The Auk 118(4):849-863, 2001 PHYLOGENY AND SYSTEMATICS OF GROUND ROLLERS (BRACHYPTERACIIDAE) OF MADAGASCAR JEREMYJ. KIRCHMAN,L3 SHANNON J. HACKETT,• STEVENM. GOODMAN,L2 AND JOHN M. BATES• tFieldMuseum of Natural History, 1400 South Lake Shore Drive, Chicago, Illinois 60605, LISA; and 2WorldWide Fund for Nature,Boite Postale 738, Antananarivo (101), Madagascar ABSTRACT.--Westudied relationshipsof five extant membersof the endemicMalagasy family Brachypteraciidae,the ground rollers, using severalmitochondrial genes (cyto- chrome-b,NADH dehydrogenase2, 12Sribosomal RNA, and cytochromeoxidase I). As out- groups,we usedother coraciiforms including the CuckooRoller (FamilyLeptosomatidae, Leptosomusdiscolor), several true rollers (Corcaciidae) and a tody(Todidae). Partial sequences of the Long-tailedGround Roller (LIratel0rnis chimaera) were obtained from toe pad samples taken off museumspecimens. For a combineddata set of all genes,Kimura two-parameter distancesbetween sequences of the five groundroller species were high, averaging 11% di- vergence.For several species, samples were available from widely separatedgeographic re- gionsand intraspecificsequence divergence was low (-<0.8%).Unweighted and weighted parsimonyand maximum-likelihood analyses consistently recovered monophyly of thefam- ily, a sisterrelationship between Brachypteraciidae and Coraciidae,and monophylyof one of three currentlyrecognized ground roller genera(Atelornis). At the baseof the Brachyp- teraciidaeclade, we couldnot fully resolverelationships between Uratelornis and two species currentlyplaced in Brachypteracias.Because of the uncertaintyof basalnodes in our phylo- geneticreconstructions, we recommendreturning B. squamigerato themonotypic genus Geo- biastes.High levelsof divergenceamong ground rollers are similar to levelsfound in other aviangroups endemic to Madagascar.However, we suggestthat moleculardivergences ap- pearfar too low to be consistentwith mid-Eocenefossils attributed to the family.Received 3 August2000, accepted28 February2001. MADAGASCAR LONG HAS commanded atten- phylogeneticanalyses are necessaryto better tion of biologistswith its high level of ende- understandfactors that have shapedMadagas- micity and large number of animal groupsre- car'sextraordinary endemicity. puted to be examplesof adaptiveradiations. In recentyears, there has been a dramaticin- The island last was connected to Africa -160 creasein number of field researchersconducting Mya and to India -90 Mya (Krauseet al. 1997) ecologicalstudies on Malagasybirds, particular- resultingin a long period of isolationthat al- ly the endemicradiations. Reconstructing phy- most certainlybegan prior to the evolutionof logeniesand testingmonophyly of suchgroups the vast majority of extant bird groups (Fed- is criticalto understandingtheir evolutionary his- uccia 1999). Thus, it is assumedthat ancestors tory.An exampleis the groundrollers (Brachyp- of the modern Malagasy avifaunaarrived on teraciidae),which have been the subjectof several the islandby dispersal.Given that their affini- ecologicalstudies (Appert 1968, Turner 1984, ties are believedto be primarily African (Mo- Rakotoarisoa 1998, 1999; Razafimahaimodison reau 1966, Dorst 1972, Keith 1980), dispersal 1995, Thorstrom and Lind 1999). This family acrossthe 400-550 km passageof the Mozam- long hasbeen allied to the African "true" roll- biqueChannel between Africa andMadagascar ers (Coraciidae),but little is known regarding is the most likely explanationfor the island's their phylogeny. colonization.Phylogenetic work on Malagasy Ground rollers currently are recognizedas a bird groups(Prum 1993,Houde et al. 1997,Ci- family (Brachypteraciidae,Order Coraciifor- bois et al. 1999, Fjeldsaet al. 1999,Johnson et mes) endemicto Madagascarwith five extant al. 2000) hasbegun to producea clearerpicture speciesplaced in three genera(Wetmore 1960, of the island's avifaunal history, but further Sibleyand Monroe1990). All speciesare forest- dwelling with four forms (Atelorniscrossleyi, A. 3 E-mail: [email protected] pittoides,Brachypteracias leptosomus, and B. squa- 849 850 KIRCHMANET AL. [Auk, Vol. 118 rnigera)occurring in humid forestsdominating is the sister of that clade. Although Cracraft's eastand centralhighlands, and the fifth species study primarily was concernedwith interfa- (Llratelornischirnaera) in the spiny bush of the milial relationships,he did include a hypothe- dry southwest(Langrand 1990). Both Brodkorb sis of phylogeny and "adaptive pathways" (1971) and Olson (1985) noted that there is no within the ground roller lineage.He proposed known fossil history of ground rollers. How- that Atelornis"is probablymore advancedfrom ever, well-preservedmaterial from the Eocene the primitive ground roller condition" exem- MesselFormation of Germanywas saidby Mar- plified by Brachypteracias,and that "Uratelornis tin (1983) and Feduccia (1999) to contain an ex- is unmistakedlycloser to Atelornisthan to Bra- ample of the Brachypteraciidae.Feduccia also chypteracias"(Cracraft 1971:748). Although described Eocene material from the North Cracraft did not examine characters in a cla- American Green River formation as reminis- distic framework, his work is the only modern cent of ground rollers. More recently,subfossil study to examinerelationships within the fam- material of the family has been recoveredfrom ily. Cracraft'sconclusions formed the founda- Holocene depositson Madagascar(Goodman tion of higher level relationshipsamong roller 2001). families in Maurer and Raikow's (1981) exam- Ground rollers first were considered a dis- ination of hindlimb myology within Coraci- tinct family (At61ornithin6s)by Chenu and des iformes and in Sibley and Ahlquist's (1990) Murs (1852),who placedthem at the beginning DNA-DNA hybridizationwork, both of which of the suboscinepasserines on the basisof tar- lacked a representativeof ground rollers. The sal and foot characters. The name "Atelornithi- dearth of comparative material in museums, dae," therefore, has priority over the name particularly skeletonsand tissues,has ham- "BrachypteraciidaeBonaparte, 1854." Howev- pered, until recently, further phylogenetic er, the latter name has been widely used in or- investigation. nithologicalliterature, and for the sakeof sta- The primary goal of our study was to use bility we follow Bock's(1994) recommendation DNA sequencedata from newly availabletissue that it be maintained. Ground rollers are now samples and recently developed "antique" universallyplaced in the Coraciiformesand are DNA protocolsto estimaterelationships of ex- consideredby someauthorities to be a subfam- tant groundrollers and their roller allies.In ad- ily of the Coraciidae (e.g. Peters 1945). Al- dition, we characterizepatterns of evolutionin though ground rollers differ markedly from four mitochondrialgenes increasingly used to coraciidsin their behavior,plumage, and post- reconstructphylogeny and discussrates of mo- cranial skeleton,the two groups are very sim- lecular evolutionby comparingour data to the ilar in their cranial osteology(Cracraft 1971) fossil record of rollers. Finally, we present a and in elements of the feeding apparatus classificationof extant ground rollers. (Burton 1984). Early classificationsof the family differed METHODS primarily with respectto the treatment of Bra- Taxaexamined.--We compared mtDNA sequences chypteraciassquamigera, which also was placed from the following coraciiformspecies (number of in Atelornis (Chenu and des Murs 1852, Hart- individuals in parentheses):ground rollersAtelornis laub 1861) and in the monotypic genus Geo- pittoides(5), A. crossleyi(3), Brachypteraciasleptosornus biastes(Sharpe 1871, Hartlaub 1877). However, (2), B. squarnigera(1), Uratelornischirnaera (2); the taxonomy within the family has been stable CuckooRoller (Family Leptosomatidae),Le?tosomus since the last extant form, Uratelornis chimaera, discolor(1); the true rollers (Family Coraciidae)Eu- was described in 1895. rystornusorientalis (1), E. glaucurus(1), Coraciasgar- Cracraft (1971) examined cranial and post- rulus (1); and the JamaicanTody (Family Todidae), cranial skeletal anatomy in the families Cora- Todustodus (1). In this paper,we use the term "roll- ers" to refer to all membersof the Brachypteraciidae, ciidae (true rollers), Leptosomatidae(Cuckoo Coraciidae,and Leptosomatidaecollectively. For all Roller [Leptosomusdiscolor]), and in three spe- taxa exceptUratelornis chirnaera, tissue samples were ciesof ground rollers (both speciesof Atelornis preservedin salt-DMSObuffer (Seutinet al. 1991)or and Brachypteraciasleptosomus). He concluded in liquid nitrogen at the time of specimenprepara- that the Coraciidaeis the sistertaxon of ground tion and later stored in the collectionof geneticre- rollers and that the monotypicLeptosomatidae sources of the Bird Division, Field Museum of Nat- October2001] GroundRoller Phylogeny 857[ TABLE1. Primers used in PCR amplificationand sequencingof four mitochondrialgenes. H and L refer to primerslocated on the heavy and light strandsof the mitochondrialgenome, respectively, and numbers refer to the 3' baseof the primer referencedto the completemtDNA sequenceof the chicken(Gallus gallus; Desjardinsand Morais 1990). Cyt b 5' -• 3' sequence Reference L-14990 CCATCCAACATCTCAGCATGATGAAA Modified from Kocher et al. 1989 L-15656 AACCTACTAGGAGACCCAGA Helm-Bychowskiand Cracraft 1993 L-15920 ACATGAGTCGGAAGCCAACC Helm-Bychowski and Cracraft 1993 L-15311 CTACCATGAGGACAAATATC Helm-Bychowskiand Cracraft 1993 H-15424 GGAGGAAGTGCAGGGCGAAGAATCG Hackett