Michigan). Completed cytochrome b sequences were aligned and translated DNA SEQUENCING using Se-al v1.0a1 (Rambaut 1995). Mitochondrial cytochrome b gene sequences were used to infer Wrentit relationships. Cyto- PHYLOGENETIC ANALYSES chrome b has been used extensively in other avi- Phylogenetic analyses were performed under an phylogenetic studies and has been demon- maximum parsimony and maximum likelihood strated to be a useful phylogenetic marker for as implemented in PAUP* (version 4.0b8; Swof- resolving generic and family level questions ford 2001). Maximum parsimony (MP) analysis (Avise et al. 1987, Edwards et al. 1991, Moore was performed under multiple models of evo- 1995, Moore and DeFilippis 1997). In total, 51 lution through the use of the following weight- sequences were used. Some of these sequences ing schemes: (1) equal or uniform weighting, (2) are new to this study (Table 1, GenBank acces- transition/transversion weighting, (3) codon po- sion numbers AY124538±AY124547) and others sition partitioning, and (4) six-parameter (nucle- were obtained from previous studies (see Ap- otide) step matrix. The four different weighting pendix for these GenBank numbers). DNA was schemes were used to test sensitivity of parsi- WRENTIT PHYLOGENETIC RELATIONSHIPS 743

TABLE 1. Species names, voucher numbers, GenBank numbers, and locality information of sequences not previously reported. MVZ ϭ Museum of Vertebrate Zoology at the University of California at Berkeley, FMNH ϭ Field Museum of Natural History, SDSU ϭ San Diego State University Vertebrate Collections.

Voucher GenBank Species number number Locality Bradypterus cinnamomeus FMNH 355750 AY124541 Uganda: Kasese, Nyabitaba, 10 km NW Ibanda, Mubuku Valley, Rwen- zori Mts. Chamaea fasciata SDSU 2189 AY124547 USA: California, San Diego Co., 4 miles west Corte Madera Mt., 32Њ45.2ЈN, 116Њ39.7ЈW Chamaea fasciata SDSU 2225 AY123546 USA: Oregon, Lincoln Co., 6 miles southeast of Tidewater 44Њ21.5ЈN, 123Њ48.2ЈW Illadopsis pyrrhopterum FMNH 355676 AY124539 Uganda: Kasese, Nyabitaba, 10 km NW Ibanda, Mubuku Valley, Rwen- zori Mts. Passerella iliaca MVZ 170456 AY124544 USA: California, Lake Co., Hull Mt. 4 miles S and 1 mile E Piranga ludoviciana SDSU 2212 AY124545 USA: California, Fresno Co., 4 miles west of Hume Lake, 36Њ47.8ЈN, 118Њ59.2ЈW Stachyris whiteheadi FMNH 433007 AY124542 Philippine Islands: Luzon, Camarines Sur, Mt Isarog Sylvia borin FMNH 385182 AY124540 Uganda: Kabale, Byumba Terpsiphone cinnamomea FMNH 358456 AY124538 Philippine Islands: Sibuyan, Mt. Gui- tinguitin, NW slope, 4.5 km S, 4 km E Magdiwang Toxostoma redivivum SDSU 2281 AY124543 USA: California, San Diego Co., 4 miles west Corte Madera Mt., 32Њ45.2ЈN, 116Њ39.7ЈW mony results to model and model parameters. with tree-bisection-reconnection (TBR) branch Equal weighting and codon position partitioning swapping). All parsimony analyses were boot- involved weighting characters. In equal weight- strapped for 100 replicates, with 10 random-ad- ing, all characters were assigned a weight of dition replicates for each bootstrap replicate. one. Codon position partitioning involved as- The second optimality criterion implemented signing ®rst and second positions a weight of for phylogenetic analysis was maximum likeli- ®ve and third positions a weight of one. Tran- hood (ML). Simulation studies have shown ML sition/transversion weighting and six-parameter to be a consistent estimator of phylogeny that is weighting involved weighting character-state robust with respect to small changes of model transformations through use of a step matrix. For and model parameter estimates (Huelsenbeck transition/transversion weighting, we assigned 1995, Yang 1996). Maximum likelihood takes transversions ®ve times the weight of transitions. into account important attributes of molecular Character transformation weights for the six-pa- evolution (i.e., unequal base frequency and rameter step-matrix substitution model were es- among-site rate variation) dif®cult to incorporate timated on a parsimony tree in MacClade 3.07 into parsimony analysis. Selection of a model of (Maddison and Maddison 1997) based on equal sequence evolution for maximum-likelihood weights. Thus, for the six-parameter analysis, analysis was explored in a hierarchical likeli- the following weights were used: transforma- hood framework using ModelTest v3.0 (Posada tions between A and C ϭ 3.3, between A and G and Crandall 1998). Fifty-six nested models of ϭ 1.0, between A and T ϭ 4.3, between C and DNA sequence evolution were evaluated, using G ϭ 4.3, between C and T ϭ 1.0, and between the likelihood-ratio test statistic with Bonferroni G and T ϭ 5.3. All MP analyses were imple- correction, to ®nd the simplest model that yields mented under multiple, independent heuristic a signi®cantly higher likelihood. Initial topolo- searches (100 random taxa-addition replicates gies for estimating rate parameters and evaluat- 744 DINO N. BARHOUM AND KEVIN J. BURNS ing substitution models for ModelTest v3.0 were ities are true probabilities of clades with values generated from three trees. The equal-weight of 95% or greater deemed signi®cantly support- parsimony tree, the six-parameter parsimony ed. tree (with codon partitioning) and a neighbor- joining tree (the default setting for ModelTest) RESULTS were used to test sensitivity of model selection SEQUENCE VARIATION to starting tree conditions. Using a successive All sequences aligned without gaps or inser- approach to obtain the optimal ML phylogeny tions, as expected for a protein-coding gene. Of (Wilgenbusch and de Queiroz 2000), subsequent the 1143 total aligned nucleotide positions, 555 trees generated from the three initial ML search- (49%) characters were constant, 100 (9%) were es were used to re-estimate models and param- variable but uninformative, and 488 (43%) were eters (in ModelTest) and then used in further ML parsimony informative. As expected, most vari- searches until identical topologies were gener- ation occurred at the third codon position. Of the ated in successive searches. 588 variable characters, 369 occurred at third In addition to parsimony and ML, Bayesian positions, 64 at second positions, and 155 at ®rst inference of phylogeny (as implemented in codon positions. Uncorrected sequence diver- MrBayes 2.01, Huelsenbeck and Ronquist 2001) gence (p-distance) between genus-level exem- was also used to evaluate relationships. The plars ranged from 12% (Chloropeta±Hippolais) Bayesian approach generates a pool of trees or to 20% (Anthus±Aphelocoma). Intrageneric var- a posterior probability distribution. The distri- iation ranged from 0.4% (Chamaea)to14% bution of sample points after which trees con- (Sylvia). verge on a likelihood score was used to produce a majority-rule consensus tree. Nodal support PHYLOGENETIC ANALYSIS was inferred using a clade's posterior probability In the ModelTest analyses, all three starting to- (the percentage of samples that contain that par- pologies (uniformly weighted parsimony tree, ticular clade). Multiple independent Bayesian six-parameter step matrix parsimony tree, and a analyses were conducted and compared to en- neighbor-joining tree) yielded TVM ϩ I ϩ⌫as sure analyses were not entrapped on local opti- the preferred model. Each of these three trees ma. ModelTest identi®ed TVM ϩ I ϩ⌫as the (and their parameters estimated from the preferred model for ML analysis, but MrBayes ModelTest analyses) were used as starting trees cannot implement this model. Instead the GTR in separate heuristic ML searches using a suc- ϩ I ϩ⌫(nst ϭ 6 rates ϭ invargamma) was cessive approach. The results of the three sepa- incorporated into the Bayesian analysis. GTR ϩ rate ML analyses were the same. Regardless of I ϩ⌫varies slightly from TVM ϩ I ϩ⌫, in that starting tree used, all converged on the same op- all nucleotide substitutions are unequal and in- timal ML tree (Fig. 1, Ϫln likelihood ϭ 17 dependent of each other in the GTR ϩ I ϩ⌫ 276.8; rAC ϭ 0.07, rAG ϭ 2.6, rAT ϭ 0.32, rCG ϭ model (transition substitutions are uniform with 0.20, rCT ϭ 2.6, rGT ϭ 1.00, %A ϭ 38, %C ϭ TVM ϩ I ϩ⌫). All Bayesian analyses were run 43, %G ϭ 5, %T ϭ 14, proportion invariable for 250 000 generations and sampled every 100 sites ϭ 0.37). In this tree, the Wrentit was the generations. Thus, each analysis resulted in 2500 sister taxon of a clade including three species of samples. Four Markov Chain Monte Carlo Sylvia warblers, and the Wrentit and Sylvia clade chains (being incrementally heated) were run for was sister to a clade containing six babblers (Yu- each analysis and parameters were not speci®ed hina, Stachyris, Garrulax, Leiothrix, Pomator- and thus were estimated as part of the analysis. hinus, and Illadopsis). The ML tree showed Support for phylogenetic relationships was as- broad agreement with Sibley and Monroe's sessed using nonparametric bootstrapping of MP (1990) proposed classi®cation of passerine birds. analyses and using Bayesian measures of nodal The parvorder Corvida was monophyletic as support (posterior probability values). Based on were the superfamilies Muscicapoidea and Pas- simulation studies and studies of known phylog- seroidea (but see Sheldon and Gill 1996, Groth enies, a bootstrap value of 70% or greater is gen- 1998, and Barker et al. 2001). With the excep- erally regarded as strongly supported given phy- tion of Regulus, the superfamily Sylvioidea was logenetic model assumptions (Hillis and Bull also monophyletic. Within Sylvioidea, the reed 1993). In Bayesian analyses, posterior probabil- warbler group (Acrocephalus, Chloropeta, and WRENTIT PHYLOGENETIC RELATIONSHIPS 745

FIGURE 1. Phylogenetic relationships of the Wrentit (Chamaea fasciata) as identi®ed by maximum likelihood. Branch lengths are proportional. Numbers above nodes indicate bootstrap support for the transition/transversion analysis. Nodes without values above the node were not supported in greater than 50% of bootstrap replicates. Numbers below nodes indicate Bayesian posterior probability values. Nodes without values below the node represent parts of the tree that con¯ict between the Bayesian and maximum-likelihood trees. For members of the parvorder Passerida, the letter following taxon name indicates the superfamily according to Sibley and Monroe's (1990) taxonomy (S ϭ Sylvioidea, M ϭ Muscicapoidea, P ϭ Passeroidea). Members of Sibley and Monroe's subfamily Sylviinae are indicated with an asterisk.

Hippolais) and three species of leaf warblers (2001). ML analysis of the mtDNA also showed (Phylloscopus) each formed a clade. The ML Muscicapoidea and Passeroidea as being sister analysis supports the placement of Regulus out- taxa. side of all Passerida superfamilies (Muscicapoi- Five separate Bayesian analyses were con- dea, Passeroidea and Sylvioidea) in agreement ducted and all ®ve converged onto similar, stable with Sheldon and Gill (1996) and Barker et al. log-likelihood scores well before 25 000 gener- 746 DINO N. BARHOUM AND KEVIN J. BURNS ations. Thus, we eliminated the ®rst 250 samples Importantly, regardless of weighting scheme for all analyses and constructed majority-rule used, all parsimony trees showed the Wrentit to consensus trees using the remaining 2250 sam- be the sister taxon of the clade containing three ples. All ®ve independent Bayesian analyses species of Sylvia. Bootstrap support for this re- generated identical trees. This tree was broadly lationship ranged from 54% (codon position congruent with the ML tree. Posterior probabil- weighting) to 87% (both six-parameter and tran- ities for the ®ve Bayesian analyses varied only sition/transversion weighting). The reed warbler slightly, and values of the ®rst analysis were (Acrocephalus, Hippolais, Chloropeta) and leaf mapped onto the ML tree (Fig. 1). All clades warbler (Phylloscopus) groups also were mono- with posterior probabilities of 95 or greater are phyletic in all the most-parsimonious trees. considered to be signi®cantly supported. Support However, in general, parsimony trees showed for the sister relationship of the Wrentit to a less agreement with previous hypotheses of re- clade including three species of Sylvia warblers lationships than did the ML and Bayesian trees. was high (100%) as was support for the sister In addition, fewer than half of the nodes in the relationship of these two taxa to a clade includ- most-parsimonious trees had greater than 70% ing six species of babblers (100%). In addition, bootstrap support, and most well-supported support was high for the monophyly of Passe- nodes joined pairs or trios of terminal taxa. The roidea and Muscicapoidea, as represented in this species representing the Corvida and Muscica- study. Two nodes differed between the ML tree poidea were not monophyletic in the equally and the Bayesian tree. The Bayesian tree dif- weighted tree because the genus Pomatostomus fered from the ML tree in the relationships fell within the Muscicapoidea. In all parsimony among the three Passerida superfamilies. The analyses, the Passeroidea was not monophyletic, Bayesian tree identi®ed the Sylvioidea as the sis- because Vidua always fell outside this otherwise ter taxon of the Passeroidea (posterior probabil- monophyletic group. The Sylvioidea also was ity of 48%), while the ML analyses identi®ed not monophyletic in the parsimony analyses, be- the Muscicapoidea as the sister taxon of the Pas- cause passeroid taxa were nested within this seroidea. In the ML tree, Thamnornis chloro- group. petoides and Neomixis ¯avoviridis are sister to As found in other studies (e.g., LeacheÂ and the rest of the Sylvioidea (Fig. 1), while in the Reeder 2002), posterior probability values of the Bayesian tree, these two taxa are the sister group Bayesian analyses were generally higher than to a subset of the Sylvioidea (posterior proba- the bootstrap parsimony values. In addition, pos- bility ϭ 47%). terior probabilities and bootstrap values were Like the Bayesian analyses, the parsimony only somewhat correlated. Using the bootstrap analyses agreed with the ML tree in the place- values of the transition/transversion tree (Fig. 1), ment of several taxa (bootstrap values of tran- 19 nodes had a posterior probability greater than sition/transversion weighted tree shown, Fig. 1). 95% and a bootstrap value greater than 70%. An Depending on the weighting scheme used, the additional 13 nodes had greater than 95% pos- parsimony analyses varied in degree of resolu- terior probability, but weaker than 70% boot- tion and number of bootstrap-supported nodes. strap support. Only four nodes showed strong Number of trees found, consistency indices ex- bootstrap support (greater than 70%), but a pos- cluding uninformative characters, and number of terior probability less than 95%. bootstrap nodes above 50% for each of the four analyses were as follows: equal weighting (6 DISCUSSION trees, 0.23, 22 nodes), transition/transversion COMPARISON TO PREVIOUS SYSTEMATIC weighting (2 trees, 0.20, 32 nodes), codon po- STUDIES sition weighting (1 tree, 0.25, 23 nodes), six- Previous phylogenetic studies supported mono- parameter weighting (3 trees, 0.21, 33 nodes). In phyly of the following groups: (1) Aphelocoma, general, the weighting schemes that attempted to Corvus, Lophorina, Pomatostomus, and Terpsip- account for saturation showed more resolution hone (Corvida; Sibley and Monroe 1990, but see and higher bootstrap values than the equally Barker et al. 2001); (2) Phylloscopus (Price et weighted tree. This is not surprising, given the al. 1997, 1998); (3) Acrocephalus, Chloropeta, distant relationships of many of the taxa includ- and Hippolais (Sibley and Monroe 1990, Helbig ed in this study. and Seibold 1999); (4) Anthus, Vidua, Melospi- WRENTIT PHYLOGENETIC RELATIONSHIPS 747 za, Motacilla, Passerella, and Piranga (Passe- raphyletic. More genera, including representa- roidea; Sibley and Monroe 1990); (5) Catharus, tives from all Sylvioidea families, are needed to Copsychus, Erithacus, Ficedula, Muscicapa, To- provide conclusive evidence for the paraphyly xostoma, and Turdus (Muscicapoidea; Sibley of these families as well as other relationships and Monroe 1990); and (6) Thamnornis, Regu- within Sylvioidea. lus, Neomixis, Bradypterus, Locustella, Ortho- tomus, Hypergerus, Cisticola, Cettia, Phyllos- WRENTIT RELATIONSHIPS AND copus, Scotocerca, Hippolais, Acrocephalus, BIOGEOGRAPHIC HISTORY Chloropeta, Chamaea, Sylvia, Yuhina, Stachy- Chamaea (Wrentits) and a clade of three species ris, Garrulax, Leiothrix, Pomatorhinus, and Illa- of Sylvia share a more recent common ancestor dopsis (Sylvioidea; Sibley and Monroe 1990). with each other than with any other taxa in the Our parsimony analysis generated slightly vary- data set. All optimal trees from ML, MP, and ing topologies and all deviated from these Bayesian analyses included this pairing, and groupings to some extent (all demonstrated Cor- Bayesian analyses provided signi®cant levels of vida, Muscicapoidea, Passeroidea, or Sylvioidea support (100% posterior probability) for this paraphyly). Both the ML tree and the Bayesian pairing. In addition, Wrentits, Sylvia, and bab- trees showed much broader agreement with the blers (Pomatorhinus, Garrulax, Leiothrix, clades described above. The ®rst ®ve clades Stachyris, Illadopsis, and Yuhina) formed a were found in the ML and Bayesian trees, and clade in all optimal trees. Thus the results of our Sylvioidea was monophyletic with the exception study, in part, corroborate the earlier hypotheses of Regulus (see below). More extensive sam- of Sibley and Ahlquist (1982, 1990). Sibley and pling is necessary (see Barker et al. 2001) to Ahlquist (1990) considered the Wrentit, the ge- fully address the monophyly of most of these nus Sylvia, and a large group of traditional bab- groups. blers to form a monophyletic group. However, In addition to the similarities with Sibley and Sibley and Ahlquist (1982, 1990) did not ex- Monroe (1990) outlined above, our study also plicitly recognize a close relationship between identi®ed differences. In agreement with Shel- Sylvia and the Wrentit to the exclusion of bab- don and Gill (1996) and Barker et al. (2001), the blers. UPGMA trees and ⌬T values indicated ML tree and the Bayesian tree placed Regulus that the Wrentit was actually closer to Sylvia outside of Sibley and Monroe's (1990) Sylvioi- than to Timaliini (Sibley and Ahlquist 1990). dea. ML phylogenetic hypotheses show that Ga- However, Sibley and Monroe's (1990) classi®- rrulax is closely related to babblers assigned to cation did not indicate the close relationship be- Timaliini with strong Bayesian support (100% tween Sylvia and the Wrentit in relation to bab- posterior probability). This conclusion is con- blers, probably because the branch connecting cordant with results from Cibois et al. (1999, Sylvia and the Wrentit was short and within the 2001), but contrasts with Sibley and Monroe's realm of error. Sibley and Monroe (1990) ranked (1990) classi®cation. Also in contrast to Sibley all three groups at an equivalent taxonomic lev- and Monroe (1990), in none of the phylogenetic el, and together they compose the three tribes analyses was Neomixis monophyletic nor did it within Sylviinae. Our results indicate that rec- form a close association to other Timaliini. Ci- ognizing a monophyletic group that consists of bois et al. (1999, 2001) also demonstrated Neo- Sylvia and the Wrentit to the exclusion of other mixis paraphyly with regards to intrageneric babblers would be appropriate, and that the babbler relationships. Inferred relationships be- branching diagrams illustrated in Sibley and tween and within families of Sylvioidea (except Ahlquist (1990) do re¯ect the relationships of for the distinction of the subfamily Sylviinae, these taxa. i.e., Wrentit, Sylvia, and Timaliini) were sup- Overall similarity based on morphology and ported weakly. For example, the families Cisti- ecology lends additional support to an associa- colidae (Cisticola, Hypergerus, and Scotocerca) tion between the Wrentit and Sylvia (as ®rst not- and Sylviidae (Thamnornis, Neomixis, Bradyp- ed by Sibley 1970). The Wrentit and some mem- terus, Locustella, Orthotomus, Cettia, Phyllos- bers of Sylvia are similar in plumage, iris col- copus, Hippolais, Acrocephalus, Chloropeta, oration, and overall body size and shape. Sylvia Chamaea, Sylvia, Yuhina, Stachyris, Garrulax, undata (Dartford Warbler) has dark brown to Leiothrix, Pomatorhinus, Illadopsis) were pa- brownish-gray upperparts with lighter streaked 748 DINO N. BARHOUM AND KEVIN J. BURNS underparts (similar to Wrentits in the northern- LITERATURE CITED most end of their distribution), while Sylvia sar- AVISE, J. C., J. ARNOLD,R.M.BALL,E.BERMINGHAM, da (Marmora's Warbler) has bluish-gray upper- T. L AMB,J.E.NEIGEL,C.A.REEB, AND N. C. parts with a lighter, mottled underbelly (similar SAUNDERSET. 1987. Intraspeci®c phylogeography: to Wrentits found in the southern end of their the mtDNA bridge between population genetics and systematics. Annual Review of Ecology and distribution). All three species have small round- Systematics 18:489±522. ed wings with long slim tails, which they hold BARKER, F. K., G. F. BARROWCLOUGH, AND J. G. GROTH. cocked. Juveniles all have lighter, creamy or- 2001. A phylogenetic hypothesis for passerine ange or brown, speckled irises (only outer iris birds: taxonomic and biogeographic implications for the Wrentit) that become brighter with age. of an analysis of nuclear DNA sequence data. Pro- ceedings of the Royal Society of London Series All Sylvia species are primarily insectivorous B 269:295±308. and partially frugivorous, build open cup-shaped CIBOIS, A., B. SLIKAS,T.S.SCHULENBERG, AND E. PAS- nests in dense vegetation, and reach their highest QUET. 2001. An endemic radiation of Malagasy density in the Mediterranean region. These char- songbirds is revealed by mitochondrial DNA se- acteristics are similar for the Wrentit except it quence data. Evolution 55:1198±1206. CIBOIS, A., E. PASQUET, AND T. S. SCHULENBERG. 1999. reaches highest species density in California Molecular systematics of the Malagasy babblers (with a similar, Mediterranean climate). Al- (Timaliidae) and warblers (Sylviidae), based on though these characteristics are morphological cytochrome b and 16S rRNA sequences. Molec- and ecological generalities, they do agree with ular Phylogenetics and Evolution 13:581±595. EDWARDS, S. V., P. ARCTANDER, AND A. C. WILSON. the hypotheses generated in this study. 1991. Mitochondrial resolution of a deep branch The phylogeny presented in this study pro- in the genealogical tree for perching birds. Pro- vides a framework for addressing the biogeo- ceedings of the Royal Society of London Series graphic history of the Wrentit. Current distri- B 243:374±385. ERICSON, P. G. P., U. S. JOHANSSON, AND T. J. PARSONS. butions are allopatric (Wrentit: western North 2000. Major divisions in the oscines revealed by America, Sylvia: primarily Europe, the Middle insertions in the nuclear gene c-myc: a novel gene East, and Africa). Both groups are more closely in avian phylogenetics. Auk 117:1069±1078. related to babblers of Timaliini than to any other GAMBEL, W. 1845. Descriptions of new and little Sylvioidea member. Babblers are widespread, known birds, collected in upper California. Pro- ceedings of the Academy of Natural Sciences, but reach their greatest diversity and density in Philadelphia 2:263±266. Asia and have been proposed to have Asian an- GRINNELL, J. 1913. Call-notes and mannerisms of the cestry. One possible biogeographic scenario is Wren-tit. Condor 15:178±181. that the Wrentit and Sylvia ancestor also origi- GROTH, J. G. 1998. Molecular phylogenetics of ®nches and sparrows: consequences of character state re- nated in Asia. As each group dispersed, they col- moval in cytochrome b sequences. Molecular onized new yet similar climates and habitats. Phylogenetics and Evolution 10:377±390. Wrentit ancestors dispersed east to western HARSHMAN, J. 1994. Reweaving the tapestry: what can North American and the chaparral habitat of we learn from Sibley and Ahlquist (1990)? Auk California, and Sylvia ancestors dispersed west 111:377±388. HELBIG, A. J., AND I. SEIBOLD. 1999. Molecular phy- to Europe and the maquis habitat of the Medi- logeny of the Palearctic-African Acrocephalus and terranean. Hippolais warblers (Aves: Sylviidae). Molecular Phylogenetics and Evolution 11:246±260. ACKNOWLEDGMENTS HILLIS, D. M., AND J. J. BULL. 1993. An empirical test of bootstrapping as a method for assessing con®- Funding for this project was provided by the Frank M. dence in phylogenetic analysis. Systematic Biol- Chapman fund of the American Museum of Natural ogy 42:182±192. History and the San Diego State University Founda- HOWARD, R., AND A. MOORE. 1991. A complete check- tion. The California Department of Fish and Game, the list of the birds of the world. Academic Press, Oregon Department of Fish and Game, and the U.S. London. Fish and Wildlife Service provided collecting permits. HUELSENBECK, J. P. 1995. Performance of phylogenetic We thank the Museum of Vertebrate Zoology at Uni- methods in simulations. Systematic Biology 44: versity of California at Berkeley (N. Johnson and C. 17±48. Cicero) and the Field Museum of Natural History (S. HUELSENBECK,J.P.,AND F. R . R ONQUIST. 2001. Hackett) for contributing additional tissue samples. We MRBAYES: Bayesian inference of phylogenetic also thank M. Martinez-Vergara for providing the trees. Bioinformatics 17:754±755. Spanish translation of the abstract. For comments on LEACHEÂ , A. D., AND T. W. R EEDER. 2002. Molecular the manuscript, we thank T. Reeder, B. Yang, D. Dob- systematics of the eastern fence lizard (Sceloporus kin, and two anonymous reviewers. undulatus): a comparison of parsimony, likeli- WRENTIT PHYLOGENETIC RELATIONSHIPS 749

hood, and Bayesian approaches. Systematic Biol- SIBLEY, C. G., AND J. E. AHLQUIST. 1982. The relation- ogy 51:44±68. ships of the Wrentit as indicated by DNA-DNA LOVETTE, I. J., AND E. BERMINGHAM. 2000. C-mos var- hybridization. Condor 84:40±44. iation in songbirds: molecular evolution, phylo- SIBLEY, C. G., AND J. E. AHLQUIST. 1990. Phylogeny genetic implications, and comparisons with mito- and classi®cation of birds. Yale University Press, chondrial differentiation. Molecular Biology and New Haven, CT. Evolution 17:1569±1577. SIBLEY, C. G., AND B. L. MONROE JR. 1990. Distribu- MADDISON,W.P.,AND D. R. MADDISON. 1997. Mac- tion and taxonomy of birds of the world. Yale Clade: analysis of phylogeny and character evo- University Press, New Haven, CT. lution. Ver 3.07. Sinauer, Sunderland, MA. SWOFFORD, D. L. 2001. PAUP*. Phylogenetic analysis MOORE, W. S., AND V. R. DEFILIPPIS. 1997. The win- using parsimony (*and other methods). Version dow of taxonomic resolution for phylogenies 4.0b8. Sinauer Associates, Sunderland, MA. based on mitochondrial cytochrome b, p. 84±120. WALSH, P. S., D. A. METZGER, AND R. HIGUCHI. 1991. In D. P. Mindell [ED.], Avian molecular evolution Chelex 100 as a medium for simple extraction of and systematics. Academic Press, San Diego, CA. DNA for PCR-based typing from forensic mate- MOORE, W. S. 1995. Inferring phylogenies from rial. Biotechniques 10:506±513. mtDNA variation: mitochondrial-gene trees versus WETMORE, A. 1960. A classi®cation for the birds of nuclear-gene trees. Evolution 49:718±726. the world. Smithsonian Miscellaneous Collections 139:1±37. POSADA, D., AND K. A. CRANDALL 1998. MODEL- WILGENBUSCH, J., AND K. DE QUEIROZ. 2000. Phylo- TEST: testing the model of DNA substitution. genetic relationships among the phrynosomatid Bioinformatics 14:817±818. sand lizards inferred from mitochondrial DNA se- PRICE, T., H. L. GIBBS,L.DE SOUSA, AND A. D. RICH- quences generated by heterogeneous evolutionary MAN. 1998. Different timing of the adaptive radi- processes. Systematic Biology 99:592±612. ations of North American and Asian warblers. YANG, Z. 1996. Phylogenetic analysis using parsimony Proceedings of the Royal Society of London Se- and likelihood methods. Journal of Molecular ries B 265:1969±1975. Evolution 42:294±307. PRICE, T. D., A. J. HELBIG, AND A. D. RICHMAN. 1997. Evolution of breeding distributions in the Old APPENDIX World leaf warblers (genus Phylloscopus). Evo- GenBank accession numbers for sequences that come lution 51:552±561. from other studies. GenBank numbers for sequences RAMBAUT, A. 1995. Se-Al v1.0a1: sequence alignment unique to this study are reported in Table 1. program. Department of Zoology, University of AF094618, AF094622, AF094625, AF094627, Oxford, Oxford, UK. AF094628, AF094630, AF094631, AF094636, SHELDON, F. H., AND F. B. GILL. 1996. A reconsidera- AF094637, AF094640, AF094642, AF135052, tion of songbird phylogeny, with emphasis on the AF139826, AF197835, AF197837, AF276776, evolution of titmice and their sylvioid relatives. AF290159, AJ004318, AJ004762, AJ004764, AJ004765, Systematic Biology 45:473±495. AJ004789, AJ004797, AJ004798, AJ299689, AJ299690, SIBLEY, C. G. 1970. A comparative study of the egg- NC࿞000880, U25733, U46773, U46777, U77335, white proteins of passerine birds. Bulletin of the U86031, X60937, X60938, X74261, Y08058, Y10732, Peabody Museum of Natural History 32:1±131. Y10739, Z73474, Z73487, Z73494.