A Molecular Phylogeny of Anseriformes Based on Mitochondrial DNA Analysis

Home , Aix (bird), Alopochen, Anas, Anatidae, Anatinae, Anserinae

MOLECULAR PHYLOGENETICS AND EVOLUTION Molecular Phylogenetics and Evolution 23 (2002) 339–356 www.academicpress.com

A molecular phylogeny of anseriformes based on mitochondrial DNA analysis

Carole Donne-Goussee,a Vincent Laudet,b and Catherine Haanni€ a,*

a CNRS UMR 5534, Centre de Genetique Moleculaire et Cellulaire, Universite Claude Bernard Lyon 1, 16 rue Raphael Dubois, Ba^t. Mendel, 69622 Villeurbanne Cedex, France b CNRS UMR 5665, Laboratoire de Biologie Moleculaire et Cellulaire, Ecole Normale Superieure de Lyon, 45 Allee d’Italie, 69364 Lyon Cedex 07, France

Received 5 June 2001; received in revised form 4 December 2001

Abstract

To study the phylogenetic relationships among Anseriformes, sequences for the complete mitochondrial control region (CR) were determined from 45 waterfowl representing 24 genera, i.e., half of the existing genera. To confirm the results based on CR analysis we also analyzed representative species based on two mitochondrial protein-coding genes, cytochrome b (cytb) and NADH dehydrogenase subunit 2 (ND2). These data allowed us to construct a robust phylogeny of the Anseriformes and to compare it with existing phylogenies based on morphological or molecular data. Chauna and Dendrocygna were identified as early offshoots of the Anseriformes. All the remaining taxa fell into two clades that correspond to the two subfamilies Anatinae and Anserinae. Within Anserinae Branta and Anser cluster together, whereas Coscoroba, Cygnus, and Cereopsis form a relatively weak clade with Cygnus diverging first. Five clades are clearly recognizable among Anatinae: (i) the Anatini with Anas and Lophonetta; (ii) the Aythyini with Aythya and Netta; (iii) the Cairinini with Cairina and Aix; (iv) the Mergini with Mergus, Bucephala, Melanitta, Callonetta, So- materia, and Clangula, and (v) the Tadornini with Tadorna, Chloephaga, and Alopochen. The Tadornini diverged early on from the Anatinae; then the Mergini and a large group that comprises the Anatini, Aythyini, Cairinini, and two isolated genera, Chenonetta and Marmaronetta, diverged. The phylogeny obtained with the control region appears more robust than the one obtained with mitochondrial protein-coding genes such as ND2 and cytb. This suggests that the CR is a powerful tool for bird phylogeny, not only at a small scale (i.e., relationships between species) but also at the family level. Whereas morphological analysis effectively resolved the split between Anatinae and Anserinae and the existence of some of the clades, the precise composition of the clades are different when morphological and molecular data are compared. Ó 2002 Elsevier Science (USA). All rights reserved.

Keywords: Anseriformes; mtDNA; Control region; Waterfowl

1. Introduction North America (Olson and Feduccia, 1980). According to these paleontological data, the main radiation of Among avian orders, the Anseriformes (screamers modern ducks has taken place during Miocene, 5–23 and waterfowls) are a morphologically and biologically million years ago (Olson, 1985). diverse group containing ca. 150 species distributed The Anseriformes are traditionally divided into two worldwide. This order contains the screamers of South families, Anhimidae (2 genera and 3 species) and America, the magpie goose of Australia and Asia, and Anatidae (approximately 41 genera and 147 species). the ducks, geese, and swans known worldwide. Fossil The taxonomic division is rather complex and has been records indicate that the ﬁrst Anseriformes (genus much disputed and revised. Most available data con- Presbyornis) was present during Upper Paleocene (61– cerning Anseriformes phylogeny came from morpho- 62 million years ago), whereas the ﬁrst Anatidae was logical, anatomical, and behavioral analyses (Delacour found in the Upper Eocene (40–50 million years ago) in and Mayr, 1945; Del Hoyo et al., 1992; Livezey, 1986, 1997b). Molecular data such as DNA–DNA hybridiza- * Corresponding author. Fax: +33-4-72-44-05-55. tion studies were also used to decipher the relationships E-mail address: [email protected] (C. Ha¨nni). between these birds (Sibley and Ahlquist, 1990). More

1055-7903/02/$ - see front matter Ó 2002 Elsevier Science (USA). All rights reserved. PII: S1055-7903(02)00019-2 340 C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356 recently, phylogenies based on DNA sequence analysis nae (Fig. 1B). A radically different view has been pro- of mitochondrial genes were proposed for some taxa of posed by Sibley and Ahlquist (1990), who cluster together the Anseriformes such as the Australasian waterfowl Chauna and Anseranas to form the suborder Anhimides. (Sraml et al., 1996), the genus Anas (Johnson and As Livezey (1997b), these authors consider Dendrocygna Sorenson, 1998, 1999), some extinct species such as an independent lineage. Because of these discrepancies moa-nalos (Sorenson et al., 1999a), or the endangered the composition of the family Anatidae is still a matter of Laysan ducks (Cooper et al., 1996). The phylogenies discussion, as for example the inclusion of Anseranas. obtained by these different approaches differ mostly on The Anatidae have been traditionally divided into two (i) the relative positions of Anhimidae (with Ansera- subfamilies, Anatinae and Anserinae, the latter including natidae) and Anatidae, (ii) the grouping of Anatidae in geese, swans, and Dendrocygna (Delacour and Mayr, two main subfamilies, Anatinae and Anserinae, and (iii) 1945; Del Hoyo et al., 1992; Fig. 1A and Appendix A). the phylogeny inside each of these subfamilies, in par- This view as been challenged by Livezey (1997b), who ticular the composition and relationships of so-called recognized four main clades (Fig. 1B): (i) the Anatinae; ‘‘tribes’’ clustering several genera within Anatidae. (ii) a clade called Tadornini including the genera Tad- The traditional classification of the Anseriformes of orna, Chloephaga, and Alopochen; (iii) the Anserinae; and Delacour and Mayr (1945), based on morphological and (iv) Dendrocygna. In addition other minor clades (Tha- behavioral characters, later modified by Johnsgard lassorninae, Stictonettinae, and Plectropterinae) were (1978), has been followed by many others (Del Hoyo et al., also noticed inside Anatidae. The relationships between 1992). According to these authors, the genus Chauna, all these clades remain poorly resolved. According to this representative of the Anhimidae, diverged first, and was author, the Tadornini, which were previously included then followed by Anseranas and the crown group con- inside the Anatinae by Del Hoyo et al. (1992), represent taining ducks, geese, and swans (Fig. 1A and Appendix an independent lineage. The definition of Anserinae by A). According to these authors the whistling-ducks Del Hoyo et al. (1992) and Livezey (1997b) differs only (Dendrocygna) are placed inside the subfamily Anserinae. by the inclusion, or not, of Dendrocygna inside this clade. This scheme has been mainly confirmed by Livezey The DNA–DNA hybridization results of Sibley and (1997b), who nevertheless proposed that Dendrocygna Ahlquist (1990) give rise to an even different scheme with represents an independent lineage, unrelated to Anseri- three, main lineages (Fig. 1C and Appendix A): (i) the

Fig. 1. Schematic phylogenies of the Anseriformes according to (A) Del Hoyo et al. (1992); (B) Livezey (1997b); and (C) Sibley and Ahlquist (1990). Only the 24 genera analyzed in our study plus Anseranas are depicted in these trees. The subfamilies Anatinae and Anserinae (or the tribes Anatini and Anserini) deﬁned by the various authors are indicated. C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356 341

Anatini, which has a composition different from that of verging first and then a split between Anserinae (Anser, the Anatidae of Del Hoyo et al. (1992) and Livezey Branta, and Cygnus) and Anatinae (Sorenson et al., (1997b); (ii) the Anserini, including Tadornini and Cai- 1999a). Within Anatinae four main clades were found: (i) rina; and (iii) the genera Cygnus and Coscoroba which Anas and related genera; (ii) a clade containing Tador- together form an independent grouping called Cygninae. nini (Tadorna and Chloephaga) allied with Cairina and Given the strong level of discrepancy existing between Aix (this group is then clustered with Mergini (Bucep- the three main published phylogenies at the family and hala)); (iii) Aythyini (Aythya, Netta and Marmaronetta) subfamily levels, it is not surprising that the relation- with other genera such as Cyanochen and Pteronetta; and ships inside each subfamily are also a matter of con- (iv) Chenonetta and Sarkidiornis. Despite their interest in tention (Fig. 1). As shown in Fig. 1A, Del Hoyo et al. clearly indicating that a robust phylogenetic signal exists (1992) found five main tribes in the Anatinae. Some of in mtDNA to resolve waterfowl phylogeny, most of these these tribes such as Mergini (Mergus, Bucephala, Mel- molecular analyses were based on a limited sampling of anitta, Somateria, and Clangula) were also found by species or only marginally discuss the relationships be- Livezey (1997b). The tribe called Aythyini has different tween the various groups of Anseriformes. For these compositions for the two authors since Livezey (1997b) reasons we sequenced and analyzed the mtDNA control includes Marmaronetta in this clade, whereas Del Hoyo region from 45 Anseriformes representing 24 genera. Our et al. (1992) consider it as more closely related to Anas. sampling includes representatives of all of the main tribes Other tribes such as those containing the genera Anas, of Anseriformes and half of the existing genera, allowing Aix,orCairina were completely different for the two us to propose a phylogeny of this order. In addition, we authors. Concerning the Anserinae, Del Hoyo et al. sequence the cytb and ND2 genes in a more limited set of (1992) cluster together Anser, Branta, Cygnus, and Co- species to constitute, in addition to the sequencing done scoroba with the exclusion of Cereopsis and Dendrocy- by Johnson and Sorenson (1998), a data set containing gna which form two independent lineages (Fig. 1A). the mitochondrial control region (CR), cytb, and ND2 Livezey (1997b) clusters Anser and Branta on one hand for 18 species of the crown group Anatinae. This allows and Cygnus and Coscoroba on the other hand. He also us to compare the resolution and robustness of phylog- considers Cereopsis as an independent lineage (Fig. 1B). enies based on CR or protein-coding genes. This clearly Using DNA–DNA hybridization, Sibley and Ahlquist shows that the CR appears to be an efficient tool with (1990) cannot resolve the phylogeny at the generic level. which to decipher the phylogeny of Anseriformes, not This complex situation calls for the completion of a only at the species and genus levels but also at the family molecular phylogeny of Anseriformes using DNA se- level. This analysis allowed us to test the various sce- quence analysis. Several recent reports have proposed narios proposed by other authors based on morpholog- partial schemes that were mainly focused on Anatinae or ical or DNA–DNA hybridization analysis. even on the genus Anas. Using complete cytochrome b (cytb) and NADH dehydrogenase subunit 2 (ND2) genes, Johnson and Sorenson (1998, 1999) found that 2. Materials and methods Anas is not a monophyletic genus since it also contains species of the genera Lophonetta, Amazonetta, Specul- 2.1. Taxa examined anas,andTachyeres. Despite the use of extensive out- groups containing sequences of 11 genera of Anatidae, Investigation of 45 Anseriformes species from 24 gen- no information with regard to the grouping of these taxa era was done. Cytochrome b and ND2 of 14 sequences and was given. The analysis of a short segment of the cytb the complete mitochondrial DNA of the redhead (Aythya gene of Australasian waterfowl suggests some grouping americana) were obtained from GenBank whereas 4 cytb inside the Anatidae despite the fact that, because of the and 4 ND2 sequences were determined (see Table 1; small size of the sequenced fragment, the overall ro- Johnson and Sorenson, 1998; Sorenson et al., 1998). bustness level of this analysis is weak (Sraml et al., 1996). The complete control region of the snow goose (Anser These authors found that Anseranas and Dendrocygna caerulescens) was published by Quinn and Wilson (1993). were two independent lineages diverging early from other Anatidae. The Anatidae are then split into two main 2.2. DNA extraction lineages that correspond to Anatinae and Anserinae. Given the weak resolution of this analysis, within these Feather samples were collected from live birds in the two subfamilies only two groups, namely Cairina and wild and in captivity in France at the ‘‘Parc de Cleres’’ of Aix in Anatinae and Cygnus and Branta in Anserinae, are MNHN, Zoological Museum of Lille, and the ‘‘Parc robustly found (Sraml et al., 1996). A more recent anal- Ornithologique Ker Anas’’ (Table 1). DNA was extracted ysis of a longer set of mitochondrial sequences containing from feathers using the method described by Taberlet and clustered fragments of 12S, cytb, and tRNAs gave rise to Bouvet (1991). Feathers were digested in a total volume of the same scheme with Anseranas and Dendrocygna di- 400 ll of buffer (10 mM Tris–HCL, pH 8.0, 2 mM EDTA, 342 C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356

Table 1 Scientific name, common name, region of mtDNA sequenced, origin and accession number of the relevant sequences for the various samples studied Species Common name Region of mtDNA sequenced Origin of sample Accession number Aix galericulata Mandarin duck D loop Lille Museum AY112953 Alopochen aegyptiacus Egyptian goose D loop Ker Anas Park AY112964 Anas acuta Pintail D loop Cytba ND2a Ker Anas Park AY112939, AF059055, AF059116 Anas bahamensis Bahama pintail D loop Cytba ND2a Lille Museum AY112940, AF059058, AF059119 bahamensis Anas clypeata Nothern shoveler D loop Cytba ND2a Ker Anas Park AY112941, AF059062, AF059174 Anas crecca Eurasian greenwinged teal D loop Cytba ND2a Ker Anas Park AY112942, AF059064, AF059124 Anas platyrhynchos Mallard D loop Cytba ND2a Lille Museum AY112938, AF059081, AF059141 Anas sibilatrix Chiloe wigeon D loop Cytba ND2a MNHN AY112943, AF059108, AF059168 Anas strepera Gadwall D loop Cytba ND2a Ker Anas Park AY112944, AF059109, AF059169 Anser albifrons White-fronted goose D loop MNHN AY112967 Anser anser Greylag goose D loop MNHN AY112966 Anser caerulescens Snow goose D loopc ACMTTPGF Anser erythropus Lesser white-fronted goose D loop MNHN AY112970 Anser indicus Bar-headed goose D loop Lille Museum AY112971 Anser rosii Ross goose D loop Ker Anas Park AY112972 Anser canagicus Emperor goose D loop MNHN AY112969 Aythya americana Redhead D loopb Cytbb ND2b NC000877 Aythya marila Greater saup D loop Lille Museum AY112947 Aythya nycora Ferruginous duck D loop Ker Anas Park AY112948 Branta bernicla Brent goose D loop Lille Museum AY112973 Branta canadensis Canada goose D loop Lille Museum AY112974 Branta leucopsis Barnacle goose D loop Ker Anas Park AY112975 Branta ruficollis Red-breasted goose D loop MNHN AY112976 Bucephala clangula Goldeneye D loop Cytb ND2 MNHN AY112959, AF515261, AF515265 Cairina moschata Muscovy duck D loop Cytba ND2a Toulouse (INRA) AY112952, AF059098, AF059158 Callonetta leucophrys Ringed teal D loop Cytba ND2a Lille Museum AY112960, AF059157, AF059097 Cereopsis Cape barren goose D loop MNHN AY112977 novaehollandiae Chauna torquata Crested screamer D loop MNHN AY112982 Chenonetta jubata Maned goose D loop Cytba ND2a Lille Museum AY112951, AF059100, AF059160 Chloephaga picta Magellan goose D loop Cytb ND2 Lille Museum AY112965, AF515262, AF515266 Clangula hyemalis Long-tailed duck D loop MNHN AY112954 Coscoroba coscoroba Coscoroba swan D loop MNHN AY112979 Cygnus atratus Black swan D loop Lille Museum AY112978 Dendrocygna bicolor Fulvous whistling duck D loop MNHN AY112980 Dendrocygna eytoni Eyton whistling duck D loop Lille Museum AY112981 Lophonetta Crested duck D loop Cytba ND2a MNHN AY112945, AF059102, AF059162 specularoides Marmaronetta Marbled teal D loop Cytba ND2a Ker Anas Park AY112950, AF059104, AF059164 angustirostris Melanitta nigra Black scoter D loop Cytb ND2 MNHN AY112961, AF515263, AF515267 Mergus albellus Smew D loop Ker Anas Park AY112957 Mergus cucullatus Hooded merganser D loop MNHN AY112958 Mergus serrator Red-breasted merganser D loop Lille Museum AY112956 Netta rufina Red-crested pochard D loop Ker Anas Park AY112949 Somateria mollissima Eider D loop Cytb ND2 MNHN AY112955, AF515264, AF515268 Tadorna tadorna Common shelduck D loop Cytba ND2a Ker Anas Park AY112962, AF059113, AF059173 Tadorna tadornoides Australian shelduck D loop Lille Museum AY112963 a Johnson and Sorenson (1998). b Sorenson et al. (1998). c Quinn and Wilson (1993). pH 8.0, 10 mM NaCl, 1% sodium dodecyl sulfate, and 2.3. Gene amplification and sequencing 0.4 mg/ml proteinase K) by incubation with constant agitation at 42 °C for 1–3 h. Samples were extracted twice The complete mitochondrial control region of each to standard phenol/chloroform extraction and isopro- species was amplified with the primers listed in Table 2. panol precipitation and dissolved in 100 ml of distilled To complete the range of species available, the mito- water (Haanni€ et al., 1995). For some samples, genomic chondrial cytb (1047 bp) and ND2 (1041 bp) genes were DNA was also isolated with a QIAamp DNeasy Kit amplified for four species, goldeneye (Bucephala clan- (QIAGEN) according to the manufacturer’s protocol. gula), Magellan goose (Chloephaga picta), black scoter C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356 343

Table 2 Primers used for ampliﬁcation and sequencing of D loop, Cytb, and ND2 Namea Sequenceb Refc D loop L16722 50-ACTACCCGAGACCTACGGCT-30 H1254 50-TCTTGGCAGCTTCAGTGCCA-30 L128 50-CATGCACGGACTAAACCCAT-30 L481 50-CCCCCTAAACCCCTCGCCCT-30 L718 50-TAAGCCTGGACACACCTGCG-30 H738 50-CGCAGGTGTGTCCAGGCTTA-30 H501 50-AGGGCGAGGGGTTTAGGGGG-30 H319 50-TGAATGCTCTAATACCCAAC-30 Cytb L14990 50-AACATCTCCGCATGATGAAA-30 1 H16064 50-CTTCGATTTTTGGTTTACAAGACC-30 1 L15191 50-ATCTGCATCTACCTACACATCGG-30 1 L15517 50-CACGAATCAGGCTCAAACAACC-30 1 L15710 50-CCMMCMCAYATCAARCCMGAATG-30 2 H15742 50-TGCTAGTACGCCTCCTAGTTTGTTTGGGATTGA-30 1 H15545 50-GTATGGGTGAAATGGAATTT-30 1 H15298 50-CCCTCAGAATGATATTTGTCCTCA-30 1 ND2 L5219 50-CCCATACCCCGAAAATGATG-30 1 H6313 50-CTCTTATTTAAGGCTTTGAAGGC-30 1 L5524 50-AGGCCTGGTCCCATTTCACT-30 L5758 50-GGCTGAATRGGMCTNAAYCARAC-30 1 L6022 50-CCAAAGTGACTCATCATCCA-30 H6031 50-CACTTTGGTATAAACCCTGT- 30 H5766 50-GGATGAGAAGGCTAGGATTTTKCG-30 1 H5544 50-AGTGAAATGGGACCAGGCCT-30 a L and H numbers designate the location of the 30 base in the light or heavy strand, respectively, of the published chicken mtDNA sequence (Desjardins and Morais, 1990). b Degenerate primer positions are as follows: M ¼ AorC;Y¼ CorT;R¼ AorG;N¼ A, C, T, or G; K ¼ GorT. c 1, Johnson and Sorenson (1998); 2, Sorenson et al. (1999b).

(Melanitta nigra), and eider (Somateria mollissima), us- ble, complete cytb, ND2, or control region sequences ing the primers listed in Table 2. PCR amplification was were systematically compared with partial sequences made in 50 ll total volume with 1 unit AmpliTaq DNA determined by other authors, available in GenBank. Polymerase (Sigma), the manufacturer’s buffer, 2 mM The presence of nuclear insertions of mtDNA MgCl2, 0.25 mM each dNTP, 200 lg=ml bovine serum sequences (called Numts) in our amplified sequences, abumin, and 1 lM each primer. The PCR cycle was as which can lead to the wrong phylogeny (Quinn, follows: denaturation at 94 °C for 1 min, annealing at 1997; Sorenson and Quinn, 1998; Zhang and He- 55 °C for 1.5 min, and extension at 72 °C for 2 min for 40 witt, 1996), has been tested using the following cri- cycles. Five-microliter aliquots of the amplification teria. (i) All the sequences were sequenced directly products were electrophoresed in 2% agarose gels and from the PCR product, cloned, and sequenced. visualized via ethidium bromide staining. PCR products All the clones exhibit the same sequences which are were purified with QIAquick PCR Purification Kit. identical to the direct sequence of the PCR product. PCR products were cloned with TOPO cloning (Invi- This suggest that only one fragment was amplified trogen). Double-stranded PCR product was used in from each sample. (ii) DNA was extracted from cycle sequencing reactions using fluorescent dye termi- feather and not blood, a tissue known to be prone to nators and AmpliTaq (Applied Biosystems). Reaction amplification of Numts since it is poor in mtDNA products were run on an ABI 373 automated DNA (Arctander, 1995; Quinn, 1992; Sorenson and sequencer. Fleischer, 1996). (iii) The cytb and ND2 fragments that we amplified are coding proteins of regular 2.4. Authenticity of the sequences size, indicating that no mutations disrupting the reading frame took place. (iv) The control region To avoid contamination between sample extrac- sequences are relatively rich in transitions, a situ tions, PCR amplifications and sequence analysis were ation that is reminiscent of mtDNA, but not of performed in different dedicated rooms. When possi- Numts. 344 C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356

2.5. Phylogenetic analysis blocks; CSBs). Following this division, on our alignment, domain I runs from the 50 end of the CR light Sequences were aligned by eye using SEAVIEW strand to position 470, the central domain runs from (Galtier et al., 1996) and CLUSTAL_W (Thompson position 471 to 1050, and domain IIs runs from position et al., 1994). All positions containing gaps were excluded 1051 to the 30 end of the CR (Fig. 2). The conserved from the analysis using either a pairwise or a global structural features are clearly evident on our alignment removal scheme (Hillis et al., 1996). (1424 bp). Within the control region, four CSBs were To estimate saturation, scatter plots that compared clearly identified (Fig. 2): F-, D-, and C-boxes are lo- pairwise percentage sequence divergence to pairwise cated in the central domain, and CSB-1 is located in transversion (TV) and pairwise transition (TS) diver- domain II (Baker and Marshall, 1997). gences were drawn. Saturation plots using Kimura’s The complete sequences of the control regions of (1980) two-parameter genetic distances were drawn. Anseriformes range in size from 970 bp in the maned According to Hackett (1996) saturation is determined to goose (Chenonetta jubata) to 1230 bp in the crested have occurred if the scatter of points shows a clear screamer (Chauna torquata), with an average size of leveling off of changes as sequence divergence increases. 1100 bp. Within Anatidae, there is a deletion of ca. 100– The aligned sequences were treated by distance 130 bp in Anatinae (Aix, Anas, Alopochen, Aythya, Bu- (neighbor-joining (NJ); Saitou and Nei, 1987) and cephala, Cairina, Callonetta, Chenonetta, Chloephaga, maximum-parsimony (MP) analyses as implemented in Clangula, Lophonetta, Marmaronetta, Melanitta, Mer- PHYLO_WIN (Galtier et al., 1996) and PAUP (version gus, Netta, Somateria, Tadorna) compared to other 3.1) (Swofford, 1993). We employed the random taxon groups of Anserinae (Anser, Branta, Cereopsis, Cosco- addition option to prevent the tree from getting stuck in roba, Cygnus, and the divergent genera Chauna and a local optimum during the heuristic search. For dis- Dendrocygna). Other small (1–20 bp) deletions in do- tance analysis we employed log-determinant (LogDet) main I are also present in Anatinae versus other species distances that allow correction for heterogeneous base (Fig. 2). Quinn and Wilson (1993) also reported rela- composition (Steel, 1993). The parsimony analysis was tively large deletions in both 50 (61 bp) and 30 (38 bp) assessed using the heuristic search method (character regions of the lesser snow goose compared to the do- optimization ACCTRAN, MULPARS, and TBR mestic chicken. This has been confirmed also by Ra- branch-swapping options) with gaps treated as a fifth mirez et al. (1993), who reported large deletions in both base and all uninformative positions excluded. We em- adjacent regions in the Peking duck versus the domestic ployed the random taxon addition option to prevent the chicken. The average sequence divergences between se- tree from getting stuck in a local optimum during the quences in different genera used in this study in domain heuristic search. Parsimony analyses weighted all char- I, central region, and domain II are 25%, 9%, and 22%, acters equally. ‘‘Classical’’ maximum-likelihood (ML) respectively. analysis, as implemented in DNAML of PHYLIP for The complete alignment of the control region se- example, excludes all positions with gaps and is thus quences of the 45 Anseriformes used in this study gave very lowly resolved, as are the distance and parsimony rise to an alignment containing 1424 sites, among which analyses excluding these positions. Maximum-likelihood are 1143 variable sites, 836 sites containing at least one estimation was thus performed using quartet puzzling gap, and 937 sites that are informative for parsimony (Strimmer and Von Haeseler, 1996) as implemented in when all events, transitions, transversions, and gaps are TREE-PUZZLE which allows pairwise gap removal. considered. When all positions with gaps are deleted The robustness of inferences was assessed through from the analysis, 201 parsimony-informative characters bootstrap resampling (BP) (Felsenstein, 1985) with the remain. The mean frequency of nucleotides in the distance (1000 replicates) and parsimony (100 replicates) compared sequences show a paucity of G compared to with one random addition per replicate. the other bases (28% A, 31.2% C, 15.4% G, and 25.5% T) this compostion does not vary among the 45 sequenced species but it varies among regions of the D 3. Results loop. Domain I is rich in A and C, the central domain is rich in C and T, and domain II is AT rich and very low 3.1. Control region sequence analysis in G as observed for other birds (Baker and Marshall, 1997). The average TS/TV ratio is 1.1. Characters were The Anseriformes control region has many avian thus equally weighted for the parsimony analysis. features that have been reported in other birds (Des- Two data sets were used in the phylogenetic analysis: jardins and Morais, 1990). Typically, the vertebrate CR (i) a data set containing the complete control region is subdivided into three domains (domain I, central sequence of the 45 sequenced species and (ii) a reduced domain, and domain II), characterized primarily by data set containing only 1 sequence for each genus, i.e., different structural features (e.g., conserved sequence only 24 sequences. This last data set contains 1424 sites C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356 345

Fig. 2. Structure of the mitochondrial DNA replication control region in three representative species of Anseriformes used in our study: an anatine, the mallard Anas platyrhynchos; an anserine, the common goose Anser anser; and an animid, the screamer Chauna torquata. The tRNA Glu and Phe that surround the control region are indicated. The three domains discussed in the text are differentially shaded with the central conserved domain depicted with a darker shading. The various conserved sequence blocks, F-, D-, and C-boxes, and the CSB-1 are indicated as small boxes. The numbering system refers to the alignment of the 45 species. The gaps that are present in domain I and domain II are discussed in the text and shown as small slashed boxes. Sizes of the CR sequence for the three regions are indicated below each species name. of which 1114 are variable and 790 informative for significant rate differences among our sequences parsimony when all events are considered. To test (P > 0:05), which suggests that long-branch attraction is whether a robust phylogenetic signal was present in this not a problem in this data set. We notably compared data set, we recorded g1 statistic values after con- suspicious groupings by relative-rate tests (Robinson- structing 10 independent sets of 1000 random trees using Rechavi and Huchon, 2000). Taken together, all these PAUP. We obtained g1 statistics 1:0502772 data indicate the existence of a phylogenetic signal even 0:0463358 for the complete data set of 45 species and of for the profound dichotomies in the tree and clearly 0:5362543 0:0982904 for the reduced data set. Both support the use of the control region for Anseriformes values are robust according to Hillis and Huelsenbeck phylogeny. (1992), suggesting that the control region of Anserifor- mes contains some structured signal. To investigate 3.2. Phylogenetic reconstruction using the control region saturation we used the method developed by Hackett sequences (1996) (see Section 2). We obtained a linear increase of both transitions and transversions as sequence diver- The pairwise deletion scheme results in a much better gence increases (Fig. 3). We thus conclude that there is resolved tree of the reduced data set (compare Figs. 4 A no evidence of saturation in our data set, an observation and B). For example, in the tree constructed after global that is in accordance with the range of sequence diver- gap removal, the clustering of Melanitta, Callonetta, gence that we observed (ca. 30% at most). There are no Bucephala, Mergus, Somateria, and Clangula that we observed in the pairwise deletion scheme supported by 66% bootstrap values is not found. We also found that the resolution of the tree containing the whole control region is much better than any isolated domain or combination of domains (not shown). For all other analyses we thus used the complete control region with a pairwise deletion scheme when applicable. For neighbor-joining, with corrections for multiple substitutions, we observed very little influence on topology or robustness (not shown). We used the LogDet distance for all subsequent distance analyses. The overall topology of the tree is identical for NJ (Fig. 4B), MP (Fig. 4C), and ML (Fig. 4D) analyses. We Fig. 3. Analysis of the saturation present in the 45-species data set. For found Chauna torquata and Dendrocygna bicolor at ba- each pair of species the number of observed differences in the distance sal positions in both types of analysis. This basal matrix was plotted against the number of inferred substitutions that is placement was confirmed by the rooting of the tree with given by the patristic distance after a parsimony analysis. The upper outgroup sequences of other bird orders (Galliformes, points separated from the main plot correspond to the comparison of Anserinae with Chauna torquata, whereas the lower points corresponds Gruiformes, and Passeriformes). In a tree based on the to the comparison of Anatinae with Chauna torquata. central region only, due to sequence divergence of the 346 C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356

Fig. 4. Phylogenetic reconstruction of the relationships among the various genera of waterfowl from the reduced data set containing CR sequences from only 24 species. (A) Distance analysis calculated with the neighbor-joining method using a LogDet distance and a global gap removal option. A total of 630 sites remain in this analysis; 1000 bootstrap replicates were calculated. (B) Neighbor-joining analysis using a LogDet distance and a pairwise gap removal option (Hillis et al., 1996). A total of 988 sites remain in this analysis; 1000 bootstrap replicates were calculated. (C) Maximum- parsimony analysis calculated using PAUP with a pairwise gap removal option. A total of 1100 sites remains; only 100 bootstrap replicates were performed and the bootstrap tree is shown. Two equally parsimonious trees were obtained (tree length 4259). The tribes discussed in the text and the two subfamilies Anatinae and Anserinae are indicated when they are supported by bootstrap values above 50%. Only bootstrap values above 50% are indicated. (D) Maximum-likelihood estimation performed using quartet puzzling as implemented in TREE-PUZZLE. The numbers on each branch indicate quartet puzzling support values. Unresolved branches according to likelihood criteria were polytomized. C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356 347 domains I and II, we consistently found C. torquata etta, and Marmaronetta group is strongly supported basal to all the Anseriformes. In all subsequent analyses with MP (94% bootstrap), although weakly with NJ Chauna was thus used as the outgroup sequence. It is (BP < 50%) and not at all in ML. Tadorna and related thus clear from these results that D. bicolor does not species appear as a basal offshoot of Anatinae, with belong to the Anserinae as proposed by Del Hoyo et al. strong support in MP (99%) but low in NJ, and is not (1992). Most species split into two groups, which cor- found in ML. respond to the Anatinae and Anserinae subfamilies. The analysis of the complete data set by NJ (Fig. 5A) Within Anserinae all trees also give congruent and or MP (Fig. 5B) gives essentially the same results, sug- robust results, with two groups of species: the geese gesting a very weak influence of species sampling for this (Anser and Branta) in one group and the swans and phylogeny. The NJ and MP trees again found Chauna Cape barren goose (Cygnus, Coscoroba,andCereopsis) and then the two Dendrocygna as basal species and then in the other. Within this group Coscoroba coscoroba and the Anatinae/Anserinae split. Within Anserinae, the Cereopsis novae-hollandiae are sister species. various Anser are clearly monophyletic and closely re- The situation is more complex for the larger Anatinae lated to Branta. In both analyses Coscoroba and Cere- subfamily. Groups found in all types of analysis include opsis cluster together, suggesting that, in this group, Anas + Lophonetta, Aix + Cairina, Bucephala + Mergus, Cygnus diverged first. Within Anatinae both analyses Somateria + Clangula, and Alopochen + Tadorna + Chlo- found the tribes Anatini (Anas and Lophonetta), Ay- ephaga. Melanitta and Callonetta group with high thyini (Aythya and Netta), Cairinini (Cairina and Aix), bootstrap support in MP. The grouping of Cairina and and Mergini (Mergus, Bucephala, Melanitta, Callonetta, Aix with the Anas, Lophonetta, Netta, Aythya, Chenon- Somateria, and Clangula). The topology inside the

Fig. 5. Phylogenetic reconstruction of the relationships among the 45 studied species of waterfowl from the complete data set of CR sequences. (A) Distance analysis calculated with the neighbor-joining method using a LogDet distance and a pairwise gap removal option. A total of 997 sites remain in this analysis; 1000 bootstrap replicates were calculated. (B) Maximum-parsimony analysis calculated using PAUP with a pairwise gap removal option. A total of 1218 sites remains; only 100 bootstrap replicates were performed. The tribes discussed in the text and the two subfamilies Anatinae and Anserinae are indicated when they are supported by bootstrap values above 50%. Only bootstrap values above 50% are indicated. 348 C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356

Mergini is different in NJ and MP and this tribe was not and Marmaronetta; (iv) Mergini with six genera; and (v) found in the MP analyses of the reduced data set (see Tadornini which is the first to diverge inside the Anatinae. Fig. 4C) but is observed in the ML analyses (Fig. 4D). Both MP and NJ analyses also found Lophonetta inside 3.3. Comparison of control region-based phylogeny with the Anas genus, an observation that was already made ND2 and cytb data by Johnson and Sorenson (1998) using the ND2 and cytb genes. The tribe Tadornini is found in NJ with 82% Since ND2 and cytb, two protein-coding mito- support, but not in MP. Yet, this tribe was robustly chondrial genes, were sequenced and analyzed in 18 found in the MP analysis of the reduced data set. Anatinae species belonging to the main tribes (Johnson Taken together these results suggest the existence of and Sorenson, 1998) we compared the topologies found five tribes in the Anserinae (Fig. 6): (i) Anatini and (ii) using these two genes with those found using the control Aythyini which are linked; (iii) Cairinini which forms a region (Fig. 7). We also analyzed a data set containing monophyletic group with Anatini, Aythyini, Chenonetta, cytb, ND2, and the control region together. The NJ tree based on control region sequences for the 18 species found the same clustering as the 24- or 45-species data sets (compare Fig. 7A with Figs. 4B and 5A, respectively), with the tribes Anatini and Mergini well supported and Tadornini recovered with less than 50% BP. The alignment of the two protein-coding genes (Cytb/ ND2) comprised a total of 2103 sites, of which 790 were variable and 623 phylogenetically informative for parsimony. Since Johnson and Sorenson (1998) show that there are no differences between cytb and ND2 with respect to their phylogenetical signal, we combined them. Comparing TS and TV, and first and second codon position versus the three positions, we found that the most robust result was found using the three codon positions and all differences (not shown), by NJ with LogDet distance (Fig. 7B), or by maximum-parsimony (not shown). The topologies of the trees that we obtained with cytb and ND2 are comparable with those described by Sorenson et al. (1999a) using a large number of sequences. In both NJ and MP analyses, the resolution power of these two genes appears very weak compared to that of the control region. The only grouping found using NJ or MP was that of Anas with Lophonetta (Anatini) and Bucephala with Melanitta, which is not observed using the control region. In the MP tree we also noticed the grouping of Marmaronetta with Aythya, which is not found in the control region. From these data it appears that the control region is a much better marker with which to trace back phylogenetic relationships among Anatinae than the protein- coding cytb and ND2 genes. Of note, the combined analysis of cytb, ND2, and CR does not improve (and even appears to decrease) the resolutive power of CR alone (Fig. 7C). This analysis Fig. 6. Schematic phylogeny of the Anseriformes that summarizes the again recovers the Anatini, but the Mergini are not main conclusions of our study. For each branch, bootstrap values found in the complete data set are indicated. The values found by the supported when the three genes are used together. distance analysis (Fig. 5A) are indicated above the branch, whereas Ironically, Aythya and Marmaronetta on the one hand those found by MP (Fig. 5B) are indicated below. The star for the and Melanitta and Bucephala on the other hand are value (55) found by MP in the branch connecting the three genera of found together as for cytb/ND2 alone. We thus conclude the Tadornini indicates that this value was found only by the study of from these data that the analysis of the control region the reduced data set (Fig. 4C). The value of the corresponding branch for the complete data set is below 50%. Branches that are unstable and/ which contains only ca.1400 bp alone is a better strategy or for which all bootstrap values are below 50% are collapsed. The with which to resolve Anseriformes phylogeny than the various tribes and subfamilies are indicated by brackets. analysis of cytb and ND2 which contains ca. 2103 bp. C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356 349

Fig. 7. Comparison of the resolutive power of the mtDNA control region (CR) and two mitochondrial proteins-coding genes ND2 and cytb. A data set of 18 species of Anatinae for which both CR and cytb/ND2 sequences were available has been studied. In all cases the analysis was performed using the neighbor-joining method using a LogDet distance and a pairwise gap removal option. To assess the robustness of the branches 1000 bootstrap replicates were calculated. (A) Tree obtained with the CR sequences. A total of 990 sites remain in this analysis. (B) Tree obtained with the ND2/Cytb sequences, A total of 2103 sites remain in this analysis. (C) Tree obtained with the ND2/Cytb associated with the CR sequences. A total of 3100 sites remain in this analysis. The tribes discussed in the text and the two subfamilies Anatinae and Anserinae are indicated when they are supported by bootstrap values above 50%. Only bootstrap values above 50% are indicated. 350 C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356

4. Discussion the CR may be valuable in relatively deep phylogeny reconstruction. This is true even with divergences as 4.1. The control region as a phylogenetic marker for bird high as 20%, as long as alignment is satisfactory, which phylogeny is the case for Anseriformes. Thus, the CR, although shorter than the association of cytb and ND2, appears The control region has been classically divided into as a promising tool for future phylogenetic studies. three subregions: domain I in 50, domain II in 30, and a Recent studies suggest that increased taxonomic central domain. These regions differ in their base com- sampling may improve recovery of higher-level trees, position and in rate and mode of evolution (Baker and although the importance of increased taxon sampling is Marshall, 1997; Lee et al., 1995). Due to its relatively fast debated (Graybeal, 1998; Lecointre et al., 1993; Poe and rate of evolution, the CR has been typically found to be Swofford, 1999). Our results suggest that the resolution more appropriate for intraspecific studies, especially in of phylogeny is effectively better when more samples are mammals (see Quinn (1997) for a review). It is less well included, since the bootstrap value are improved, par- appreciated that it can also resolve phylogenetic rela- ticularly at deeper nodes in the tree (compare Fig. 5 with tionships at much deeper levels. Nevertheless, several Fig. 4). Although this remains to be systematically tes- recent studies have highlighted its potential in recovering ted in the case of Anseriformes, it suggests that taxon phylogeny at the family level (see Douzery and Randi sampling has a much more visible affect on phylogenetic (1997) and Saunders and Edwards (2000) for specific results than, for example, the type of distance correction examples in Cervidae and Corvidae, respectively). The used in NJ analysis. Again this confirms the analysis main structural and evolutionary features of the control done on New World jays that suggests that the partic- region of Anseriformes, such as the division into three ular weighting scheme used has a much more modest regions with different base composition, the variable impact on tree robustness than taxon sampling (Saun- amounts of gaps in these regions, the structure of the ders and Edwards, 2000). Our data set also confirms conserved blocks, the respective amounts of transitions that increasing sequence size increases resolution since and transversions, or the average intrageneric diver- the use of the three domains gives rise to better-resolved gence, are similar to those described for other birds such trees than the separate use of each domain. as Corvidae (Saunders and Edwards, 2000) and other groups (Baker and Marshall, 1997). Indeed, the CR has 4.2. Chauna and Dendrocygna as three early diverging been recently demonstrated to be very efficient in re- genera covering the phylogeny of New World jays (Saunders and Edwards, 2000). The comparison of the dynamics of All the tree topologies based on CR, irrespective of CR and cytb made by these authors has revealed that the sampling or the method used, place C. torquata at the saturation of transitions is less of a problem in the CR basal position of Anseriformes. The position of the data than in the third codon positions of cytb. screamers as an early offshoot within Anseriformes has In accordance with these recent studies, our results been recognized widely by morphological studies (Del clearly show that the control region is a useful tool with Hoyo et al., 1992; Livezey, 1997b; Sibley and Ahlquist, which to construct a robust phylogeny even at a rela- 1990) and has lead to comparison with other avian or- tively deep level, such as families, in Anseriformes. We ders in attempts to discover the origin of Anseriformes observed that the trees obtained using the control region (Olson and Feduccia, 1980). It is generally believed that either by the NJ or by the MP methods are consistently there are enough synapomorphies for screamers to be more stable (i.e., less variable when the sampling or the designated a distinct family, comprising three exclusively tree reconstruction methods are changed), more re- South American species. Another molecular analysis has solved (i.e., fewer nodes with bootstrap values below also confirmed this basal placement (Sraml et al., 1996). 50%, irrespective of the sampling or method used), and The whistling ducks (Dendrocygna) diverged more re- more robust (i.e., the resolved nodes are supported by cently from the main lineage and represent one of the higher bootstrap values) than the trees constructed using most distinctive genera of the Anatidae. Several mor- protein-coding genes such as ND2 or cytb. As discussed phological features such as erect posture, relatively above, the phylogeny that we obtained with the CR data elongated necks and legs, and conspicuous perching tree is reasonable given the debated issues with regard to habits distinguish them from most other waterfowl Anseriformes phylogeny. Even if a detailed comparison (Delacour, 1954). Our molecular results corroborate is still impossible given the large difference of taxonomic morphological phylogenies, suggesting that this group sampling between the CR and the cytb/ND2 data sets, it diverged from other Anatidae earlier than the Anatinae/ seems that well-resolved nodes in both phylogenies are Anserinae split (Livezey, 1997b; Madsen et al., 1988). The in agreement, suggesting that no obvious conflict exists separation of Dendrocygna from Anserinae is also con- between the two types of data. These points support the sistent with an early divergence of the whistling ducks idea that fast-evolving DNA sequences such as those of based on allozyme data (Numachi et al., 1983), DNA– C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356 351

DNA hybridization (Sibley et al., 1988; Sibley and Ahl- the Southern Hemisphere, and Cygnus comes from the quist, 1990), and analysis of concatenated mtDNA frag- Northern Hemisphere. It would be interesting to study a ments from three different genes (Sorenson et al., 1999a). larger sample of species from Cygnus to confirm this position, notably to test the monophyly of Cygnus. 4.3. Two subfamilies: Anserinae and Anatinae The position of Coscoroba has also been much disputed. Johnsgard (1978) used behavioral characteristics Within Anatidae, our analysis supports the conven- to place this species in the Anserini tribe (geese and tional division between Anatinae (Anas, Lophonetta, swans). In an extensive morphological study, Livezey Netta, Aythya, Chenonetta, Marmaronetta, Aix, Cairina, (1986) found only 6 characters of 120 studied, supporting Melanitta, Callonetta, Bucephala, Mergus, Somateria, a sister group relationship between Coscoroba and Ciangula, Alopochen, Tadorna, and Chloephaga) and swans, but his topology of Cygnus remains unresolved. Anserinae (Anser, Branta, Cereopsis, Coscoroba, and More recently, the complete mitochondrial srRNA gene Cygnus). This basal dichotomy, is on the one hand was shown to support the branching of Coscoroba prior strongly supported in all of our analyses with high to the divergence of geese and swans or, depending on bootstrap values and on the other hand confirmed by the method used, the association with Cygnus (Zimmer several insertion/deletion events. For example, we et al., 1994). These authors discuss the relatively close observed a large deletion on the CR sequence of ca. 100– branching times among Coscoroba, swans, and geese. 130 bp in Anatinae compared to Anserinae. This di- Our data allow a firm resolution of the branching orders chotomy between Anserinae and Anatinae was also among Coscoroba and Cygnus, since in all cases we found observed by other molecular studies (Sorenson et al., Cygnus splitting out first and then the clade Cereopsis 1999a), but based on a relatively small set of Anserinae. and Coscoroba. Nevertheless, the short length of the Our results strongly favor the definition of Anserinae branch connecting Cygnus, Coscoroba, and Cereopsis given by Livezey (1986), with Anserinae paraphyletic to and the low bootstrap value of this branch in parsimony the rest of the family, in contrast to the monophyly suggest that the Cygnus lineage diverged rapidly after the suggested by Delacour and Mayr (1945). The majority split between geese and Cygnus/Coscoroba/Cereopsis. of the convergences of this group are associated with adaptations for diving (see Fig. 1B). Indeed, in all other 4.5. Five main clades inside Anatinae morphological analyses Anserinae either contains Dendrocygna (Delacour and Mayr, 1945; see Fig. 1A) or Within the Anatinae, we found five consistent clades is totally different (Sibley and Ahlquist, 1990). when all types of analyses and/or data sets are considered (Fig. 6): (i) Anatini (Anas and Lophonetta) and (ii) 4.4. Relationships within Anserinae: The problem of Aythyini (Aythya and Netta), which form a larger clade; Cygnus, Coscoroba, and Cereopsis (iii) Cairinini (Cairina and Aix) + Anatini + Aythy- ini + Chenonetta + Marmaronetta; (iv) Mergini, with six In all of our analyses Cygnus diverged first, and C. genera (Mergus, Bucephala, Melanitta, Callonetta, So- Coscoroba and C. novae-hollandiae are sister species, materia, and Clangula); and (v) Tadornini (Tadorna, whereas traditionally Coscoroba and Cygnus are con- Alopochen, and Chloephaga), which is the first to split sidered sister species (Del Hoyo et al., 1992; Livezey, from the basal Anatinae lineage. 1997b; Sibley and Ahlquist, 1990). The unique species of Phylogenetic relationships of the tribe Anatini (dab- the genus Cereopsis, the Cape barren goose (C. novae- bling ducks) remain controversial despite intensive study hottandiae), is an Australian endemic goose of disputed (Johnson and Sorenson, 1998, 1999; Livezey, 1991). affinities. It was formerly considered an aberrant shel- Livezey recognized the tribe Anatini in which he in- duck and thus included in the tribe Tadornini (Delacour cluded all of the dabbling ducks and many of the and Mayr, 1945). It is now more commonly regarded as perching ducks (Anas, Lophonetta, Cairina, Aix, Cal- distantly related to the swans and true geese: it is oc- lonetta, and Chenonetta; see Fig. 1B). He classified the casionally included in the tribe Anserini, but more often genus Anas and a few other closely related genera separated in its own tribe, Cereopsini (Del Hoyo et al., (Amazonetta, Callonetta, Lophonetta, Speculanas, and 1992; Livezey, 1997b). This species has never been in- Mareca) in the subtribe Anateae. Other authors either cluded in molecular analyses. The position that we ob- did not resolve the distribution of Anatinae into tribes serve, closely related to C. coscoroba, was never (DNA–DNA hybridizations; Sibley and Ahlquist, 1990) observed previously. However, Livezey (1997b) men- or found Anas allied to Lophonetta and Marmaronetta tions an unpublished phylogeny of Anseriformes by in the tribe Anatini (Del Hoyo et al., 1992). Our mo- Harshman, which places Coscoroba and Cereopsis as lecular phylogenies do not support any of these views, sister genera as in our CR-based trees. This cluster is in but are in accordance with a recent detailed phylogeny accordance with the geographical origin of these species of dabbling ducks based on ND2 and cytb (Johnson and since C. Coscoroba and C. novae-hollandiae come from Sorenson, 1999; Sorenson et al., 1999a). We found that 352 C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356

Lophonetta is closely related to (parsimony; see Fig. 5B) etta,andChenonetta. According to Del Hoyo et al. or even located inside (NJ analysis, 79% bootstrap; see (1992), Aix and Cairina are clustered with Chenonetta Fig. 5A) the Anas genus. In the study of Johnson and and Callonetta since these birds have more characteris- Sorenson (1999) using ND2 and cytb the position of tics in common with each other than they have with the Lophonetta related to Anas is not robustly resolved. In members of any other tribe, particularly in the aspects of our trees using ND2 and cytb (see Fig. 7) based on a general behavior and breeding biology. This group has a more limited number of species we found that Lophon- cosmopolitan distribution and is most closely related to etta is included within Anas with a relatively low boot- the dabbling ducks (Delacour and Mayr, 1945; Del strap support. The close relationship between Hoyo et al., 1992). Livezey (1997b) includes Aix and Lophonetta and Anas was also found in morphological Cairina in the Anatini but proposes a subtribe, Cairin- analysis since in some works the crested duck Lophon- ina, clustering these two species together on the basis of etta specularoides is called Anas specularoides. Cairina, a single osteological synapomorphy. Our molecular re- Aix, and Callonetta are clearly excluded from the Ana- sults supported this view since we found that the two tini in all our trees. The case of Marmaronetta and genera always grouped with high bootstrap support, and Chenonetta is less clear since the position of these species this tribe grouped with Anatini and Aythyini. Other remains unresolved. It is clear that both genera are re- molecular analyses based on three concatenated short lated to Anatini, Aythyini, and Cairinini but their pre- mtDNA fragments confirmed the close association be- cise affiliation remain unknown. We thus cannot tween Aix and Cairina (Sorenson et al., 1999a; Sraml formally reject the definition of Anatini proposed by Del et al., 1996) but, in contradiction with most morpho- Hoyo et al. (1992) (Anas, Lophonetta, and Marmaron- logical studies, found this group related to Tadornini etta) although we find no statistical support for it. and Mergini with low bootstrap support (52 and 54%; Relationships within the genus Anas are rather intri- Sorenson et al., 1999a). Our data are in accordance with cate, as some species have very wide geographical ranges morphological data, although the relatively low boot- and occur in a number of strains such as the mallard strap support suggests that a more thorough analysis, (Anas platyrhynchos). Molecular phylogeny divides the including a more complete sampling, may be needed to dabbling ducks into several groups that are strongly confirm or exclude this proposal. supported (Johnson and Sorenson, 1999). The pintails The fourth clade comprises the Mergini (Mergus, (Anas bahamensis/Anas acuta), the wigeons (Anas stre- Bucephala, Melanitta, Clangula, Somateria, and Cal- pera/Anas sibilatrix), and the mallard (A. platyrhynchos) lonetta). Traditionally, Callonetta, which contains a represent the major clade of Anatini. The remaining unique species, Callonetta leucophrys, is associated with species, green-winged teals (Anas crecca) and blue-win- the Anatini (Livezey, 1997b; Sibley and Ahlquist, 1990) ged ducks (Anas clypeata), are unresolved in the tribe or the Cairinini (Del Hoyo et al., 1992; Johnsgard, 1978) Anatini. Our analyses based on CR and on cytb/ND2 but this species has never been ascribed to the tribe also found that A. acuta grouped with A. bahamensis Mergini and is closely related to Melanitta, as suggested and that A. sibilatrix grouped with A. strepera. The in our strongly supported MP analysis. The unambigu- positions of the other studied species are less clear, ous position of Callonetta within Mergini in our CR whereas we consistently found A. crecca and A. clypeata phylogeny was confirmed in the cytb/ND2 tree, since we as sister species, an association which is not resolved found Callonetta associated with either Melanitta or using cytb/ND2 (Johnson and Sorenson, 1999). This Bucephala, and is found using distance, MP, and ML again highlights the strong resolutive power of the CR analyses (Fig. 4). The position of this species in an in- when compared with the protein-coding genes. dependent study (Johnson and Sorenson, 1999) clearly The second tribe that we recover is Aythyini, with excludes the placement that we observed in the CR tree Aythya and Netta. Del Hoyo et al. (1992) divided the as the result of a misidentification or a contamination. modern pochards (Aythyini) into these two genera, The phylogenetic relationships of the remaining species whereas Livezey (1996), by the analysis of skeleton, of modern sea ducks (Mergini) based on control region trachea, natal plumage, and definitive integument, sequence confirmed the previously reported composition placed Marmaronetta inside this tribe, a suggestion first of the group (Del Hoyo et al., 1992; Livezey, 1997b). made by Johnsgard (1961). Our molecular analysis is in The sawbills, Mergus, is monophyletic and despite their accordance with the association of Netta and Aythya but markedly different external appearance, they seem to be we found no support for the inclusion of Marmaronetta closest to the goldeneyes (Bucephala; Johnsgard, 1978; in this tribe since the position of this species remain Livezey, 1995). The eiders (Somateria) are sometimes unresolved in our analysis. It will be probably important separated from the rest of the sea ducks in their own to sample other species closely related to Aythya, Netta, tribe Somateirini, (Delacour, 1959; Cramp and Sim- and Marmaronetta to correctly resolve this issue. mons, 1977). More recently, Livezey (1995) presented a The third tribe, Cairinini, grouping Aix and Cairina, phylogenetic analysis of modern Mergini using charac- forms a large clade with Anatini, Aythyini, Marmaron- ters of the skeleton, trachea, and natal and definitive C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356 353 plumage. On that analysis, Somateria is monophyletic logeny which is reasonably congruent with previous and constitutes the sister group of all other sea ducks in morphological analysis. This suggests that analysis of a subtribe Somaterina. Our analysis clearly suggests that the remaining species with the same method and using Somateria is close to Clangula and that both genera other genes, including nuclear genes, will probably form an early offshoot inside the Mergini. contribute to further clarify the relationships inside this Tadornini contains Tadorna, Alopochen, and Chlo- group. It is interesting to note that specific problems of ephaga and is the sister group of all other Anatinae relationships between living species of Anseriformes can tribes with moderate support in our study (51–72% also benefit from the study of extinct species using an- bootstrap). The monophyly of the tribe itself is better cient DNA analysis. This kind of analysis has already supported (from 55 to 89% bootstrap). Our phyloge- proven to be useful in the study of the moa-nalos from netic relationships inferred in the molecular analysis of Hawaii (Sorenson et al., 1999a) and will probably be the CR agree with most recent classifications, separat- fruitful for other extinct taxa. ing the sheldgeese (Chloephaga and Alopochen) and shelducks (Tadorna) (Livezey, 1997a). The southern hemisphere shelgeese are considered ‘‘intermediate’’ Acknowledgments between Anserinae and Anatinae in anatomy and behavior by Delacour and Mayr (1945) and Livezey We are grateful to Michel Saint Jalme, Patrick (1986), Alopochen had been clearly separated from Rambaud, Yves Gaumetou, and Geerard Guy for help in Tadorna by the allozyme study of Numachi et al. collecting the specimens used in this study and to (1983) and associated with the Anserini on behavioral Aureelie Theenot for invaluable technical help. We thank grounds by Johnsgard (1961). Nevertheless, the two Ceecile Mourer-Chauvire Marc Robinson-Rechavi and main morphological classifications depicted Fig. 1 Ludovic Orlando for critical reading of the manuscript consider Tadorna, Chloephaga, and Alopochen to form and two anonymous reviewers for helpful comments. a monophyletic group, in accordance with our molec- We warmly appreciate the implication of the grand- ular analysis. The case of the other genera of the mothers for babysitting during the redaction of the Tadornini tribe such as Cyanochen, which we have not manuscript. We thank CNRS, MENRT, UCBL, IBL, studied, is probably more problematic (see Sorenson and ENS-Lyon for financial support. et al., 1999a). In our phylogeny we cannot resolve correctly the trichotomy among Tadorna, Alopochen, and Chloephaga, which suggests that the three genera Appendix A originated from a rapid cladogenesis event. The detailed analysis of the relationships among the Different Taxonomic Arrangements According to (A) 24 studied genera of Anseriformes inferred from our Del Hoyo et al. (1992), (B) Livezey (1997b), and (C) analysis of the mtDNA control region supports a phy- Sibley and Ahlquist (1990)

(A) Suborder Anhimae Family Anhimidae Anhima, Chauna Suborder Anseres Family Anatidae Subfamily Anseranatinae Anseranas Subfamily Anserinae Tribe Dendrocygnini Dendrocygna, Thalassornis Tribe Anserini Branta, Anser, Cygnus, Coscoroba Tribe Cereopsini Cereopsis Tribe Stictonettini Stictonetta Subfamily Anatinae Tribe Tadornini Cyanochen, Chloephaga, Alopochen, Neochen, Tadorna Tribe Tachyerini Tackyeres Tribe Cairinini Sarkidiornis, Pteronetta, Cairina, Plectropterus, Nettapus, Callonetta, Amazonetta, Chenonetta, Aix Tribe Merganettini Merganetta Tribe Anatini Anas, Lophonetta, Hyemenolaimus, Malacorhynchus, Marmaronetta Tribe Aythyini Netta, Aythya Tribe Mergini Somateria, Polysticta, Melanitta, Histrionicus, Clangula, Bucephala, Mergus Tribe Oxyurini Oxyura, Biziura, Heteronetta 354 C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356

(B) Suboder Anhimae Family Anhimidae Anhima, Chauna Suboder Anseres Family Anseranatidae; Anseranas Family Anatidae Subfamily Dendrocygninae Tribe Dendrocygnini Dendrocygna Tribe Thalassornithini Thalassornis Subfamily Anserinae Tribe Cereopsini Cereopsis Tribe Anserini Branta, Anser Tribe Cygnini Cygnus, Coscoroba Subfamily Stictonettinae Stictonetta Subfamily Tadorninae Tribe Merganettini Hyemenolaimus, Merganetta, Tachyeres Tribe Plectropteini Plectropterus, Sarkidiornis Tribe Tadornini Subtribe Tadornina (Tadoma), Subtribe Chloephagina (Cyanochen, Alopochen, Neochen, Chloephaga) Subfamily Anatinae Tribe Malacorhynchini Malacorhynchus Tribe Anatini Subtribe Cairinina (Cairina, Pteronetta, Aix), Subtribe Nettapodina (Chenonetta, Nettapus), Subtribe Anatina (Amazonetta, Callonetta, Lophonetta, Anas) Tribe Aythyini Subtribe Marmaronettina (Marmaronetta), Subtribe Rhodonessina (Netta, Rhodonessa), Subtribe Aythyina (Aythya) Tribe Mergini Subtribe Somaterina (Somateria, Polysticta), Subtribe (Histrionicus Melanitta, Clangula, Bucephala, Mergellus, Mergus, Lophodytes, Camptorhychus) Tribe Oxyurini Subtribe Heteronettina (Heteronetta), Substribe Oxyurina (Nomonyx, Oxyura, Biziura)

(C) Infraoder Anhimides Superfamily Anhimoidae Family Anhimidae Anhima, Chauna Superfamily Anserantoidea Family Anseranatidae Anseranas Infraoder Anserides Family Dendrocygnidae Dendrocygna, Thalassomis Family Anatidae Subfamily Oxyurinae Oxyura, Biziura Subfamily Stictonettinae Stictonetta Subfamily Cygninae Cygnus, Coscoroba Subfamily Anatinae, Tribe Anserini Branta, Anser, Cereopsis, Cyanochen, Chloephaga, Alopochen, Neochen, Tadorna, Tachyeres, Plectropterus, Cairina, Pteronetta, Sarkidiornis, Nettapus Tribe Anatini Callonetta, Aix, Chenonetta, Amazonetta, Merganetta, Hyemenolaimus, Salvadorina, Anas, Malacorhynchus, Marmaronetta, Rhodonessa, Netta, Aythya, Somateria, Polysticta, Histrionicus, Clangula, Melanitta, Bucephala, Mergellus, Lophodytes, Mergus, Heteronetta

References (Ed.), Avian Molecular Evolution and Systematics. Academic Press, San Diego, pp. 51–79. Arctander, P., 1995. Comparison of a mitochondrial gene and a Cooper, A., Rhymer, J., James, H.F., 1996. Ancient DNA and island corresponding nuclear pseudogene. Proc. R. Soc. Lond. B. 262, 13– endemics. Nature 381, 484. 19. Cramp, S., Simmons, E.L., 1977. In: Handbook of the Birds of Europe Baker, A.J., Marshall, H.D., 1997. Mitochondrial control region and the Middle East and North Africa, vol. 1. Oxford University sequences as tools for understanding evolution. In: Mindell, M.P. Press, Oxford. C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356 355

Delacour, J., Mayr, E., 1945. The family Anatidae. Wilson Bull. 57, 3– Madsen, C.S., McHugh, K.P., De Kloet, S.R., 1988. A partial 55. classification of waterfowls (Anatidae) based on single copy DNA. Delacour, J., 1954. In: The Waterfowls of the World, vol. 1. Country Auk 105, 452–459. Life, London. Numachi, K., Wataba, M., Kakizawa, R., Kuroda, N., Utida, S., Delacour, J., 1959. In: The Waterfowls of the World, vol. 3. Country 1983. Evolutionary genetics of the Anatidae. Tori 32, 63–74. Life, London. Olson, S.L., Feduccia, A., 1980. Presbyornis and the origin of the Del Hoyo, J., Elliot, A., Sargatal, J., 1992. Handbook of the Birds of Anseriformes (Aves: Charadriomorphae). Smithson. Contrib. the World, vol. 2. New World Vultures to Guineafowl, Lynx Zool. 323, 1–24. Edicions, Barcelona. Olson, S.L., 1985. The fossil records of birds. In: Farner, D.S., King, Desjardins, P., Morais, R., 1990. Sequence and gene organization of J.R., Parkes, K.C. (Eds.), Avian Biology, vol. 8. Academic Press, the chicken mitochondrial genome: a novel gene order in higher New York, pp. 79–238. vertebrates. J. Mol. Evol. 212, 599–634. Poe, S., Swofford, D.L., 1999. Taxon sampling revisited. Nature 398, Douzery, E., Randi, E., 1997. The mitochondrial control region of 299–300. Cervidae: evolutionary patterns and phylogenetic content. Mol. Quinn, T.W., 1992. The genetic legacy of Mother Goose—Phylogeo- Biol. Evol. 14, 1154–1166. graphic patterns of Lesser Snow Goose Chen caerulescens maternal Felsenstein, J., 1985. Confidence limits on phylogenies: an approach lineages. Mol. Ecol. 1, 105–117. using the bootstrap. Evolution 39, 783–791. Quinn, T.W., Wilson, A.C., 1993. Sequence evolution in and around Galtier, N., Gouy, M., Gautier, C., 1996. SEA VIEW and the mitochondrial control region in birds. J. Mol. Evol. 33, 417– PHYLO_WIN: two graphic tools for sequence alignment and 425. molecular phylogeny. Comput. Appl. Biosci. 12, 543–548. Quinn, T.W., 1997. Molecular evolution of the mitochondrial genome. Graybeal, A., 1998. Is it better to add taxa or characters to a difficult In: Mindell, M.P. (Ed.), Avian Molecular Evolution and Syste- phylogenetical problem? Syst. Biol. 47, 9–17. matics. Academic Press, San Diego, pp. 3–28. Hackett, S.J., 1996. Molecular phylogenetics and biogeography of Ramirez, V., Savoie, P., Morais, R., 1993. Molecular characterization Tanagers in the genus Ramphocelus (Aves). Mol. Phylogenet. Evol. and evolution of a duck mitochondrial genome. J. Mol. Evol. 37, 5, 368–382. 296–310. Haanni,€ C., Brousseau, T., Laudet, V., Stehelin, O., 1995. Isopropanol Robinson-Rechavi, M., Huchon, D., 2000. RRTree: relative-rate tests precipitation removes PCR inhibitors from ancient bone extracts. between groups of sequences on a phylogenetic tree. Bioinformatics Nucleic Acids 23, 881–882. 16, 296–297. Hillis, D.M., Huelsenbeck, J.P., 1992. Signal, noise, and reliability in Saitou, N., Nei, M., 1987. The neighbor-joining method: a new method molecular phylogenetic Analyses. J. Hered. 83, 189–195. for reconstructing phylogenetic trees. Mol. Biol. Evol. 4, 406–425. Hillis, D.M., Moritz, C., Mable, B.K., 1996. Molecular Systematics, Saunders, M.A., Edwards, S.V., 2000. Dynamics and phylogenetic second ed. Sinauer, Sunderland, MA. implications of MtDNA control region sequence in New World Johnsgard, P.A., 1961. The systematic position of the Marbled Teal. jays (Aves: Corvidae). J. Mol. Evol. 51, 97–109. Bull. Br. Ornithol. Club 81, 37–43. Sibley, C.G., Ahlquist, J.E., Monroe Jr., B.L., 1988. A classification of Johnsgard, P.A., 1978. Ducks, Geese, and Swans of the World. the living birds of the world based on DNA–DNA hybridization University of Nebraska Press, Lincoln. studies. Auk 105, 409–423. Johnson, K.P., Sorenson, M.D., 1998. Comparing molecular evolution Sibley, C.G., Ahlquist, J.E., 1990. Phylogeny and Classification of in two mitochondrial protein coding genes (cytochrome b and Birds: A Study in Molecular Evolution. Yale University Press, New ND2) in the dabbling ducks (Tribe: Anatini). Mol. Phylogenet. Haven, CT. Evol. 10, 82–94. Sorenson, M.D., Fleischer, R.C., 1996. Multiple independent trans- Johnson, K.P., Sorenson, M.D., 1999. Phylogeny and biogeography of positions of mitochondrial DNA control region sequences to dabbling ducks (Genus: Anas): a comparison of molecular and nucleus. Proc. Natl. Acad. Sci. USA 93, 15239–15243. morphological evidence. Auk 116, 792–805. Sorenson, M.D., Dimcheff, D.E., Adt, J.C., Yuri, T., Mindell, D.P., Kimura, M., 1980. A simple method for estimating evolutionary rates 1998. Complete mitochondrial DNA sequences for five birds and a of base substitution, through comparative studies of nucleotide turtle (unpublished). sequences. J. Mol. Evol. 16, 111–120. Sorenson, M.D., Quinn, T.W., 1998. Numts: a challenge for avian Lecointre, G., Philippe, H., Van Le, H.L., Le Guyader, H., 1993. systematics and population biology. Auk 115, 214–221. Species sampling has a major impact on phylogenetic inference. Sorenson, M.D., Cooper, A., Paxinos, E.E., Quinn, T.W., James, Mol. Phylogent. Evol. 2, 205–224. H.F., Olson, S.L., Fleischer, R.C., 1999a. Relationships of the Lee, W.C., Coroy, J., Howell, W.H., Kocher, T.D., 1995. Structure extinct moa-nalos, flightless Hawaiian waterfowl, based on ancient and evolution of fish mitochondrial control regions. J. Mol. Evol. DNA. Proc. R. Soc. Lond. 266, 2187–2193. 41, 54–66. Sorenson, M.D., Ast, J.C., Dimcheff, D.E., Yuri, T., Mindel, D.P., Livezey, B.C., 1986. A phylogenetic analysis of recent anseriform 1999b. Primers for a PCR-based approach to mitochondrial genera using morphological characters. Auk 105, 681–698. genome sequencing in birds and other vertebrates. Mol. Phyloge- Livezey, B.C., 1991. A phylogenetic analysis and classification of net. Evol 12, 105–114. recent dabbling ducks (Tribe Anatini) based on comparative Sraml, M., Christidis, L., Easteal, S., Horn, P., Collet, C., 1996. morphology. Auk 108, 471–507. Molecular relationships within Australian waterfowl (Anserifor- Livezey, B.C., 1995. Phylogeny and evolutionary ecology of modern mes). Aust. J. Zool. 45, 47–58. seaducks (Anatidae: Mergini). Condor 97, 233–255. Steel, M.A., 1993. Recovering a tree from a leaf coloration it generates Livezey, B.C., 1996. A phylogenetic analysis of modern pochards under a Markovmodel. Appl. Math. Lett. 7, 19–23. (Anatidae: Aythyini). Auk 113, 74–93. Strimmer, K., Von Haeseler, A., 1996. Quartet puzzling: a quartet Livezey, B.C., 1997a. A phylogenetic analysis of modern shelducks and maximum likelihood method for reconstructing tree topologies. sheldgeese (Anatidae, Tadornini). Ibis 139, 51–66. Mol. Biol. Evol. 13, 964–969. Livezey, B.C., 1997b. A phylogenetic classification of waterfowl (Aves: Swofford, D.L., 1993. PAUP: Phylogenetic Analysis Using Parsimony. Anseriformes), including selected fossil species. Ann. Car. Mus. 66, Formely distributed by Illinois Natural History Survey, Cham- 457–496. paign, IL. 356 C. Donne-Gousse et al. / Molecular Phylogenetics and Evolution 23 (2002) 339–356

Taberlet, P., Bouvet, J., 1991. A single plucked feather as a source of Zhang, D.-X., Hewitt, G.M., 1996. Nuclear integrations: challenges DNA for bird genetic studies. Auk. 108, 959–960. for mitochondrial DNA markers. Trends Ecol. Evol. 11, 247–251. Thompson, J.D., Higgins, D.G., Gibson, T.J., 1994. CLUSTAL_W: Zimmer, R., Erdtmann, B., Thomas, W.K., Quinn, T.W., 1994. improving the sensitivity of progressive multiple alignment through Phylogenetic analysis of the Coscoroba coscoroba using mito- sequence weighting, position-speciﬁc gap penalties and weight chondrial srRNA gene sequence. Mol. Phylogenet. Evol. 3, 85– matrix choise. Nucleic Acid. Res. 22, 4673–4680. 91.