sign in sign up
<<
Home , Albatross, Duck, Great albatross, Laysan albatross, Mollymawk, North Pacific albatross

The Auk 113(4):784-801, 1996

EVOLUTIONARY RELATIONSHIPS AMONG EXTANT (: DIOMEDEIDAE) ESTABLISHED FROM COMPLETE CYTOCHROME-B GENE SEQUENCES

GARY B. NUNN, •'6JOHN COOPER,2 PIERREJOUVENTIN, 3 CHRIS J. R. ROBERTSON,4 AND GRAHAM G. ROBERTSONs XDepartmentof ,American Museum of NaturalHistory, Central Park West at 79th Street, New York, New York 10024, USA; 2PercyFitzPatrick Institute of AfricanOrnithology, University of CapeTown, Rondebosch7700, ; 3CentreNational de la RechercheScientifique, Centre D'Etudes Biologique de Chize, F79360 Villiers en Bois,; 4Departmentof Conservation,Science and Research Division, ConservationSciences Centre, Wellington, ;and 5AustralianAntarctic Division, Channel Highway, Kingston, ,

ABSTR•Cr.--Completemitochondrial cytochrome-b gene sequences(1,143 bp) were deter- mined from the 14 extant speciesin the Diomedeidae(albatrosses and )and in two outgroupspecies from the ( and ). Phylogenetic analysis usingmaximum parsimony and maximumlikelihood methods identified a singlebest-sup- ported hypothesisof evolutionaryrelationships within the Diomedeidae,namely that two lineagesarose early in the of the Diomedeidae.A further bifurcationin each of theselineages resulted in four monophyleticgroups of albatrosses:(1) southernmollymawks, (2) sootyalbatrosses, (3) North Pacificalbatrosses, and (4) "great" albatrosses.Monophyly of the southernmollymawks (Diomedea bulleri, D. cauta,D. chlororhynchos,D. chrysostoma, and D. melanophris)and sootyalbatrosses ( fusca and P. palpebrata)indicates that Diomedea is paraphyletic.Resurrection of two genera,dropped historically in taxonomyof the Diome- deidae, resultsin a total of four genera. Calibrationsbased on the record indicate that cytochrome-bevolutionary rates in albatrossesare slow comparedwith thoseof mostmam- mals. Received21 August1995, accepted 10 May 1996.

THE ALBATROSSESAND MOLLYMAWKS( The discovery of a small population of an Diomedeidae) are the most familiar and best undescribed"great" albatrosson Amsterdam Is- studiedgroup of procellariiform(or tube-nosed) land in the Indian (D. amsterdamensis; seabirdsdue largely to their highly philopatric Roux et al. 1983) brought the total number of nature and diurnal attendance at breeding lo- speciesto 14 (Sibley and Monroe 1990). Much calities, where their surface nests are easily controversyexists regarding the exacttaxonom- monitored (Warham 1990). The 13 traditionally ic status of D. amsterdamensis.In particular, its acceptedspecies of albatrossesare widely dis- relationship to subspecifictaxa of D. exulans, tributed throughout the southern ,the which breed on low- islands in the North Pacific Ocean, and, in a single case,the southern Pacific and Atlantic Oceans, is not well tropical Pacific Ocean (Harris 1973, Harrison understood (Bourne 1989, Robertson and War- 1983, Marchant and Higgins 1990). Fossil evi- ham 1992). An affinity among thesetaxa seems dence of albatrossspecies present in the North likely because,to differing degrees, they share (Lydekker 1891a,b; Wetmore the retention of dark juvenal or immatureplum- 1943), which are most similar to the extant Di- age as reproducingadults ( paedomor- omedea albatrus of the North Pacific Ocean, in- phosis). In contrast, populations of the larger- dicates that the Diomedeidae once were truly sized D. exulans, which occur on cosmopolitanin oceanicdistribution. islandsat higher , have a white plum- age asbreeding adults.Additional geneticanal- ysis of individuals from all populations of ex- E-mail:[email protected] ulanswill help resolve the much-debated tax- 784 October 1996] AlbatrossMolecular Phylogeny 785

TABLE1. The introductionand principalchanges to describedgenera within the Diomedeidae.

Authority Generaand changes Linnaeus (1758) Diomedeagert. nov. Reichenbach (1852) Thalassarchegen. nov.; Phoebastriagen. nov.; and Phoebetriagen. nov. Coues (1866) SubBurned Thalassarche and Phoebastffa into Diomedea Baird et al. (1884) Thalassagerongen. nov. Mathews (1912) ResurrectedThalassarche; Diomedella gen. nov.; Nealbatrusgen. nov. Murphy (1917) Rhothoniasubgen. nov. Mathews (1934) Resurrected Phoebastria Mathews and Hallstrom SubBurnedRhothonia into Diomedea;transferred from Phoebastriato (1943) Julietatagen. nov. Mathews (1948) SubBurned all albatrosses into Diomedea Boetticher (1949) in Galapagornisgen. nov.; Laysanornisgert. nov.; Penthireniagert. nov. Jouaninand Mougin (1979) Alexander et al. (1965) Standardized use of Diomedea and Phoebetria onomy of the amsterdamensis-exulanscomplex (G. higher-level phylogeneticrelationships among Nunn unpubl. data). the 14 extantspecies of Diomedeidae.The choice Based on elements of biogeographic distri- of the mitochondrial cytochrome-b(cyt-b) gene bution, a simple allometricrelationship of wing as an evolutionary marker for our study was and tail length, and charactersof the divided basedon severalfactors. First, the completegene plates making up the rhamphothecaof the bill, sequence,as well as flanking regions, are well the 14 albatrossspecies fall into four natural characterizedin (Desjardins and Morals groups: (1) the southern mollymawks, (2) the 1990, Helm-Bychowskiand Cracraft 1993, Kor- North Pacificalbatrosses, (3) the "great" alba- negay et al. 1993, Nunn and Cracraft 1996) and trosses,and (4) the sooty albatrosses(Warham other vertebrate groups (Jermiin et al. 1994), 1990).On the basisof completeadult fuliginous enabling the design of oligonucleotideprimers plumage coloration, longer wedge-shapedtail, that amplifyby polymerasechain reaction (PCR) cuneate body form, and presenceof a colored in a broad phylogenetic range of birds. Second, fleshy sulcus separating the ramicorns of the the fast evolutionary rate of changein cyt-bhas lower mandible (a morphologicalfeature found proven most suitable for studying recently di- in other procellariiforms),a traditional hypoth- vergent groups (Meyer 1994) and in birds has esis of relationships within the Diomedeidae successfullyresolved relationships from the recognizesa simple demarcationof two genera: specieslevel (Richman and Price 1992, E. Smith the "primitive" sooty albatrossgenus Phoebetria et al. 1992, Blechschmidtet al. 1993) to generic and a more comprehensivegenus Diomedea that and familial levels (Krajewskiand Fetzner 1994, envelopes the North Pacific albatrosses,the Lanyonand Hall 1994,Murray et al. 1994).Third, "great" albatrosses,and the southern molly- cyt-bis one of the larger protein-codinggenes mawks (Coues1866, Peters 1931, Murphy 1936, in tl•e avian mitochondrial genome(Desjardins Alexanderet al. 1965,Jouanin and Mougin 1979, and Morals 1990), and, so far, presentsno prob- Warham 1990).Consideration of other morpho- lem of alignment amongbirds. Finally, the ex- logicalfeatures historically have led to the split- panding use of cyt-bgene sequencesas a source ting of current members of Diomedeainto ad- of qualitative data for studiesof avian system- ditional generic groups (seeTable 1), although atics ensures that in the near future a dense none has gained common acceptance.Indeed, sampling of diversetaxa will be available,lead- Mathews (1948)went on to producean entirely ing to a common improvement of phylogeny- lumped albatrossgenus Diomedeacomprised of building within birds. all known ,including Phoebetria. Our genetic study explored several basic In view of both the traditional hypothesisof questionsconcerning the patterns and rates of albatrossrelationships based on a small number evolution among extant albatrossspecies: (1) Is of morphological features in this conservative the traditional classificationcongruent with a group, and the confusingtaxonomy within the molecularphylogeny, i.e. is Phoebetriaa sister- comprehensivegenus Diomedea,we used mi- group to the remaining Diomedeidae, as ten- tochondrial DNA sequences to investigate uously surmisedby a handful of "primitive" 786 Nu ET^•. [Auk, Vol. 113

characters(see Murphy 1936) that are shared i0 mM •5-mercaptoethanol;i mM eachof dGTP,dATP, with other petrels?(2) Does the molecularev- dTTP, and dCTP; ! M of each primer; i0-i,000 ng of idence offer any support for morphologically completegenomic DNA; and 2.5 units of Taqpoly- definedgroups previously delimited within Di- merase(Thermus aquaticus DNA polymerase,Perkin- omedea(Coues 1866)? (3) What absolute evolu- Elmer-Cetus). The dsDNA products were visualized in a 2% NuSieve low-melting point agarosegel (FMC tionary rate calibrationis suggestedfor cyt-bin Bioproducts)containing 2 pg. ml • ethidium-bromide the Diomedeidae,and how doesthis compare (Maniatis et al. 1982). The dsDNA product was cut with other estimates of the molecular clock for directlyfrom the geland resuspendedin 300/xLwater this gene (Irwin et al. 1991,Martin et al. 1992)? by heating to 73øCfor 15 min. Further primer pairswere usedto amplify double- METHODS strandedDNA subfragmentsfrom the isolatedcyt-b gene (i.e. Li4863/H15104, Li4863/Hi5298, L14990/ Studyorganisms.--Samples of fresh blood or liver H15298, L15236/H15505, Li5311/H15710, L15656/ tissue were collected from the extant speciesof al- H16065); for original primer descriptionssee Helm- batrossesand two speciesof the Procellariidae. Most Bychowskiand Cracraft(1993) and referencesthere- blood samples were collected from chicks or incu- in. A large degree of fragment overlap, as well as the bating adults at nesting colonies to ensure known sequencingof both DNA strands,ensured accurate breeding provenance,although sampleswere taken data collection.An air thermocycler(Idaho Technol- from adult birds at for three of the species.All ogies) was used to perform 10-/xLamplifications of tissue sampleswere stored in 100% ethanol in the dsDNA in glassmicrocapillary tubes using standard field and transportedwithout freezing. The binomen, buffers described elsewhere (Wittwer et al. 1989, Wit- or current trinomen where appropriate,and collec- twer 1992). All subfragmentswere amplified with tion locality of each taxon in this study are listed in conditions:! secat 94øC,0 secat 48øC(i.e. a drop to the Appendix. Basedon simple morphologicalcom- 48øCwithout pausebefore climbing to extensiontem- parisons (Murphy 1936), our two outgroup species perature), 10 sec at 72øC,and 35 cycles at slope 9 from the Procellariidae (Southern Giant [Ma- (machine-specificfastest temperature ramping rate cronectesgiganteus ] and Gray Petrel [Procellariacinerea ]) available).Subfragments were visualized,isolated, and representmembers of the mostlikely sistergroup to resuspendedas described above. Concurrent negative the Diomedeidae. and positive controlswere performed for each ex- DNA isolation.--We extracted DNA suitable for en- periment. zymatic amplificationby boiling a minute piece of Single-strandedDNA for direct sequencingwas tissue(<5/xg) or suspendedblood (5/xL) for 15 min generatedusing i:100 dilutions of one primer in 50- in 500/xL of a 5%w/v Chelex-beadsuspension (Sing- /xLamplification reactions together with !/xL of the er-Sam et al. 1989, Walsh et al. 1991). After brief vor- resuspendedsubfragment of dsDNA (Gyllenstenand texing to break up the tissue,the beadswere pulse- Erlich 1988). Amplification reagents were the same centrifuged for a few seconds,and 300/xL of the su- as describedabove for initial gene isolation and were pernatantwere removedas a sourceof templateDNA. performedin a Peltlet-effectthermocycler with con- Mitochondrialcytochrome-b gene isolation and sequenc- ditions: i rain at 94øC, i rain at 52øC, 2 rain at 72øC, ing.--The cyt-bgene and shortflanking regionswere and 35 cycles.Products were concentratedand de- amplified and isolatedas a singlefragment using the saltedby spin-dialysis(Millipore 30,000NMWL) be- PCR primers L14863 5'-TTTGCCCTATCTATCCT- fore sequencingby the Sanger termination-dideoxy CAT-3' situated at the end of ND5 (numbered follow- method (Sanger et al. 1977) using Sequenase• 2.0 ing the mitochondrialgenome [Desjardins (U.S. Biochemical).Sequencing products were sub- and Morais1990]) and designedfrom a consensusof jectedto denaturinggel electrophoresisfollowed by Procellariidaeand Diomedeidae partial ND5 sequenc- autoradiography. es(unpubl. data),and H15915 in tRNA-threonine (Ed- Correcteddistance computation.-We computed codon wards and Wilson 1990). The human mitochondrial third-positioncorrected distances for comparisonwith genome (Anderson et al. 1981) numbered primer previouslyestimated values (Irwin et al. 1991,Thomas H15915 is equivalent to chicken mitochondrialnum- and Martin !993). Corrected distanceswere made us- ber H16065 (Desjardinsand Morais 1990).Of several ing DNADIST from the Phylip 3.5 set of programs experimentaltemperature-cycling parameters tried we (Felsenstein 1993), set to the ML substitution model found that a four-step cycle most efficiently and re- (Felsenstein 1981) with a 10:i explicit transition bias liably amplified this approximately1.2-kilobase frag- (a conservativeestimate for birds [Kocher et al. 1989]) ment in a Peltier-effectthermocycler (MJ Research): and empirical nucleotidefrequencies of the total an- viz. i rain at 94øC, i rain at 40øC, 1 rain at 60øC, and alyzed gene sequencesused in the distancecompu- 3 rain at 72ø(2,for 35 cycles.Amplifications were per- tation. formed in 50-/xLreaction volumes containing 67 mM Phylogeneticanalysis.--Phylogenetic relationships Tris-HC1(pH 8.8);6.7 mM MgC12;16.6 mM (NH,)2SO4; were estimatedusing maximum parsimonyand max- October1996] AlbatrossMolecular Phylogeny 787 imum likelihood methods with bootstrappingto as- within constrainedgroups using MacClade 3.1 (Mad- sesssupport for internal branches(Felsenstein 1985, dison and Maddison 1993). The resultant topologies Hillis and Bull 1993). The two methods have "basic were then comparedusing parsimony as well asmax- equiprobable" (i.e. symmetrical)assumptions of the imum likelihood methods (Kishino and Hasegawa evolutionary process(i.e. cladisticparsimony [Farris 1989). 1983]) versus a suite of a priori assumptionsimplicit in a model of the evolutionary processat the DNA RESULTS level (Felsenstein1981). Parameterization of a model of evolution at the DNA level may be consideredan Alignment of cyt-b sequences(1,143 bp) of appropriate approach in phylogeny building given the 14 albatrossspecies and two membersof the mountingevidence for a high transitionbias in the Procellariidae shows no nucleotide insertions avian mtDNA mutationalspectrum (Kocher et al. 1989, Edwards and Wilson 1990, E. Smith et aL 1992, Nunn or deletions among sequences(data available and Cracraft1996) and a generalappreciation of the from Genbank, inclusive accession numbers effectsof base-compositionalbias upon character-state U48940 to U48955).For a variety of reasonswe change,i.e. character-statebias (Collins et al. 1995). considerthe cyt-b gene sequencesreported to Maximum parsimony analysesand bootstrapping be solely mitochondrial in origin. The entire (100 replicates) were accomplishedwith PAUP 3.1.1 cyt-bgene and flanking regions initially were (Swofford1993). Ten replicateheuristic searches were isolated as a single, contiguous fragment. This performed with random addition of taxa to minimize procedure minimizes the potential amplifica- input bias.Branch-swapping was made by the tree bisection-reconnection(TBR) algorithm. Con- tion of smaller fragmentsthat are more likely temporaneouschanges were favored(i.e. parallelisms to be translocated into the nuclear genome over reversals) by using delayed transformation (Quinn 1992, M. Smith et al. 1992, Kornegay et (DELTRAN) optimization. These analyses included al. 1993).The genesequences can be fully trans- all charactersand substitutionsequally weighted. lated using the chicken mitochondrial code Maximum likelihood analysesand bootstrapping (Desjardinsand Morais 1990)without nonsense (100 replicates) were performed with the program or intervening stop codons.Finally, the align- fastDNAml 1.1.1 (Felsenstein 1981, Olsen et al. 1994). ment revealed neither a particular overabun- Heuristic searcheswere repeated with random ad- dance of first and second codon position dition of taxa. Overall empirical base frequenciesof changes,nor a shift in the typical avian mtDNA the cyt-bgene sequencesin this study,relative codon position evolutionary rates of 5:1:20 (lst:2nd:3rd), transition bias, phenomenaknown to occur in and a 10:1 transition (Ti) to transversion (Tv) substi- mtDNA sequencestranslocated to the nuclear tution bias, were defined a priorias parametersof the genome (Arctander 1995). nucleotide substitution model in the likelihood com- There are 333 variable nucleotide positions putations. amongthe 16 genesequences (29.1% of the total Following phylogeneticanalysis, we performeda cyt-bgene sequence). Variable nucleotides pre- parsimony-based"winning-sites" test (Templeton dominantly were at codon third positions(260 1983) to compare possible alternate arrangements sites,78.1%), followed by first positions(63 sites, among the major groupsestablished within the Di- 18.9%),and then secondpositions (10 sites,3.0%). omedeidae.Assuming a data set is potentially unin- The distribution of variable sites reflects the formative,a winning-sitestest can be usedto compare charactersupport among any number of opposing majority of substitutionsoccurring at synony- phylogenetichypotheses (i.e. topologies)with an ex- mous sites (codon third positionsand leucine pectation of stochasticallyequal support for each. codon first positions). When a comparisonof charactersupport for any two Patternsof pairwisesequence divergence and sat- particular topologiesdeparts from equality, a bino- uration.--Uncorrectedpairwise percentagedif- mial testdetermines if one is significantlybetter-sup- ference (Table 2) between species ported than the other, and is therefore more likely ranged from 0.87%(D. amsterdamensisvs. D. ex- to be the correcttopological arrangement (Templeton ulans) to 11.20%(D. irroratavs. D. chlororhynchos), 1983,Prager and Wilson 1988).This test is a conser- i.e. the largest difference occurswithin Diome- vative indicatorof significancelevels for tree support dea.Differences across the deepestnode, i.e. from (Felsenstein1988). Based on the constrainedtopolog- members of the Diomedeidae to the Procellar- ical outcome of the winning-sites test, we explored the range of rootedhypotheses for the Diomedeidae iidae, ranged from 12.60%(both Phoebetriaspe- by creatingthe five possiblerooted arrangementsin- cies vs. Macronectesgiganteus) to 15.84% (D. cludingall 16 taxa.The most-parsimonioustopologies epomophoraand D. irroratavs. Procellariacinerea), were determined by searching for rearrangements potentially indicating some rate variability 788 NUNN•r AL. [Auk,Vol. 113

among the albatrosslineages (œ = 14.31 + SD of 0.81%, n = 28). The two procellariids (Ma- cronectesvs. )differed by 11.37%,which was comparableto the largestdifference found within the Diomedeidae. Within the Diomedeidae, corrected codon third-positiondistances ranged from 2.85%(D. exulansvs. D. amsterdamensis)to 47.05% (D. ex- ulansvs. D. chrysostoma),again emphasizingthat the largestdifferences are within the traditional Diomedea(Table 2). From Diomedeidae to the Procellariidae, this computed value in- creasedsubstantially to a mean value of 117.91% (n = 28, SD = 11.64)and identifiedthe outgroup comparisonsat codon third positionshighly af- fected by multiple substitutions,i.e. within a zone of saturation. Pairwise empirical numbers of substitutions between the procellariiform cyt-b sequences partitioned into transitions (Ti) and transver- sions (Tv) revealed a consistent bias in favor of Ti changes(Table 3, Fig. 1). However, compu- tation of the ratio of Ti to Tv substitutions iden- tified a large range of variation in the relative contribution of each substitution to these comparisons.Ratios of Ti:Tv ranged from a low of 2.1:1 between distantly related taxa (both Phoebetriaspecies vs. Macronectesgiganteus), to a difference composed entirely of Ti substitu- tions(31:0) between closelyrelated taxa (D. chlo- rorhynchosvs. D. cauta).A comparisonof all nu- cleotide positionsshowed that the growth of Ti differencesbetween diverging taxadeparts from linearity at approximately 11% uncorrected to- tal pairwise difference (Fig. 1A). The cluster of points greater than 11%comprised all pairwise comparisonsof Procellariidae to Diomedeidae and indicated deeper comparisons have pro- gressedinto a zone of saturation. Dependent upon codon position, however, there is a differing pattern of divergencein the cyt-b gene sequences.The growth of Ti substi- tutions at codon first positionsincreases linear- ly betweenall taxain this study(Fig. lB) with- out evidenceof a drop or plateauin divergence occurringas more remotecomparisons are made. Codon secondpositions warrant no specialat- tention becausethey do not contributesignif- icantly to overall differences. In contrast to co- don first positions, the codon third position Ti divergenceincreases linearly to a limit of ap- proximately 25% total difference(Fig. 1C) and then plateausinto the zone of saturation.The plateau effect is most pronouncedat third po- October1996] AlbatrossMolecular Phylogeny 789

TAI•I•E3. Pairwisesubstitutional differences in cytochrome-bgenes among Diomedeidae and Procellariidae. Transitions(Ti) abovethe diagonal,and transversions(Tv) below the diagonal.

Species 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 D. chlororhynchos -- 31 27 27 31 99 96 106 113 102 115 117 83 87 105 116 2 D. bulleri 3 -- 24 28 16 97 94 98 104 98 109 113 79 79 112 115 3 D. chrysostorna 3 6 -- 18 24 98 94 100 103 94 109 11 79 81 109 120 4 D. rnelanophris 5 8 4 -- 27 92 90 98 103 94 108 109 77 81 109 111 5 D. cauta 0 3 3 5 -- 95 92 99 106 98 111 111 82 81 111 116 6 D. exulans 17 20 20 22 17 -- 8 33 66 61 69 64 85 95 111 119 7 D. arnsterdarnensis 19 22 22 24 19 2 -- 36 64 59 68 66 84 94 110 120 8 D. epomophora 14 17 17 19 14 3 5 -- 68 69 69 74 98 100 121 132 9 D. imrnutabilis 12 15 15 17 12 11 13 8 -- 19 49 40 100 103 117 129 10 D. nigripes 11 14 14 16 11 10 12 7 1 -- 46 39 95 97 118 121 11 D. irrorata 13 16 16 18 13 12 14 9 5 4 -- 47 108 112 124 135 12 D. albatrus 10 13 13 15 10 9 11 6 2 1 3 -- 106 108 118 132 13 P. palpebrata 6 9 9 11 6 15 17 12 10 9 11 8 -- 22 97 107 14 P. fusca 6 9 9 11 6 15 17 12 10 9 1! 8 2 -- 97 111 15 M. giganteus 47 48 48 48 47 50 50 49 47 46 48 45 47 47 -- 104 16 Pr. cinerea 47 50 48 50 47 50 50 49 47 46 46 45 47 47 26 -- sitions becauseof the preponderanceof syn- (followingchicken codon usage [Desjardins and onymouschanges that can occurat thesesites. Morais 1990]). In total, 46 (12.1%) of the 380 In view of the overall pattern of divergence amino acid sites were variable among the 16 among these sequences,we believe that third- translatedsequences. Variable residueslargely position comparisonsto the outgroup Procel- were confinedto the transmembraneregions of lariidae will almostcertainly exhibit someeffect the molecule.A majorityof aminoacid replace- of site saturation. Among the Diomedeidae, mentsinvolved exchangesbetween hydropho- however, comparisonsshould be largely unaf- bic residues--a replacementpattern similar to fected by saturationat any given site. that found in and other organisms Basecompositional bias.--The pattern of com- (Irwin et al. 1991,Degli Espostiet al. 1993). positionalbias (Prager and Wilson1988) at each Phylogeneticanalysis.--An identical branch- codon position in procellariiform cyt-b (Table ing topologywas best supported by both max- 4) is almost identical to that found in mammals imum parsimony(single most-parsimonious tree (Irwin et al. 1991)and other birds (Kornegayet L = 574 steps, CI [excluding uninformative al. 1993, Nunn and Cracraft 1996). First posi- characters]= 0.591) and maximum likelihood tions are little-biased (C = 0.088), being G-poor (LnL = -4103.65) analyses(phylogeny shown (• = 21.1 ___1.15%) and slightly C-rich (œ= 30.5 in Fig.2). Furtherparsimony analyses including + 0.64%).Second positions are more biasedthan a greaternumber of outgrouptaxa from the first (C = 0.219),again G-poor (• = 12.9 + 0.00%) Procellariidae,Hydrobatidae, and Pelecanoidi- but T-rich (• = 39.6 + 0.17%).The highest com- dae (a total of 85 taxa) resulted in no changein positionalbias is found at third positions(C = root location or branching pattern within the 0.428),which havevery low G (• = 3.8 ___0.77%) Diomedeidae (data not shown). The "phylo- and low T (• = 14.1 + 1.41%) content, and are genetictree" indicatedan initial bifurcationin rich in A (• = 37.4 ___1.13%) and C (•?= 44.7 + the Diomedeidae.The main lineageseach sub- 1.65%).The overall GC contentof eachsequence divided once more, resulting in four phyloge- (• = 0.468 ___0.009, range 0.450 [Procellariaci- neticgroups: (1) southernmollymawks, (2) sooty nerea] to 0.485 [Diomedeairrorata]) falls within albatrosses,(3) "great"albatrosses, and (4) North the range of other known cyt-bgenes and Pacificalbatrosses. The full descriptionfor this averagesslightly higher than available passer- tree is: ((((Diomedeachlororhynchos, ((D. bulleri, ine sequencesand lower than the MuscovyDuck D. cauta),(D. chrysostoma,D. melanophris,(Phoe- (Cairinamoschata) and most phasianids(sum- betriapalpebrata, P. fusca)),((D. epomophora,(D. marized in Jermiin et al. [1994]). amsterdamensis,D. exulans)),((D. immutabilis,D. Cytochrome-bprotein sequence variation.--There nigripes),(D. irrorata,D. albatrus)) )), (Macronectes was a low number of amino-acid replacements giganteus,Procellaria cinerea )). among the translatedcyt-b protein sequences Maximum parsimony (MP) and maximum 79O NUNN ETAL. [Auk, Vol. 113

150 A

Transitions 125 Transversions lOO

75

50

25

o o% 6% 8% 10% 12% 14% 16% 18%

40 B c

30-

20 •

10,

.'... ::...__, 0q0 2q0 4•0 6% 8 10% 12% 0% 10% 20% 30% 40%

FzC.1. Empiricalnumbers of transition(Ti) and transversion(Tv) substitutions(on y axis)plotted against total uncorrectedpairwise percentagedifference (on x axis) for (A) all codonpositions (i.e. 1,143sites), (B) codonfirst positions(i.e. 381 sites),and (C) codonthird positions(i.e. 381 sites). likelihood (ML) bootstrap analyses identified letic groups within the Diomedeidae all were broadly concordant levels of support for highly supported:southern mollymawks (100% brancheswithin the phylogeny. Bootstrapsup- MP, 100%ML; five taxa),sooty albatrosses (100% port for the first lineage,comprised of southern MP, 100%ML; two taxa),great albatrosses(100% mollymawksand sooty albatrosses,was rela- MP, 100% ML; three taxa), and North Pacific tively low (79% MP, 52% ML) comparedwith albatrosses (99% MP, 99% ML; four taxa). valuesfor other deep branches.Complete boot- Further branching events within the three strapsupport (i.e. from all replicates)was found groups with more than two members (i.e. ex- for monophyly of the secondlineage, contain- cluding sooty albatrosses)were robustly sup- ing "great" and North Pacificalbatrosses (100% ported.Within southernmollymawks, the para- MP, 100%ML). In addition, the four monophy- phyletic branching of D. chlororhynchoswas October1996] AlbatrossMolecular Phylogeny 791 792 Nuuu ETAL. [Auk,VoL 113

tganteus Procellariidae

Diomedea exulans Great albatrosses

•19• Diomedeaimmutabilis

North Pacific albatrosses Diomedeanigtipes (Phoebastria )

SootyAlbatrosses palpebrata I.

SouthernMollymawks risi (Thalassarche) FIG. 2. Phylogeneticrelationships among the Diomedeidaebased on maximum parsimony (MP) and maximumlikelihood (ML) analysesof cytochrome-bgene sequences (rooted to the outgroupProcellariidae). Identicalbranching patterns were determinedby both analyses(most parsimonious tree L = 574,CI [excluding uninformative characters]= 0.591; maximum likelihood tree LnL = -4103.65). Percentagebootstrap support found in phylogeneticanalyses are shown to the left of internal branches(MP/ML). marginally supported (65% MP, 61% ML) as oc- four high bootstrap-supportedgroups shown in curring before the origin of the remaining four Figure 2. Assuming monophyly of the four speciesin this group. Among the remaining groupsand including all 14 albatrosssequences, four taxa, a sister-taxarelationship of D. bulleri 120 possiblecombinations of taxa exist for this and D. cautawas highly supported (96% MP, test. We extracted from the data set and tested 93% ML), and D. chrysostomaand D. melanophris separatelyeach possiblecombination of four also formed sistertaxa supportedby relatively taxa.Character support was assessedfor the three high bootstrapvalues (73% MP, 87% ML). In- topologicalarrangements possible among each spection of replicates revealed minor support four-taxon combination (i.e. we computed 360 for the branching of D. chlororhynchosbetween separatevalues; results summarized in Fig. 3A). the two species-pairsdescribed above, i.e. res- For all 120 combinations of taxa, a monophy- olution at the base of the three lineages(chlo- letic origin of sooty albatrossesand southern rorhynchos,cauta /bulleri, chrysostoma/melano- mollymawks(i.e. TopologyI; Fig. 3A) was best phris)could be consideredproblematic based on supported in comparisonto the two alternative the current dataset. Similarly, amongthe North arrangements (i.e. Topologies II and III; Fig. Pacific albatrossesmost replicates supported 3A). Four combinations of sampled taxa tied monophyly of D. irrorataand D. albatrus(60% equally highest supportwith a ratio of 42 char- MP, 69% ML), although the remaining repli- acters (38 or 39 transitions and 3 or 4 transver- catessupported a paraphyletic branching order sions) supporting the Topology I arrangement for these taxa (D. irrorata basal) before the well- to 7 characters(only transitions) supporting a supportedmonophyletic apical origin of D. im- best alternative arrangement (three combina- mutabilisand D. nigripes(96% MP, 94% ML). tions for TopologyII and one combinationfor To further assessthe higher-level relation- Topology III). The charactersupport for these ships determined from our phylogenetictree four identical higher-level topologies when we testedcharacter support for an unrootednet- compared with the nearest scoring alternative work of one taxon sampled from each of the topology (i.e. 42:7) was highly significant (P < October 1996] AlbatrossMolecular Phylogeny 793

A TopologyI TopologyII Topologym

southern great southern great southern sooty mollyma•trossmollymawkalbatrossmollymaw•batross sooty NorthPacific NorthPacific sooty great NorthPacific albatross• albatrossalbatross • albatross Rangeof total albatross• albatross charactersupport 3242 4-11 4-10

Characterpartition transitions 29-39 4-11 4-10 transversions 24 0 0

B

Tree 2 Tree 3 Best tree Length=578steps (+4) Length=579steps (+5) Length=574steps LnL=-4106.65 Ln -L=-4111.48 LnL=-4103.65 (difference=-2.99,sd=8.33) (difference=-7.83,sd=6.98) • Procellariidae• Procellariidae• Procellariidae southernmollymawk southernmollymawk sootyalbatross sootyalbatross NorthPacific albatross NorthPacific albatross greatalbatross sootyalbatross sgrouea•ealmbam•o•i;maw k

Tree 4 Tree 5 Length=592steps (+18) Length=592steps (+18) LnL=-4129.61 LnL=-4129.61 (difference=-25.96,sd=9.42) (difference=-25.96, sd=9.42) • Procellariidae• Procellariidae greatalbatross NorthPacificalbatross sootyalbatross southernmollymawk sootyalbatross S;oU•thi•ac7•Y2aimks s greatalbatross Fig. 3. (A) Four-taxontest of charactersupport for the internalbranch of topologicalarrangements among sampledtaxa of the higher-levelgroups and (B) the five possiblerooted arrangements of Fig. 3A. Topology I. Most-parsimonioustrees include all 16taxa and were determined based on the higher-levelgroup constraints shown(i.e. rearrangementswere allowedonly within thesegroups). Exact tree descriptionsare: Tree 1, best tree as describedin Results;Tree 2, (((((Diomedeaepomophora, (D. amsterdamensis,D. exulans)), (((D. immutabilis, D. nigripes),D. albatrus),D. irrorata)),(Phoebetria palpebrata, P. fusca)),(D. chlororhynchos,((D. bulleri,D. cauta), (D. chrysostoma,D. melanophris)))),(Macronectes giganteus, Procellaria cinerea)); Tree 3, ((((D. chlororhynchos,((D. bulleri,D. cauta),(D. chrysostoma,D. melanophris, ((D. epomophora,(D. amsterdamensis,D. exulans)), (((D. immutabilis, D. nigripes),D. albatrus),D. irrorata))),(P. palpebrata,P. fusca)),(Macronectes giganteus, Procellaria cinerea)); Tree 4, (((((P. palpebrata,P. fusca),(D. chlororhynchos,((D. bulleri,D. cauta),(D. chrysostoma,D. melanophris)))),((D. immutabilis,D. nigripes),(D. albatrus,D. irrorata))),(D. epomophora,(D. amsterdamensis,D. exulans))), (Macronectes giganteus,Procellaria cinerea)); and Tree 5, (((((P.palpebrata, P. fusca),(D. chlororhynchos,((D. bulleri,D. cauta),(D. chrysostoma,D. melanophris)))),(D. epomophora,(D. amsterdamensis,D. exulans))), ((D. immutabilis,D. nigripes),(D. albatrus,D. irrorata))),(Macronectes giganteus, Procellaria cinerea)). 794 NUNNET AL. [Auk,Vol. 113

0.001)based on a binomial test(Templeton 1983). . 1 but with trivial apical rearrangementsoccur- The four-taxon higher-level topology was con- ring within the higher-level groups). gruent in branching pattern with the phylo- genetic tree derived from analysesof the com- DISCUSSION plete data set. We estimatedthe root locationto TopologyI Generaof albatrosses.--Thestatus and number by creating complete 16-taxa trees constrained of genera in the Diomedeidae (Table 1) have to this topology. A branch to the Procellariidae varied widely since the formal descriptionof outgroup was attached to the five possible the first known taxon Diomedeaexulans (Lin- branch positionsof Topology I and parsimony naeus 1758). Nearly a century elapsed before and likelihood support computed for the dif- Reichenbach (1852) introduced three additional ferent higher-level group arrangements (Fig. genera (i.e. Phoebetria,Phoebastria, and Thalas- 3B). As expected,our phylogenetictree (Tree 1; sarche),assigning a member of Diomedeainto L = 574 steps) is the most parsimonious and eachof the four. In a taxonomicsynopsis of the most likely rooted hypothesisof relationships petrelsthat followed Reichenbach'swork, Coues among these birds (Fig. 3B). Basedon a parsi- (1866) presented unique morphological char- mony criterion, the four alternative trees (i.e. acters that establishedmonophyly of known Tree 2 to Tree 5) are rejected as hypothesesof speciesof Diomedeidaeamong the procellari- relationshipsbecause they are less parsimoni- iforms. Further, he (1866:187-188)developed a ous (L = 578-592 steps). Of the four rejected hierarchicallydefined classification based on the trees, numbers4 and 5 are 18 stepslonger (L = presenceor absenceof well-describedmorpho- 592 steps) and also are significantly rejected logical charactersamong albatrosses.Although based on log-likelihood comparisons(i.e. the Couespresented evidence for "morphological- log-likelihood difference is outside a 95% con- groups" within Diomedea,he did not name them fidence interval compared with the maximum formally at that time. In fact, Couesarbitrarily likelihood tree [Kishino and Hasegawa 1989]). rejected two of Reichenbach'sgenera (Phoe- Although rejectedby a parsimonycriterion, the bastriaand Thalassarche)and adopted a more two remaining trees (2 and 3), at four stepslon- conservativearrangement, admitting only Dio- ger (L = 578), are not significantlyrejected based medea(southern mollymawks, North Pacifical- on log-likelihood differences.We note, how- batrosses,and "great" albatrosses)and Phoe- ever, that the inclusionof many invariant sites-- betria(for the single sooty albatrossknown at in the caseof mtDNA primarily the first and that time; see Table 1). secondcodon positions--createsa considerable The discovery of new albatrosstaxa, partic- inflation of the estimate of likelihood variance. ularly from the southern oceansand the Aus- This inflation in turn affects the potential re- tralia/New Zealand region, led to an increased jection of alternative hypothesesby unrealist- interestin the taxonomyof the group,and sev- ically widening the confidenceinterval based eral genera either were reintroducedor created on the computed standard error (e.g. see the (seeTable 1). Generic-levelrevisions finally cul- discussionof codon first and second position minated in Mathews' (1934) treatment of the charactersgrouped together as "class-2sites" family Diomedeidae, which admitted all eight by Hasegawaand Kishino [1989]).Interestingly, genera that had been describedpreviously (Ta- however, Tree 2, which postulates southern ble 1). Subsequently,however, Mathews and mollymawks as the sister-groupto remaining Hallstrom (1943) subsumedthe monotypic ge- albatrosses,is more likely (LnL = -4105.15) nus Rhothoniaback into Diomedea,reuniting the than Tree 3, which postulatesthe sooty alba- morphologically similar "great" albatrosses.In trossesas the sister-group (LnL = -4111.48). addition, Mathews and Hallstrom elected a new Further tree-searching determined the exis- monotypic genusJulietata for the unique equa- tence of an additional 10 trees (not including torial speciesPhoebastria irrorata based on mor- Tree 1, the most parsimonioustree) that were phologicaldifferences from other North Pacific more parsimoniousthan Tree 2 and Tree 3 (i.e. albatrosses(albatrus, immutabilis, nigripes), which in the range L = 575,577steps; LnL = -4104.81, again resulted in a total of eight genera for the -4116.65). These 10 trees were congruent with 18 known taxa. Shortly thereafter, in a wave of the phylogenetictree in terms of their higher- lumping by avian taxonomists(including G. M. level branching pattern (i.e. a topology of Tree Mathews), the "generically oversplit" taxono- October1996] AlbatrossMolecular Phylogeny . 795 mies of Diomedeidae (Mathews 1934, Mathews logenetichypothesis supports the notion of an and Hallstrom 1943) were abandoned in favor early basaldichotomy among members of the of a single, all-encompassinggenus Diomedea Diomedeidae. This initial dichotomy led to a (Mathews 1948). lineage limited in distributionto the southern In responseto the chaoticand in many cases oceanscomprised of two groups,southern mol- confusing changesthat Mathews' introduced lymawks and sootyalbatrosses, versus a more into the ornithological literature (e.g. seeMur- geographicallywidespread lineage also with two phy 1945,Serventy 1950),a standardizationof groups,the North Pacific albatrossesand the procellariiform arose (Alexander et "great" albatrossesfrom the southern oceans. al. 1965). The taxonomic revision of Alexander The molecularphylogeny provides for the first et al. (1965) essentially was an "agreed state- time evidenceof monophylyof sootyalbatross- ment" among a committee of leading esand southernmollymawks and indicatesthat taxonomiststo rejectthe plethora of superfluous the traditional genusDiomedea is paraphyletic. naming in the "Mathews" classificationsin fa- The current taxonomyof albatrosses,traceable vor of prior taxonomictreatments of the order to Coues (1866), provides an illustrative avian Procellariiformes(e.g. Alexander 1928, Peters exampleof an "either-or"type of classification. 1931,Murphy 1936).Although Mathews' work Historically,the genusDiomedea received all al- certainly sufferedfrom nomenclaturalzealous- batrosstaxa that were not Phoebetria,regardless ness, numerous higher-level procellariiform of the clearmorphological affinities that united groupswere returned to less-sophisticatedtax- groups within Diomedea.Indeed, Coues pre- onomic treatments despite being founded on sented the morphological evidence for these well-defined morphological features. In the groups but refrained from providing formal processof circumventing Mathews' "inconsis- names.Unfortunately, the later and more so- tent new classifications,"Alexander et al. (1965) phisticatedtaxonomies of albatrossesthat rec- returned the generaof albatrossesto the earlier ognized these affinities and provided named classificationthat essentiallyhad been devised groups(e.g. Mathews 1934,Mathews and Hall- by Coues (1866) and in which Reichenbach's strom 1943)were rejectedduring the taxonomic genera Thalassarcheand Phoebastriahad been "cleansing"of the Procellariiformes(Alexander subsumedinto Diomedea.Since the revision by et al. 1965). Alexander et al. (1965), little research has oc- It is interesting to note that the molecular curred on albatrossevolutionary relationships, phylogeny is concordantwith monophyly of and the generic designationsadvocated have Coues'morphologically defined groupswithin persisted in the literature (e.g. Jouanin and the paraphyletic traditional genus Diomedea. Mougin 1979, Sibley and Monroe 1990). Sug- Based on discrete characters of the bill and sim- gestions for defining "subgeneric" groups ple diagnosticallometric measurements,Coues within the comprehensivegenus Diomedea have proposed two natural groups within the Di- resurfaced, however (e.g. Wolters 1975, War- omedea(i.e. with the prior exclusionof Phoebe- ham 1990). tria). Coues'"Genus I. Diomedea:Group A" (i.e. A phylogeneticclassification of the Diomedei- D. exulans,D. albatrus,D. irrorata [the first de- dae.--Our resultsfrom phylogeneticanalysis of scription of a partial cranium of this speciesand cyt-b gene sequencesrevealed very clear evi- correctlyassigned to the group],and D. nigripes) denceof four higher-level species-groupsin the is concordant with a primary lineage in our Diomedeidae (see Fig. 2). These phylogenetic phylogeny. This group, some of which were species-groupsare congruentwith sometradi- later transferredto the resurrectedgenus Phoe- tional morphologicallydefined groups of al- bastria(albatrus, immutabilis, nigripes, and irrorata; batrosses(i.e. current or discardedgenera, or Mathews 1934), containsthe largestalbatrosses unnamed morphological groups).However, the and can be characterizedby featuresincluding traditional taxonomic framework for albatross a relatively short tail, laterally broadened bill relationships, i.e. sooty albatrossesPhoebetria (particularlya wide boss-likebroadening of the versusall other Diomedea(e.g. Coues 1866,Mur- culminicorn posterior of the nostrils), as well phy 1936, Alexander et al. 1965) is not sup- as large, wide nostrils.The later division of taxa ported based on most-parsimoniousrooting of to Phoebastriaalso is congruent with the cyt-b the arrangement among the well supported phylogeny.Coues' second major group "Genus higher-level groups(see Fig. 2). Our cyt-bphy- Diomedea:Group B" (i.e.D. melanophris,D. cauta, 796 NUNNœT AL. [Auk,Vol. 113

TABLE5. A phylogenetic classificationof the Dio- The evolutionary relationships among alba- medeidae. trossesinferred from their traditionaltaxonomy presentsa subjectivehypothesis of groupings. Family Diomedeidae We therefore recommend a formal revision of Genus Thalassarche Species T. chlororhynchos the taxonomyof Diomedeidaeto achievea clas- T. bulleri sificationcongruent with the new phylogenetic T. cauta hypothesisof relationships.Our revision con- T. chrysostoma stitutesfour genera of coordinatephylogenetic T. melanophris Genus Phoebetria rank, eachequivalent to one of the higher-level Species P. palpebrata phylogeneticgroups, and eliminatesparaphyly I•. fusca of the traditional genus Diomedea.We resurrect Genus Diomedea two previously described genera: (1) Thalas- Species D. exulans D. amsterdamensis sarcheReichenbach 1852 (type speciesmelano- D. epomophora phrisTemminck 1828), by original designation, Genus Phoebastria to include the southern mollymawks; and (2) Species P. immutabil• PhoebastriaReichenbach 1852 (type species P. nigripes P. &rorata brachyuraTemminck 1829 [synonym albatrus P. albatrus Pallas1769]), by original designation,to include the North Pacificalbatrosses. These genera have D. chlororhynchos,and D. chrysostoma)is the sec- historical precedenceover later synonyms (see ond clade within the other primary cyt-b lin- Table 1). Thus, a total of seven traditional spe- eage.This group is now more commonlyknown cies-level taxa are transferred from the genus as the southern mollymawks (including D. bul- Diomedeato the resurrectedgenera. This clas- leri, discovered after this date) and can be char- sification leaves the "great" albatrossesas the acterizedin comparisonwith Group A by a lat- sole members of the genus DiomedeaLinnaeus erally compressedand much weaker bill with 1758 and retains the genus PhoebetriaReichen- narrow culminicorn,and a relatively longer and bach 1852 for the sooty albatrosses.The adop- slightly rounded tail. Interestingly, Phoebetria tion of phylogenetically defined higher-level shares some of these morphological features species-groupsrefines the nomenclature of the with the southern mollymawks (Thalassarche), Diomedeidae and establishesfour comparable providing useful corroborative evidence for biological units within the family. The recom- their monophyly. For example, even Coues mended traditional species-leveltaxa within the (1866) remarked on the unclear distinction in genera are given in Table 5. bill morphology between Thalassarcheand Phoe- Albatrossnest building.--The evolution of be- betria,pointing out similar generalfeatures such havioral and life-history (BLH) charactersin as extreme overall lateral compressionand the birds has been shown to map closely to phy- acute narrowing of the culminicorn posterior logeneticrelationships (Prum 1990,Winkler and to the nostrils. Indeed, Coues considered the Sheldon 1993, Paterson et al. 1995). Members bill of Phoebetria"hardly separable"from that of Diomedeidae possessa diversity of stereo- of some members of Diomedea(i.e. Thalassarche) typed courtship and specialized nest-building but eventually relied upon additional "features behaviorsthat are well described(e.g. Marchant radically distinct from...those presented by and Higgins 1990).The inspectionof a complex Diomedeaproper" to distinguishat a generic nest-building character among the albatrosses level the singleknown speciesof sootyalba- in our study provides a valuable example of the tross (Phoebetriafuliginosa [palpebrata]) estab- benefits gained from a comparative phyloge- lished at that time. It seems clear that the sup- netic approach to analysis of BLH characters posedlyprimitive morphological features that (Brooks and McLennan 1991). Within the Di- Couesused to define Reichenbach'sgenus Phoe- omedeidae,only sootyalbatrosses and southern betria(i.e. completefuliginous plumage, pres- mollymawks build a tall pedestal-shapednest ence of a sulcus in the lower mandible, and made primarily from earth but with occasional acuminateelongation of the rectrices)are in fact rock and plant material included. Birds return plesiomorphicin origin,as they canbe found eachbreeding seasonto the previousyear's ped- in various combinationsin several other petrel estal-nest,which is both repaired and increased lineages. in size, and from which both members of the October1996] AlbatrossMolecular Phylogeny 797 pair perform elaboratecourtship and territorial (Warheit 1992), indicate displays. This contrasts with the low nest the existenceof a diversealbatross fauna during mounds of gathered vegetation of the "great" the lateTertiary period. A stratigraphicallydated albatrossesand the scantily lined nest scrapes fossil from suggeststhe presenceof of North Pacific albatrosses, which in both cases an albatross that shared affinities with D. alba- are rebuilt eachbreeding season (Warham 1990). trus (= Phoebastria)in the North Pacific Ocean Given that no other petrel builds a nest ped- as early as the mid-, approximately 15 estal,the traditional taxonomicarrangement of million years before present (MYBP; Miller albatrosseswould lead us to expect either that 1962).Relatively few procellariiformshave been an ancestorof all albatrossesbuilt a pedestal discoveredfrom Tertiary depositsbordering the nest and subsequentlywas lost from a lineage southern oceans (Olson 1985a,b), in contrast to within the genus Diomedea,or that both Phoe- the abundanceof Northern Hemisphere . betriaand a lineage within Diomedeaindepen- However, the discovery of a cranial fragment dently gained the pedestal nest-building be- of a southern of surprisingly mod- havior from an ancestor that did not build a ern appearancein Australia indicatesthe pres- pedestal nest. However, it seemsmore likely ence of a member of the genus Thalassarcheat that the complexand stereotypednest-building the Miocene-Plioceneboundary about 10 MYBP behaviorarose only once.Our cyt-bphylogeny (Wilkinson 1969). resolvesa single, i.e. monophyletic,origin of Previous assessment of corrected mitochon- this behavioral character in an ancestor of Phoe- drial cyt-b codon third-position divergencein betria and Thalassarchethat has persisted homeothermssuggested an evolutionaryrate of throughout this majoralbatross lineage. Similar approximately 10%per million yearsin a num- to some BLH charactersanalyzed by Paterson ber of mammalian lineages(Irwin et al. 1991); et al. (1995),it appearsthat building a pedestal this value has been corroboratedby further nest has remained stable over a considerable studiesof ground squirrels(Thomas and Martin evolutionary period. It is likely that the evo- 1993). On the basis of the fossil evidence out- lution of a laterally compressedbill morphol- lined above,we wereable to calibratecyt-b third- ogy also arose in the ancestorof Thalassarche positionrate estimatesfor the phylogenetically and Phoebetria.The bill morphologyis of par- defined higher-level groups Phoebastriaand ticular importance becausenest building is ac- Thalassarche,given well-supported relation- complishedby a lateral plasteringaction of the ships to their respective sister groups, i.e. Di- bill while the bird rotateson top of the pedestal omedeaand Phoebetria.The correctedthird-po- nest (G. Nunn pets. obs.).Further comparative sition pairwise divergence between members analysesof other such charactersamong alba- of (1) Phoebastriaand Diomedea(œ = 23.62 + trossesmay provide valuable insight into the 1.21%,n = 12) and (2) Thalassarcheand Phoebetria patterns of evolution of complex BLH charac- (œ = 28.58 + 0.88%, n = 10) leads to rate cali- ters in (e.g. see Patersonet al. 1995). brations of 1.58%and 2.86% per million years, Calibrationof evolutionaryrate.--Considerable respectively. These independently derived es- evidencesuggests that molecularevolutionary timates are not perfectly concordant with one rates vary among taxonomic lineages (Britten another but are lower than estimates derived in 1986,Li et al. 1987).In particular,the assump- mammals (Irwin et al. 1991). The inclusion of tion of a molecularclock is unlikely to extend a wider array of mammalianlineages suggests throughout a phylogeny whose members ex- that rates may be substantiallyslower in those hibit diverse reproductive or metabolic rates, of larger body size (i.e. lower basal metabolic the latter being inverselycorrelated with body rate) such as the baleen (Martin and size (Martin et al. 1992, Martin and Palumbi Palumbi1993). Low metabolicand reproductive 1993). Within the Diomedeidae, however, both rates of albatrossesmay be causal factors ex- basal metabolic rate (Adams and Brown 1984) plaining the low cyt-brate calibrations, but these and age at first breeding (Jouventin and Wei- observations are not idiosyncratic. Rates of merskirch 1988) vary by much lessthan an or- mtDNA evolution establishedfrom the split be- der of magnitude, and evolutionary-ratecali- tween the anseriform genera Anser and Branta brations among the now well-establishedphy- (Shieldsand Wilson 1987)suggest that, in com- logenetic lineages are possible. parisonwith their body size,geese have a slower Fossilremains of seabirds,mostly from the mtDNA substitution rate than other homeo- 798 NUNNItr AL. [Auk, Vol. 113 therms (Martin and Palumbi 1993). Similarly, MURPHY, F. SALOMONSEN,K. H. Voous, G. E. comparative restriction fragment analyses of WATSON, W. R. P. BOURNE,C. A. FLEMING, N. H. mtDNA evolutionary rates among a selection KURODA, M. K. ROWAN, D. L. SERVENTY,W. L. N. of passefineand nonpasserinegroups also sup- TICKELL,J. WA•H•M, AND J. M. WIN•-RSOTTOM. port a rate slow-down in comparisonwith non- 1965. The families and genera of petrels and their names. Ibis 107:401-405. avian vertebrate groups (Kessler and Avise ANDERSON, S., A. T. BANKIER, B. O. BARRELL,M. H. L. 1985). The surprisingly low cyt-b calibrations DE BRUIIN,A. R. COULSON,J. DROUIN,I. C. EPERON, obtained here support the hypothesisthat avian D. P. NIERLICH, B. A. ROE, F. SANGER, P. H. mitochondrial genomic evolutionary rates are SCHREIER,A. J. H. SMrrH, R. STADEN,AND I. G. considerablyslower in birds than in mammals. YOUNG. 1981. Sequenceand organisationof the human mitochondrialgenome. Nature 290:457- 465. ACKNOWLEDGMENTS ARCTANDER,P. 1995. Comparisonof a mitochondrial Financialsupport for this researchcame from a Frank gene and a correspondingnuclear pseudogene. M. Chapman PostdoctoralFellowship to G. B. Nunn. Proceedingsof the Royal Societyof London Se- Principal tissuecollections were made under the aus- ries B 262:13-19. pices of the South African Department of Environ- BAIRD, S. F., T. M. BREWER,AND R. RIDGWAY. 1884. mental Affairs and Tourism, New Zealand Depart- The water birds of North America, vol. 2. Little, ment of Conservation,Australian Department of En- Brown and Company, . vironment ( Division), and the Galapagos BLECHSCHMIDT,K., H.-U. Pm'ER,J. DE KORTE,M. WINK, National Park; we thank each of these agenciesfor I. SERoOLD,AND A. J. HELBIG. 1993. Untersu- granting permission to conduct collecting and for chungen zur loekularen Systematikder Raub- providing logisticalsupport. We thank P. Arctander toowen (Stercorariidae).Zoologische Jahrbuech- (University of Copenhagen,Denmark) for providing er Abteilung fuer Systematik Oekologie und tissue-collectingequipment, and the staff of the Ga- Geographieder Tiere 120:379-387. lapagosResearch Station for providing expert logis- BOURNE,W. R. P. 1989. The evolution, classification tical support in the GalapagosIslands. In addition, and nomenclature of the great albatrosses.Ger- we thank H. Hasegawa,N. Brothers,the Burke Mu- faut 79:105-116. seum,and the LouisianaState University Museum of BaITTEN,R.J. 1986. Rates of DNA sequenceevolu- Natural Sciencefor the generousdonations and loans tion differ between taxonomic groups. Science of tissuesamples for this study. G. Olsen provided 231:1393-1398. valuable assistanceon implementation of the com- BROOKS,D. R., ANDD. A. MCLENNAN.1991. Phylog- puter program fastDNAml. We gratefully acknowl- eny, ecology,and behavior:A researchprogram edge G. Barrowdough,J. Bates,W. R. P. Bourne,R. in comparativebiology. University of Chicago Brooke, J. Cracraft, C. Griffiths, J. Groth, S. Hackett, Press,Chicago. N. Klein, S. Stanley, and W. L. N. Tickell for their COLLINS,T. M., P. H. WIMBERGER,AND G. J.P. NAYLOR. commentsand suggestionson this manuscript and 1995. Compositionalbias, character-state bias, and for discussion of saturation rates in mtDNA. A. Baker, character-statereconstruction using parsimony. C. Krajewski, and an anonymousreviewer also pro- SystematicBiology 43:482-496. vided thoughtful commentson the manuscript.The COUES,E. 1866. Critical review of the Family Pro- researchreported in this paper is a contributionfrom cellariidae, part V, embracing the Diomedeinae the Lewis B. and Dorothy Cullman ResearchFacility and the Halodrominae.Proceedings of the Acad- at the AmericanMuseum of Natural History and has emy of Natural Sciencesof Philadelphia 18:172- receivedgenerous support from the LewisB. and Do- 197. rothy Cullman Program for Molecular Systematics DEGLI ESPOSTI,M., S. DE ValEs, M. CaiMI, A. GHELLI, Studies,a joint initiative of The New York Botanical T. PATARNELLO,AND A. MEYER. 1993. Mitochon- Garden and The American Museum of Natural His- drial cytochromeb: Evolution and structure of tory. the protein. Biochimicaet BiophysicaActa 1143: 243-271.

LITERATURE CITED DESIARDINS,P., ANDR. MORAIS. 1990. Sequenceand gene organisationof the chicken mitochondrial ADAMS,N.J., AND C. R. BROWN. 1984. Metabolic rates genome: A novel gene order in higher verte- of sub-Antarctic Procellariiformes:A compara- brates.Journal of MolecularBiology 212:599-634. tive study.Comparative Biochemistry and Phys- EDWARDS,S. V., AND A. C. WILSON. 1990. Phyloge- iology 77A:169-173. netically informative length polymorphismand ALEXANDER,W.B. 1928. Birds of the ocean. Putnam, sequencevariability in mitochondrial DNA of New York. Australiansongbirds (Pomatostomus). 126: ALEXANDER,W. B., R. A. FALLA, C. JOUANIN,R. C. 695-711. October1996] AlbatrossMolecular Phylogeny 799

FARRIS,J.S. 1983. The logical basisof phylogenetic 1986. National Museum of Natural Science, Ot- analysis.Pages 7-36 in Advancesin cladistics(N. tawa. I. Platnick and V. A. Funk, Eds.). Columbia Uni- KESSLER,L. G., ANDJ. C. AvISE. 1985. A comparative versity Press,New York. description of mitochondrial DNA differentia- FELSENSTEIN,J. 1981. Evolutionary trees from DNA tion in selectedavian and other vertebrate gen- sequences:A maximum likelihood approach. era. Molecular Biology and Evolution 2:109-125. Journal of Molecular Evolution 17:368-376. KISHINO, H., AND M. HASEGAWA. 1989. Evaluation of FELSENSTEIN,J. 1985. Confidence limits on phylog- the maximum likelihood estimate of the evolu- enies: An approachusing the bootstrap.Evolu- tionary tree topologiesfrom DNA sequencedata, tion 39:783-791. and the branchingorder in Hominoidea. Journal FELSENSTEIN,J. 1988. Phylogenies from molecular of Molecular Evolution 29:170-179. sequences:Inference and reliability. Annual Re- KOCHER, T. D., W. K. THOMAS, A. MEYER, S. V. ED- view of Genetics 22:521-565. WARDS,S. PAABO, F. X. VILLABLANCA,AND A. C. FELSENSTEIN,J. 1993. PHYLIP (Phylogeny Inference WILSON.1989. DynamicsofmitochondrialDNA Package)3.55c. Distributed by the author. Uni- evolution in mammals: Amplification and se- versity of Washington,Seattle. quencing with conservedprimers. Proceedings GYLLENSTEN,U. B., AND H. A. ERLICH. 1988. Gener- of the National Academy of SciencesUSA 86: ation of single-strandedDNA by the polymerase 6196-6200. chain reaction and its application to direct se- KORNEGAY,J. R., T. D. KOCHER,L. A. WILLIAMS,AND quencing of the HLA-DQA locus.Proceedings of A. C. WILSON. 1993. Pathwaysof lysozymeevo- the National Academyof SciencesUSA 85:7652- lution inferred from the sequencesof cyto- 7656. chrome b in birds. Journal of Molecular Evolution HARRIS,M.P. 1973. The biology of the Waved Al- 37:367-379. batross Diomedea irrorata of Hood Island, Gala- KRAIEWSKI,C., ANDJ. W. FETZNER.1994. Phylogeny pagos.Ibis 115:483-510. of cranes (: Gruidae) basedon cyto- HARmSON,P. 1983. Seabirds:An identificationguide. chrome-bDNA sequences.Auk 111:351-365. Houghton Mifflin, Boston. LANYON,S. M., ANDJ. G. HALL. 1994. Reexamination HASEGAWA,M., AND H. KISHINO. 1989. Confidence of barbetmonophyly using mitochondrial-DNA limits on the maximum-likelihood estimate of the sequencedata. Auk 111:389-397. hominoid tree from mitochondrial-DNA se- LI, W.-H., M. TANIMURA, AND P.M. SH,•RP. 1987. An quences.Evolution 43:672-677. evaluationof the molecularclock hypothesisus- HELM-BYCHOWSKI,K., AND J. CRACRa•FT.1993. Re- ing mammalian DNA sequences.Journal of Mo- covering phylogenetic signal from DNA se- lecular Evolution 25:330-342. quences:Relationships within the cotvine assem- LINNAEUS,C. 1758. Systemanaturae per regna tria blage (ClassAves) as inferred from completese- naturae, secundum classes,ordines, genera, spe- quencesof the mitochondrial DNA cytochrome- cies cum characteribus,differentils, synonymis, b gene. MolecularBiology and Evolution 10:1196- locis. Laurentii Salvii, Stockholm. 1214. LYDEKKER,R. 1891a. Catalogueof the fossil birds in HILLIS,D. M., ANDJ. J. BULL. 1993. An empirical test the . British Museum of Natural of bootstrappingas a method for assessingcon- History, London. fidence in phylogenetic analysis. SystematicBi- LYDEKKER,R. 1891b. On British fossil birds. Ibis Se- ology 42:182-192. ries 6:394-395. IRWIN, D. M., T. D. KOCHER,AND A. C. WILSON. 1991. MADDISON, W. P., AND D. R. MADDISON. 1993. Evolution of the cytochromeb gene of mammals. MacClade 3.1 Interactive analysisof phylogeny Journal of Molecular Evolution 32:128-144. and character evolution. Sinauer, Sunderland, JERMIIN,L. S., D. GRAUR,R. M. LOWE,AND R. H. CROZIER. Massachusetts. 1994. Analysis of directional mutation pressure MANIATIS,T., E. F. FRITSCH,AND J. S•BROOK. 1982. and nucleotide content in mitochondrial cyto- Molecular cloning: A laboratorymanual. Cold chromeb genes.Journal of Molecular Evolution Spring Harbor Laboratory,Cold Spring Harbor, 39:160-173. New York. JOUAN•N,C., AND J.-L. MOUCIN. 1979. Order Procel- MARCHANT,S., AND P. J. HIGGINS. 1990. Handbook lariiformes.Pages 48-121 in Check-listof birds of Australian, New Zealand and Antarctic birds, of the world, vol. 1, 2nd ed. (E. Mayr and G. W. vol. 1. Ratitesto .Oxford University Press, Cottrell, Eds.). Harvard University Press,Cam- Oxford. bridge, Massachusetts. MARTIN,A. P., G. J. P. NAYLOR,AND S. R. PALUMBI. JOUVENTIN,P., AND H. WEIMERSKIRCH.1988. Demo- 1992. Rates of mitochondrial DNA evolution in graphicstrategies of southernalbatrosses. Pages sharksare slow comparedwith mammals.Nature 857-868 in Acta XIX CongressusInternationalis 357:153-155. Ornithologici(H. Ouellet, Ed.). Ottawa,Ontario, MARTIN,A. P., ANDS. R. PALUMBI.1993. Body size, 800 NUNNET AL. [Auk,Vol. 113

metabolicrate, generationtime, and the molec- DNA and amino acid sequences.Journal of Mo- ular clock.Proceedings of the NationalAcademy lecular Evolution 27:326-335. of Sciences USA 90:4087-4091. PRUM,R. O. 1990. Phylogeneticanalysis of the evo- MATHEWS,G. M. 1912. , vol. 2. H. lution of display behavior in the Neotropical F. & G. Witherby, London. manakins(Aves: Pipridae). 84:202-231. MATHEWS, G.M. 1934. Remarks on albatrosses and QUINN,T. W. 1992. The genetic legacy of Mother mollymawks. Ibis Series 13(4):807-816. :Phylogeographic patterns of LesserSnow MATHEWS,G.M. 1948. Systematicnotes on the pet- Goose Chen caerulescens caerulescens maternal lin- rels.Bulletin of the BritishOrnithologists' Club eages.Molecular Ecology1:105-117. 68:155-170. REICHENBACH,L. 1852. Avium Systema Naturale. MATHEWS,G. M., ANDE. J. L. HALLSTROM.1943. Notes Leipzig, Germany. on the orderProcellariiformes. The FederalCap- RICHMAN, A.D., AND T. PRICE. 1992. Evolution of ital Press of Australia Limited, Canberra. ecologicaldifferences in the Old World leaf war- MEYER,A. 1994. Shortcomingsof the cytochromeb blers. Nature 355:817-821. gene as a molecularmarker. Trends in Ecology ROSERTSON,C. J. R., AND J. WAIU-IAM. 1992. Nomen- and Evolution 9:278-280. clature of the New Zealand Wandering Alba- MILLER, L. 1962. A new albatross from the Miocene trosses Diomedea exulans. Bulletin of the British of California. Condor 64:471-472. Ornithologists'Club 112:74-81. MURPHY, R.C. 1917. A new albatross from the west Roux, J.-P., P. JOUVENTIN,J.-L. MOUGIN, J.-C. STAHL, coast of . Bulletin of the American AND H. WEIMERSKIRCH. 1983. Un nouvel alba- Museum of Natural History 37:861-864. trossDiomedea amsterdamensis n. sp. d•couvert sur MURPHY, R.C. 1936. Oceanic birds of South America, File Amsterdam (37ø50'S,77ø35'E). Oiseau 53:1- vol. 2. AmericanMuseum of NaturalHistory, New i1. York. SANGER, F., S. NICKLEN, AND A. R. COULSON. 1977. MURPHY, R. C. 1945. Review. Notes on the Procel- DNA sequencingwith chain termination inhib- lariiformes. Auk 62:153. itors. Proceedingsof the National Academy of MURRAY,B. W., W. B. McGILLIVRAY,J. C. BARLOw,R. Sciences USA 74:5463-5467. N. BEECH, AND C. TROSECK. 1994. The use of SERVENTY,D.L. 1950. Taxonomic trends in Austra- cytochromeb sequencevariation in estimationof lian ornithology with special reference to the phylogeny in the Vireonidae. Condor 96:1037- work of GregoryMathews. Emu 49:259-267. 1054. SHIELDS,G. F., AND A. C. WILSON. 1987. Calibration NUNIX/,G. B.,AND J. CRACRAFT.1996. Phylogenetic of mitochondrialDNA evolution in geese.Jour- nal of Molecular Evolution 24:212-217. relationshipsamong the majorlineages of the Birds-of-Paradise(Paradisaeidae) using mito- SIBLEY,C. G., AND B. L. MONROE, JR. 1990. Distri- chondrialDNA genesequences. Molecular Phy- bution and taxonomyof birds of the world. Yale logeneticsand Evolution 5:445-459. University Press, New Haven, Connecticut. SINGER-SAM,J., R. L. TANGUAY,AND A.D. RIGGS. 1989. OLSEN,G. J., H. MATSUDA,R. HAGSTROM,AND R. OVER- SEEK. 1994. FastDNAml: A tool for construction Use of Chelex to improve the PCR signalfrom a small number of cells. Amplifications:A forum of phylogenetictrees of DNA sequencesusing for PCR users 3:11. maximum likelihood. Computing for the Ap- SMITH, E. F. G., P. ARCTANDER,J. FJELDS.•.,AND O. G. plied Biosciences10:41-48. AMIR. 1992. A new speciesof shrike (Laniidae: OLSON,S. L. 1985a. An early marine avi- Laniarius)from Somalia, verified by DNA se- fauna from Duinefontein, Cape Province,South Africa. Annals of the South African Museum 95: quencedata from the only known individual. Ibis 133:227 -235. 147-164. SMITH, M. F., W. K. THOMAS,AND J. L. PATRON. 1992. OLSON,S.L. 1985b. Early PlioceneProcellariiformes MitochondrialDNA-like sequencein the nuclear (Aves)from Langebaanweg,southwestern Cape genome of an akodontine rodent. Molecular Bi- Province, South Africa. Annals of the South Af- ology and Evolution 9:204-215. rican Museum 95:123-145. SWOFFORD,D. L. 1993. Phylogeneticanalysis using PATERSON,A.M., G. P. WALLIS,AND R. D. GRAY. 1995. parsimony, 3.1.1. Distributed by the author. ,petrels, and parsimony:Does cladistic SmithsonianInstitution, Washington,D.C. analysisof behaviorreflect seabird phylogeny? TEMPLETON,A. R. 1983. Phylogeneticinference from Evolution 49:974-989. restrictionendonuclease cleavage site mapswith PETERS,J. L. 1931. Check- of the world, particular reference to the evolution of humans vol. 1. Harvard University Press,Cambridge, and the apes. Evolution 37:221-244. Massachusetts. THOMAS, W. K., AND S. L. MARTIN. 1993. A recent PRAGER,E. M., AND A. C. WILSON. 1988. Ancient origin of marmots.Molecular Phylogeneticsand origin of lactalbuminfrom lysozyme:Analysis of Evolution 2:330-336. October1996] AlbatrossMolecular Phylogeny 801

WALSH, P.S., D. A. METZGER, AND R. HIGUCH•. 1991. WINKLER, D. W., AND F. H. SHELDON. 1993. Evolution Chelex 100 as a medium for simple extraction of. of nest construction in swallows (Hirundinidae): DNA for PCR-basedtyping from forensicmate- A molecular phylogenetic perspective.Proceed- rial. Biotechniques10:506-513. ings of the National Academy of SciencesUSA WARHAM,J. 1990. The petrels: Their ecology and 90:5705-5707. breeding systems.Academic Press,London. WITTW•R,C.T. 1992. Buffers and reaction compo- WAP,HEIT, K. I. 1992. A review of the fossil seabirds nents for rapid cycling. The Rapid Cyclist 1:6-8. from the Tertiary of the North Pacific:Plate tec- WITTWO, C. T., G. C. FILLMORE,AND D. R. HaLLYARD. tonics,paleoceanography, and faunal change.Pa- 1989. Automated polymerasechain reaction in leobiology 18:401-424. capillary tubes with hot air. Nucleic Acids Re- WETlYlORE,A. 1943. Fossil birds from the Tertiary search 17:4353-4357. depositsof Florida. Proceedingsof the New En- WOLTERS,H.E. 1975. Die vogelarten der Erde: Eine gland Zoological Club 22:59-68. systematischeListe mit Verbreitungsangaben WILKINSON,H.E. 1969. Descriptionof an upper Mio- sowie deutschenund englischenNamen. Parey, cene albatross from Beaumaris, Victoria, Austra- Berlin. lia, and a review of fossil Diomedeidae. Memoirs of the National Museum Victoria 29:41-51.

APPENDIX. Individual taxa sequenced(with binomen or current trinomen for traditional polytypic species) and their collectionlocalities and dates.Taxonomic scheme of genera follows our suggestedphylogenetic classificationof higher-level groups in Diomedeidae.

Taxon Collection locality and date Diomedeidae Thalassarche(southern mollymawks) Thalassarchechlororhynchos chlororhynchos (Yellow-nosedMollymawk) , Atlantic Ocean; October 1990. Thalassarchebulleri bulleri (Buller's Molly- mawk) Snares Islands, New Zealand; March 1993. Thalassarchechrysostoma (Gray-headed Mol- lymawk) Marion Island, ; April 1993. Thalassarchemelanophris melanophris (Black- browed Mollymawk) Near , South Africa; April 1992. Thalassarchecauta cauta (Shy Mollymawk) Pedra Branca,Tasmania; August 1992. Phoebetria(sooty albatrosses) Phoebetriapalpebrata (Light-mantled ) Marion Island, Indian Ocean; April 1993. Phoebetriafusca (Dark-mantled SootyAlba- tross) Marion Island, Indian Ocean; April 1993. Diomedea"great" albatrosses) Diomedeaexulans dabbenena (Wandering Al- batross) Gough Island, Atlantic Ocean;October 1990. Diomedeaamsterdamensis (Amsterdam Alba- tross) Ile Amsterdam, Indian Ocean; July 1993. Diomedeaepomophora sanfordi (Royal Alba- tross) Forty-fours,Chatham Islands, New Zealand;March 1992. Phoebastria(North Pacific albatrosses) Phoebastriaimmutabilis ( Albatross) North PacificOcean (40ø06'N,161ø30'E); August 1991. Phoebastrianigripes (Black-footed Albatross) North PacificOcean (WashingtonState, USA); 1988. Phoebastriairrorata () Isla Espafiola,Galapagos Islands; October 1993. Phoebastriaalbatrus (Short-tailed Albatross) Torishima, ;April 1993. Procellariidae Macronectesgiganteus (Southern Giant Pe- trel) Marion Island, Indian Ocean; October 1990. Procellariacinerea (Gray Petrel) Gough Island, Atlantic Ocean;October 1990.