Phylogeny and Biogeography of Ratite Birds Inferred from DNA Sequences of the Mitochondrial Ribosomal Genes Marcel van Tuinen,* Charles G. Sibley,² and S. Blair Hedges* *Department of Biology and Institute of Molecular Evolutionary Genetics, Pennsylvania State University; and ²Santa Rosa, California The origin of the ¯ightless ratite birds of the southern continents has been debated for over a century. Whether dispersal or vicariance (continental breakup) best explains their origin depends largely on their phylogenetic rela- tionships. No consensus has been reached on this issue despite many morphological and molecular studies. To address this question further we sequenced a 2.8-kb region of mitochondrial DNA containing the ribosomal genes in representative ratites and a tinamou. Phylogenetic analyses indicate that Struthio (Africa) is basal and Rhea (South America) clusters with living Australasian ratites. This phylogeny agrees with transferrin and DNA hybrid- ization studies but not with sequence analyses of some protein-coding genes. These results also require reevaluation of the phylogenetic position of the extinct moas of New Zealand. We propose a new hypothesis for the origin of ratites that combines elements of dispersal and vicariance. Introduction The living ratites include two species of ostriches cause a UPGMA tree joined Rhea and Struthio, whereas (Struthio) in Africa and formerly in Asia, the Australian trees constructed with the Fitch-Margoliash (1967) al- emu (Dromaius), three species of cassowaries (Casuar- gorithm joined Rhea with the Australasian clade (Sibley ius) in New Guinea and northeastern Australia, three and Ahlquist 1990). In contrast, mitochondrial DNA se- species of forest-dwelling kiwis (Apteryx) in New Zea- quence data (12S rRNA; 400 bp) supported a basal po- land, and two rheas (Rhea) in South America (Sibley sition for Rhea (Cooper et al. 1992). However, reanaly- 1996). All lack a keel on the sternum, a character as- sis of those sequence data by Rzhetsky, Kumar, and Nei sociated with ¯ightlessness. Based on anatomical anal- (1995) and analysis of an additional 600 bp from a nu- yses, ratite phylogeny has been controversial for over a clear gene (Cooper and Penny 1997) indicate that such century. This stems from subjective interpretations of small data sets are unlikely to resolve the early history anatomical characters and from dif®culties in determin- of ratites. Therefore, we sequenced the complete mito- ing polarity in characters shared by ratite birds (Kuroch- chondrial ribosomal genes (2.8 kb) from representative kin 1995). Sibley and Ahlquist (1990) provide a histor- ratite species to address the question of the nearest living ical review of ratite systematics. relative of the Australasian ratites. These genes have Over the last two decades, several molecular stud- proven useful for addressing higher-level phylogenetic ies of ratites have clari®ed some aspects of ratite phy- questions in birds and other vertebrates (e.g., Hedges logeny. There is general agreement that living ratites are 1994; Hedges and Sibley 1994; Springer et al. 1997). monophyletic and that the weakly ¯ying tinamous are their closest living relatives (Prager et al. 1976; Sibley Materials and Methods and Ahlquist 1981, 1990; Caspers, Wattel, and De Jong 1994). Monophyly of the living Australasian ratites is A 2.8-kb region of mitochondrial DNA was se- also supported by molecular data (Prager et al. 1976; quenced for each of three ratite species (Rhea ameri- Sibley and Ahlquist 1981; 1990; Cooper et al. 1992), cana, Struthio camelus, and Dromaius novaehollandiae) although consensus on this issue has not been reached and a gray tinamou (Tinamus tao). We assumed mono- with anatomical data (Cracraft 1974; Bledsoe 1988; Ku- phyly of the living Australasian ratites based on molec- rochkin 1995). ular evidence (Prager et al. 1976; Sibley and Ahlquist The relationships of the three major lineages, rheas, 1990; Cooper et al. 1992), with Dromaius as the rep- ostriches, and Australasian ratites, have a direct bearing resentative of the living Australasian ratite clade. The on the biogeographic history of the ratites (Cracraft sequenced region includes the entire 12S rRNA, t- 1974), but these relationships remain unclear. Analysis RNAVal, and 16S rRNA genes. The corresponding se- of transferrin immunological data grouped the Rhea quences from a domestic fowl (Gallus gallus; accession with the Australasian clade and thus identi®ed Struthio number X52392, sites 1274±3994 in Desjardins and as the most basal living ratite (Prager et al. 1976). The Morais 1990) and an American alligator (Alligator mis- relationships of these three lineages with DNA-DNA hy- sissippiensis, accession number L28074) were obtained bridization data were considered to be unresolved be- from GenBank for comparison. The new sequences re- ported here have been deposited in EMBL with acces- Key words: molecular phylogeny, ratite evolution, vicariance, dis- sion numbers AJ002921±AJ002924. persal, Gondwana. DNA ampli®cation was performed with the follow- Address for correspondence and reprints: S. Blair Hedges, De- ing primer pairs:12L9/12H2, 12L9/12H5, 12L10/12H5, partment of Biology, 208 Mueller Laboratory, Pennsylvania State Uni- 12L1/12H4, 12L7/16H11, 16L11/16H5, 16L17/16H17, versity, University Park, Pennsylvania 16802. E-mail: [email protected]. 16L10/16H10, 16L9/16H3, 16L8/16H13, and 16L4/ Mol. Biol. Evol. 15(4):370±376. 1998 16H12 (Hedges 1994; Hedges and Sibley 1994; Hedges q 1998 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038 et al. 1995). Primers not previously described are 370 Phylogeny of Ratites 371 [IUPAC]: 12L7 (GAA GGW GGA TTT AGY AGT AAA), 12L9 (AAA GCA HRR CAC TGA ARA TGY YDA GA), 12L10 (CMC AMG GGA MWC AGC AGT GAT WAA HAT T), 12H5 (TTA GAG GAG CCT GTC CTA TAA TCG), 16L17 (CCW AMC GAR CYT RGT GAT AGC TGG TT), and 16H17 (TGT TTA CCA AAA ACA TMY CCY YYS GC). DNA was PCR-ampli®ed as follows: 30±40 cycles (948C for 15 s, 508Cor558C for 15 s, 728C for 45 s). Double-stranded DNA was gel-puri®ed and served as template for another PCR run under slightly different conditions: 25±35 cycles (948C for 15 s, 558Cor608C for 15 s, 728C for 45 s). A hot start was performed at 808C for all PCR runs. Prior to sequencing, double- FIG. 1.ÐNeighbor-joining tree of three representative ratite birds stranded DNA was ®ltered as described earlier (Hedges (Struthio camelus, Rhea americana, and Dromaius novaehollandiae), and Sibley 1994). Both complementary L- and H- a tinamou (Tinamus tao), domestic fowl (Gallus gallus), and American strands were sequenced for all primer sets. Cycle se- alligator (Alligator mississipiensis) inferred from the mitochondrial 12S, 16S rRNA, and tRNAVal genes (2.8 kb). Con®dence values are quencing reactions were performed using 39 dye-labeled indicated between brackets on the nodes in the tree (left, interior- dideoxynucleotide triphosphates (¯uorescent dye termi- branch test; middle, interior-branch test for transversions only; right, nators) and run on an ABI PRISM 377 DNA Sequencer maximum-parsimony bootstrap values). (Perkin-Elmer ABI, Foster City, Calif.). Alignments were performed with ESEE (Cabot and Beckenbach 1989). For phylogenetic analysis, neighbor- accordance with the number of varied sites and those joining (Saitou and Nei 1987) analyses were performed informative for parsimony for the three genes (16S: 705 in MEGA (Kumar, Tamura, and Nei 1994) using a Ki- varied sites, 262 sites informative for parsimony; 12S: mura (1980) two-parameter distance with transitions and 440 varied sites, 135 sites informative for parsimony; transversions included unless otherwise noted. Maxi- tRNAVal: 28 varied sites, 12 sites informative for par- mum-parsimony analyses were conducted with PAUP simony). The 12S rRNA and tRNAVal data support the (Swofford 1993), and maximum-likelihood analyses Rhea-Dromaius clade, with 71% and 70% con®dence were performed with Nucml (HKY) in MOLPHY (Ada- values, respectively, for the interior-branch analysis. chi and Hasegawa 1996). In MOLPHY, we used the de- After this study was completed, a similar study by fault transition : transversion ratio of 4:1, which is sim- Lee, Feinstein, and Cracraft (1997) was published on ilar to that estimated for mitochondrial DNA (Kumar ratite relationships inferred from DNA sequences of sev- 1996), and 100:1 (ùtransversions only). Sites corre- eral noncoding and protein-coding mitochondrial genes, sponding to alignment gaps were excluded. The interior- including a reexamination of the morphological evi- branch (SE) test was performed to assess con®dence in dence. There was partial overlap between the sequenced the reliability of the neighbor-joining trees by testing if regions (rRNA genes) in their study and ours. Compar- interior branch lengths deviate signi®cantly from zero. ison of our Struthio sequence with their sequence and In PAUP, the bootstrap method was applied (Felsenstein that of HaÈrlid, Janke, and Arnason (1997) indicated 11± 1985) with 2,000 replications. 12 nucleotide differences between each of the three se- quences. All differences between our sequence and that Results of HaÈrlid, Janke, and Arnason (1997) were transitions, whereas both of those sequences shared six transversion Out of 2,900 sites in the initial alignment of six differences with the Struthio sequence of Lee, Feinstein, taxa, 2,429 were analyzed after elimination of gaps and and Cracraft (1997). These differences did not affect the sites showing ambiguous alignment. Of those sites, there phylogeny as shown in ®gure 1. However, the sequence were 1,173 varied sites and 409 sites informative for the analysis (combined genes) of Lee, Feinstein, and Cra- parsimony method. The phylogenetic tree (®g. 1), rooted craft (1997) supported a basal Rhea (rather than Stru- with Alligator, shows high con®dence for a Rhea-Dro- thio) lineage, an unexpected discordance that we inves- maius clade. Con®dence values are signi®cant (99%) for tigated here. the interior-branch test and almost signi®cant when only An overview of the present molecular evidence (ta- transversions are used (92%).