TheAuk 116(3):702-711, 1999

STRONG DIVERSIFICATION AT THE TREELINE AMONG METALLURA

JAIMEGARCiA-MORENO, •'3 PETER ARCTANDER, • AND JONFJELDS• 2 •ZoologicalInstitute, Department of PopulationBiology, University of Copenhagen,Universitetsparken 15, DK-2100Copenhagen O, Denmark;and 2ZoologicalMuseum, University of Copenhagen,Universitetsparken 15, DK-2100 Copenhagen O, Denmark

ABSTR^CT.--Thehummingbird Metallura comprises nine . Six of themlive at the treelineand replaceeach other sharplyalong the easternslope of the tropicalAndes (williami,baroni, odomae, theresiae, eupogon, and aeneocauda). Their rangesoverlap for 3,000km alongthe Andeswith M. tyrianthina,which lives at lowerelevations and is differentiated into severalsubspecies that showclinal variation. The genusalso includes M. phoebein semi- arid westernPeru and M. iracundain the PerijaMountains of the northernAndes. The group couldbe a goodmodel to studyrelative differences in diversificationbetween montane forest assuch and the narrowtransition zone toward the barren highlands. Analysis of nucleotide sequencesfrom three different mitochondrial gene fragments (cytochrome b,ND2, and NDS) showthat Metalluraforms a monophyleticgroup whose sister taxon is the genusChalcostig- ma.The treeline forms of Metallura,including the morphologically divergent M. phoebe,group in a cladesister to M. tyrianthina,confirming the idea that montane forest and treeline forms are sistertaxa in a strictsense. Neighboring treeline species show greater morphological and geneticdifferentiation relative to neighboringmontane forest forms. The splitbetween mid- elevationand treelineforms is estimatedto haveoccurred during the Pliocene,suggesting that muchof theMetallura radiation took place during the Pleistocene.Received 6 March 1998, accepted14 December1998.

ALTHOUGH MUCH ATTENTION has been de- and the consequentvulnerability of local pop- voted to speciationin the lowlands,montane ulationsin this habitat.Intensive speciation of areasmay play a centralrole in biologicaldi- local remnant populationsat the treeline po- versificationin the tropics(Vrba 1993,Roy et al. tentially could recruit speciesinto lower ele- 1997).Among of humid Andeanforests, vational zones and contribute to the submon- two commonlyseen distributional patterns are tanepeak of biologicaldiversity (Rahbek 1997), (1) veryelongate ranges where a speciesis con- or possiblyeven recruit speciesinto the low- tinuously presentfrom northern Colombiato lands (Roy et al. 1997). Bolivia, with rather weak or clinal subspecific An idealsituation for studyingthe roleof the differentiation;and (2) sharp replacementof treeline zone in diversification would be to closely related species along the mountain comparesister taxa that have widely overlap- range,in many caseswhere the cloudforest is ping ranges but are segregatedelevationally. intersectedby a deep valley with a hot, dry One group that is especiallywell suited for tropical climate.The boundariesbetween spe- suchan analysis is thehummingbird genus Me- cies could represent initial isolating barriers tallura,which comprisesnine recognizedspe- (O'Neill 1992) or secondary contact zones, cies (Graves 1980, Heindl and Schuchmann whichwould be maintainedmost easily where 1998).Metallura is part of a species-richAndean a physicalboundary also exists (Fjeldsfi 1995). radiation (Bleiweisset al. 1997) that presum- Graves(1985, 1988) demonstrated that range ably illustrates speciationevents mainly of disjunctionsand morphologicaldifferentiation Pleistoceneage (Fjeldsfi 1992). Members of the in the Andesincrease with elevationup to the genus are characterizedby rather small, treeline,possibly reflecting the extremenar- straightbills and broadtails wherethe irides- rownessof ecologicalzones at high elevations cent colorschange according to the angle of view. As currentlyrecognized, the genuscom- 3Present address: Museum of Zoology,University prises one species,M. tyrianthina,(Tyrian Me- of Michigan,1109 Geddes Avenue, Ann Arbor,Mich- taltail),which is widespreadin humidmontane igan 48109,USA. E-mail: [email protected] forestsin the tropical Andes region, and one

702 July1999] PhylogenyofMetallura Hummingbirds 703 speciesgroup (the M. aeneocaudagroup; see erated to discussrelationships among these Graves 1980, Heindl and Schuchmann1998) birds and possiblespeciation scenarios, and we that inhabits humid treelines, elfin forests,and use the molecular data to assess the effects of adjacentshrubby paramo habitat in the same elevationin the diversificationprocess of the region.The two overlapfor 3,000 km from group. northern Colombia to Cochabamba in Bolivia (Fig. 1). The genusalso includes two ratherab- MATERIALS AND METHODS errant, large melanistic species:M. iracunda (Perij• ), which is a local endemicof Samples.--Tissuesof eight of the nine Metallura the isolatedPerij• Mountainson the borderof speciesrecognized by Graves (1980) were included Colombiaand Venezuela,and M. phoebe(Black (Table1, Fig. 1). Samplesof Metallurairacunda, as well as samplesof the subspeciesMetallura w. williami M. Metaltail),which inhabits mist vegetationand m recisa,M. theresiaeparkeri, M. aeneocaudamalagae, Polylepiswoodlands high abovethe Peruvian Metalluratyrianthina septentrionalis, M. t. oreopola,and coastal desert. M. t. districtawere not availablefor analysis.Tissue Metalluratyrianthina is commonthroughout samplesfor taxa analyzedwere obtainedfrom the much of the zone of humid montane forest and blood collectionof the ZoologicalMuseum of Co- is differentiatedinto sevensubspecies (Fjelds• penhagenand the LouisianaState University Muse- and Krabbe 1990, Heindl and Schuchmann um of Natural Science frozen tissue collection. To an- 1998)with somebroad zones of intergradation alyze the relationshipsof metaltailswith thornbills, (Zimmer 1952). In contrast,the M. aeneocaudawe included samplesof Chalcostigmaruficeps and C. herrani(Rainbow-bearded Thombill). We used Erioc- groupinhabits an ecologicalzone that is usu- nemisnigrivestis (Black-breasted Puffleg) and Helian- ally only a few kilometerswide (Graves1988) gelusviola (Purple-throated Sunangel) as outgroups. and is sharplydifferentiated into severalspe- Both speciesbelong to a different cladethan Metal- ciescharacterized by differentcombinations of lura and Chalcostigmawithin the Trochilinae (see colorsof tail and body plumage,scaliness of Bleiweisset al. 1997). the plumage,and colorsand shapeof the gor- DNA extraction.--DNA was extracted from blood get. Sharpreplacements and non-clinalcheck- samples preserved in EDTA (Arctander 1988) or erboard character variation, and indications of DMSO (Arctander and Fjeldsa 1994). Following marginal sympatry in two places in Ecuador Doyle and Doyle (1987), sampleswere washed in blood buffer (NaC1 0.1 M, Tris-HC1 0.01 M pH 8, (see below), suggestspecies rank for several EDTA 0.001 M pH 8), incubated in CTAB buffer forms (Graves1980, Fjelds• and Krabbe 1990; (Tris-HC10.1 M pH 8, NaC11.4 M, EDTA 0.02 M pH seeFig. 1): M. williami (), M. 8, 0.2% 2-mercaptoethanoland 2% hexadecyl-tri- baroni (Violet-throatedMetaltail), M. odomae methyl-ammonium-bromide[CTAB]) at 65øCfor a (), M. theresiae(Coppery Me- minimum of 1 h, and extracted with chloroform / iso- taltail), M. eupogon(Fire-throated Metaltail), pentyl alcohol(24:1 v/v). DNA was then precipitat- and M. aeneocauda(). ed with isopropanol,cleaned twice with ice-coldeth- It wasour a prioriassumption that M. tyrian- anol,and dissolvedin TE buffer (10 mM Tris-HC1pH thinaand the M. aeneocaudacomplex of metal- 7.6, 1 mM EDTA). tail hummingbirdsare sistertaxa, and that a Alternatively,DNA was extractedfrom piecesof skin of museumspecimens, or tissuesother than comparisonof the two would illustratethe ef- blood, by soaking in 250 p.L of blood buffer, dis- fectsof the extremelynarrow habitats in which rupting with a pistil, and incubatingovernight at treeline speciesdwell. However, some uncer- 65øCwith ProteinaseK (final concentration1 mg/ tainty arisesbecause the relationshipof Metal- mL) and SDS(final concentration1% w/v). The sam- lura with "thornbill" hummingbirds(Chalco- pleswere then extractedonce with phenoland once stigmaspp.) is unresolved,because one species, with chloroform/isopentylalcohol (24:1) and pre- C. ruficeps(Rufous-capped Thornbill), com- cipitatedby addingan equalvolume of isopropanol. bines charactersof the two genera (Fjeldsg PCR andsequencing.--PCR products were obtained 1992,Schuchmann and Heindl 1997). using primers L14857 (Nunn and Cracraft 1996), H15149 (Kocheret al. 1989), and H15915 (Edwards In this paper,we make a phylogeneticanal- et al. 1991)for cytochromeb (cyt b);L5215 and H5578 ysis of hummingbirdsof the genusMetallura (Hackett 1996)for ND2; and L13753(CAG GAA AAT usingDNA sequencedata from threedifferent CCG CTC AAT TCG G), L13589 (CCA GCA GCA mitochondrialgene fragments, for a totalof 855 ATA GAA GGC CC), and H14149 (CCT ATT TTG basepairs (bp). We use the phylogeny thus gen- CGG ATG TCT TGT TC) for ND5 (numbers follow 704 GARCiA-MORENO,ARCTANDER, AND FJELDS,k [Auk, Vol. 116

TABLE1. Speciesand area of collectionfor the samplesused in this study.

District Species Country (no. of samples)a Metallura tyrianthina Ecuador Loja 2 (1) Metalluratyrianthina Ecuador Azuay I (2) Metallura tyrianthina Ecuador Pichincha2 (1) Metalluratyrianthina Ecuador Carchi 1 Metalluratyrianthina Ecuador Imbabura 1 (2) Metalluratyrianthina Ecuador Napo 2 (2) Metalluratyrianthina Ecuador Carchi 1 Metallura tyrianthina Ecuador Bolivar 0 (1) Metalluratyrianthina b Peru Pasco 1 Metalluratyrianthina Peru Huancavelica2 (1) Metalluratyrianthina Peru Apurimac 3 (1) Metalluratyrianthina Bolivia La Paz 1 Metalluratyrianthina Bolivia Cochabamba0 (2) Metallura williami Ecuador Zamora-Chinchipe2 (2) Metallura williami Ecuador Carchi,Napo, Morona-Santiago 3 Metallura baroni Ecuador Azuay 1 Metallura odomae Ecuador Loja/ Zamora-Chinchipe2 (2) Metalluraphoebe Peru Lima 1 (1) Metallura theresiae Peru Huanuco 2 Metalluraeupogon b Peru Huanuco and Pasco 2 Metallura aeneocauda b Bolivia La Paz 1 Chalcostigmaruficeps Peru Cuzco 2 Chalcostigmaherrani Colombia Valle 1 Eriocnemisnigrivestis Ecuador Pichincha 1 Heliangelusviola Ecuador Loja/ Zamora-Chinchipe1 (1) • Firstnumber refers to completesequences; number in parenthesesrefers to extraindividuals that were sequenced only partially (i.e. one or twoof thegene fragments only). bSample provided by LSUcollection.

Desjardin and Morals 1990). Double-stranded PCR AF022659 to AF022711, U85723, U85730, U85731, was performed in 20 •L reaction volumes using and U85738. equalamounts of primers;5 •L of the reactionwere Analyses.--Sequenceswere aligned by eye using electrophoresedin a low meltingpoint agarosegel, the programSeqApp 1.9 (Gilbert1992). They were and the bandswere cut out followingstaining with translatedto amino acid sequenceusing MacClade ethidiumbromide and meltedin 400 •L of TE buffer. 3.06 (Maddison and Maddison 1992) to check for Thesolution (1 •L) wasused as template to generate possiblestop codons.We usedthe programCS3 (H. single-strandedDNA by asymmetricPCR amplifi- Siegismundunpubl.) to calculatethe total numberof cation. The conditions were identical to those of the substitutionsand the numberof transitions(ts) and balanced-primerreaction except that the reaction transversions(tv) between homologousDNA se- volumewas increasedto 50 •L, the limiting primer quences.These figures were comparedbecause it is wasdiluted one hundred fold, and thenumber of cy- knownthat mitochondrialand nuclear sequences ex- cleswas increased by three.Reaction products were hibit differencesin their substitutionpatterns (Arc- cleanedby washingthree times with ddH20through tander 1995). To assesssaturation of the sequences Ultrafree-MCf (Millipore Corporation) 30,000- we plottedthe observednumber of substitutions(p NMWL filters and concentrated to a final volume of distances)versus Kimura two-parameterdistances, 18•L, of which8 •L wereused for DNA sequencing. whichadjust the raw estimatebased upon differenc- An alternativefor generatingsingle-stranded PCR es in substitution rates between ts and tv. Deviation products was to produce double-stranded PCR from the line y = x was takenas an indicationthat productsusing one biotinylated primer in 35-•L re- multiple substitutionshad takenplace at somepo- actions.The productswere purified using Dynal sitions(Burns 1997). Initially, we analyzed the se- magneticbeads following the protocolof the sup- quencesof the cyt b, ND2, and ND5 genefragments plier.Dideoxy sequencing was performedfollowing separately,and we alsomade an analysiscombining the protocolfor Sequenaseversion 2.0 (UnitedStates the sequencesof the three fragments.Analyses in- Biochemicals).The products,labeled with S3s, were cludedthe use of maximum-likelihood,parsimony, resolvedin 6% acrylamidegels. Sequences have been and neighbor-joiningalgorithms. submitted to Genbank under accession numbers For the maximum-likelihoodanalysis, we used July1999] PhylogenyofMetallura Hummingbirds 705

lmbab• Pichincha,

Cafiar

,.•:•ßCuzco Flc. 1. Distributionsof all membersof the genusMetallura (from the databaseof the distributionof An- deanbirds in theZoological Museum, University of Copenhagen).Each block is 15 minutesof latitudeand longitude(ca. 28 x 28 km at this latitude).Species' ranges are madeby conservativeinterpolation between confirmedrecords (see Fjeldsfi and Rahbek1997). Locality names mentioned in the text are givenmainly on the left map.Left: the two largemelanistic forms M. iracunda(Perijfi Mountains on the Colombia/Venezuela border)and M. phoebe(woodlands and shrubsteppe high abovethe Peruviancoastal desert and locallyin intermontanebasins). Center: a complexof morphologicallydistinct forms that replace each other along the Andes in humid treelinehabitats, from the north M. williami (in Colombiaand northernEcuador, with sub- speciesrecisa, williami and primolinus),M. williamiatrigularis (continental divide in centralEcuador), M. baroni (Cajasplateau of Azuay,Ecuador), M. odomae(along the continentaldivide from Loja,Ecuador, into extreme northernPeru), M. theresiae(the easternAndes ridge betweenthe North PeruLow and Rio Huallaga,with subspeciesparkeri and theresiae),M. eupogon(eastern Andes ridge betweenthe R•osHuallaga and Apurlmac), and M. aenecauda(Cordillera Oriental from Cuzco,Peru, to Cochabamba,Bolivia, with subspeciesmalagae furthestsouth). Right: M. tyrianthina(widespread in humidmontane forest throughout the tropicalAndes, includingsome relict cloudforests on the Pacificslope, with sevensubspecies; note that subspecieschroro- pogoninhabits Venezuelan coastal mountains outside the mappedarea). program DNAml in Phylip (Felsenstein1995). The 3, and probabilitiesof 0.7, 0.25,and 0.05,respective- transition-transversionratio was estimated empiri- ly. Forthe parsimony analysis, we searchedusing the cally.We attemptedto accommodatefor rate hetero- branch-and-boundoption in PAUP 3.1.1 (Swofford geneityin the sequences(Swofford et al. 1996,Yang 1991)with the farthestaddition sequence.We per- 1996).Yang (1996) reported that among-siterate var- formedbootstrap replicates with randomaddition of iation in cyt b of deuterostomesis bestdescribed by sequences(500 replicates,but only 200 replicates a gammadistribution with a parameterof 0.44.We with maximumlikelihood). For the distanceanalysis made a coarseapproximation to a gamma distribu- a tree was producedusing the Neighbor-Joiningal- tion with a parameterof 0.5 by statingthree cate- gorithmbased on an uncorrectedp distanceusing goriesof rates(which do not necessarilycorrespond MEGA 1.01 (Kumar et al. 1993). to codonpositions) with magnitudesof 0.25,1, and We comparedour resultswith a usertree forcing 706 GARCiA-MORENO,ARCTANDER, AND FJELDS•. [Auk, Vol. 116

M. tyrianthinaand the M. aeneocaudagroup to be sis- variation,with ND2 showingthe highestvari- ter taxa with polychotomouspopulation differenti- ation between species. Fewer transversions ationand equalbranch lengths within each,with M. were detectedin ND5 comparedwith the other phoebe(and M. iracunda) representing a deeper branch. two fragments. Geneticdistances among Metallura were rel- atively small (•10%; Fig. 2), a situation in RESULTS which uncorrected(p) distancesare expected With one exception,the sequencesfor each to givea goodestimate of divergencewith the fragmentfrom eachof the taxa used in the least variance (Kumar et al. 1993, Swofford et analysiswere obtainedwithout problems. The al. 1996).Genetic distances among the M. ty- exceptionwas ND2 for Metalluraphoebe, which rianthinasubspecies ranged between 0.010 and yieldedsequences that wereunreadable due to 0.036 (f = 0.027, n = 21), and within the M. double bands from a certain base onwards. aeneocaudasuperspecies, including M. phoebe, ThesePCR products were subsequently cloned rangedfrom 0.002to 0.085(f = 0.047,n = 28). usingthe pCR-ScriptAmp SK (+) cloningkit Despitethe small distances,we detectedindi- (Stratagene).Sequencing of the clonesyielded cationsof multiplehits in somethird- andfirst- two sequencesidentical except for thepresence codon positions (as seen from uncorrectedvs. of an extra triplet (ATC) at position5362 (num- Kimura two-parameter distance plots; Burns bers follow Desjardin and Morais 1990). We 1997). choseto use the sequencewithout the extra Samplesof M. tyrianthinafrom severalEcu- triplet for our comparisons,since use of the adorian localitiesyielded identical sequences longersequence would require the insertionof ("Ecuador"in Fig.3). Birdsfrom Nape, Pichin- gapsin all othersequences (a lessparsimoni- cha,Imbabura, Bolivar, and Azuay showedthe ousoption). It is beyondthe scopeof this work samehaplotypes for 345 bp of cyt b. Similarly, to determinewhether the double-bandpattern samplesfrom Cochabambashowed no differ- resultingfrom this insertionis due to hetero- encefrom thoseof La Paz,although they were plasmyin the individualssequenced or to a dif- 200 km apart. The smallestnon-zero distance ferentphenomenon. was that between Pasco and Huancavelica Levelsand patterns of variation.--Allthree se- (0.011),which are 300 km apart. A larger dis- quencedfragments showed similar levelsof tance (0.027) was found over 200 km between

2 3 4 5 6 7 8 10 11 12 13 14 15 16 17 18 19 1- M. w.at•gularis 17/4 10/0 17/4 39/7 22/2 24/2 46/3 57/4 55/5 54/7 60/7 62/6 52/5 53/3 51/7 47/7 79/20 71/14 2. M. w.primolinus 0.025 20/4 16/3 34/8 39/6 4,3/6 60/7 60/7 59/9 57/11 62/11 66/8 55/9 56f7 54/11 45/11 84/22 75/16 3. M. baroni 0.012 0.029 18/5 44/7 33/'2 37/2 54•3 59/4 59/5 61/7 63/7 65• 57/5 58/3 61/6 51/7 87/22 77/14 4. M. odomae 0.026 0.023 0.028 34/9 32/4 35/5 55/5 5,5• 52/8 52/10 56/10 59/8 53/8 51/6 .54/9 48/7 78/22 67/15 5- M.phoebe 0.056 0.049 0.061 0.051 55/5 58/5 65/6 65/7 65/8 65/10 65/10 70/7 64/8 61/6 55/9 45/9 90/22 79/13 6. M. theresiae 0.029 0.05 0.043 0.044 0.072 2/0 42/1 52/2 52/3 49/5 55/5 57/4 51/3 48/1 52/5 53/5 89/19 72/11 7. M. eupogon 0.031 0.057 0.046 0.047 0.074 0,002 42/1 53/2 54•3 51/5 56/5 58/4 52/3 49/1 54/5 52/5 90/20 75/12 8- M. aeneocauda 0.059 0.080 0.069 0.072 0.085 0.052 0.051 55/3 57/4 55• 54/6 58/5 54/4 53/2 57/6 52/6 91/20 73/13 9- M. tyrianthina- Carchi 0.075 0.080 0.076 0.073 0.086 0.066 0.065 1O- M. tyrianthina- Ecuador 0.072 0.080 0.076 0.072 0.086 0.066 0.067 0.0700.072 0.010- 8/0- 9/2 12/2 27/228/2 26/1 28/0 ,• 56/356/4 60/362/3 95/1894/19 78/1585/14 11-M. tyrianthina-Loia 0.075 0.081 0.082 0.075 0.089 0.066 0.067 12- M. tyrianthina-Pasco 0.081 0.086 0.083 0.078 0.088 0.072 0.071 0.0730.0170.0130.071 0.034 0.033 0.036- 26/4- 24,/36/3 23/220/2 1•/44 54/659/6 56/561/5 95/2189/21 77/1779/17 13- M. tyrianthina-Huancavelica 0.082 0.087 0.084 0.079 0.090 0.073 0.073 0.0750.032 0.032 0.032 0.011 - 22/121/3 I 59/3 62/4 91/18 81/16 14. M. tyrianthina-Apufimac 0.069 0.075 0.074 0.072 0.085 0.065 0.084 15-M. tyrianthina-Bolivia 0.068 0.074 0.072 0.068 0.079 0.059 0.058 0.0690.065 0.0330.032 0.0310.029 0.0300.032 0.0260.027 0.0270.028 0.020- 15/2- 57/460/6 59/363/5 86/1987/21 76/1573/13 16-Chalcostigma rufk;eps 0.071 0.077 0.080 0.075 0.076 0.069 0.070 17-Chalcostigma herrani 0.068 0.071 0.074 0.070 0.068 0.073 0.072 0.0750.073 0.0710.080 0.0710.082 0.0720.078 0.0770.083 0.0760.083 0.0720.078 0.0780.086 0.•5•'•77/19 80114 74/1485/15 18- Eriocnemisnigrivestis 0.120 0.124 0.129 0.119 0.132 0.130 0.129 0.132 0.134 0.133 0.138 0.129 0.128 0.123 0.126 0.113 0.118 103/22 19-Heliangelus viota 0.103 0.090 0.108 0.098 0.109 0.100 0.102 0.103 0.118 0.110 0.112 0.113 0.114 0.110 0.101 0.119 0.111 0.147

FIG.2. Pairwisesequence comparisons among the differenttaxa used in thisstudy. Uncorrected sequence divergenceis shownbelow the diagonal,and the numberof transitions/transversionsis shownabove the diagonal.Based on 855bp of mtDNA sequencefrom cyt b (345bp), ND2 (290bp), and ND5 (220bp). Boxed comparisonsare thosebetween species of the M. aeneocaudacomplex (1 to 8), M. tyrianthinafrom different localities(9 to 15), and Chalcostigmaspecies (16 and 17). July1999] PhylogenyofMetallura Hummingbirds 707

73 Iwifiiami atrigularis However,both Heliangelusand Eriocnemisare rather distant from Metallura to function as proper outgroups,and they belongto a differ- 59'--•© F williamiprimolinus ent group of non-hermithummingbirds (Blei- 62 I odomaephoebeweiss et al. 1997).Having confirmedthe mono- 75 phyly of Metallura, and its close association 83 theresiae with Chalcostigma,further phylogeneticanaly- 8•eupogonIFbarøni seswere carriedout with Chalcostigma herrani aeneocauda declaredas the outgroup. A basal dichotomybetween M. tyrianthina 6•rEcuador and all treeline forms (includingM. phoebe;Fig. 3) was found in the optimal trees recovered 9•._•I• Carchi with all the differentalgorithms. According to I •Loja M. tyrianthina the Kishinoand Hasegawa(1989) test, the to- 9•J'•Pasco pologies of the maximum-likelihood tree ß I '-'-Huancavelica (-log• likelihood (L) = 2,697), the parsimony bootstrap-consensustree (-log, L = 2,719,P = b•'•Po•ii•ac 0.180),and theneighbor-joining tree (-loge L = Chalcostigmaherrani 2,721,P = 0.091)were not significantlydiffer- 981 ent. In contrast, the null model in which we Chalcostigmaruficeps forced M. tyrianthinaand the M. aeneocauda 0.01 group to be sistertaxa, with M. phoebe(and M. iracunda)representing a deeperbranch, was a Maximum-likelihood tree obtained from significantly poorer hypothesis (-log, L = 855 bp of combinedDNA sequencesof mitochondri- 2,939, P < 0.0001). al cyt b (345 bp), ND2 (290 bp), and ND5 (220 bp) In all trees,M. phoebewas groupedwith the genes.Figures above the branchesrepresent boot- strap values with the maximum-likelihoodalgo- northerntreeline forms, and the two subspe- rithm after 200 replicates.Branches are drawn pro- ciesof M. williamithat we used(atrigularis and portionateto theamount of changealong the branch, primolinus)did not group as sistertaxa. Unfor- and thosewith low bootstrapsupport (<50%) have tunately,we cannottell where M. iracundafits beencollapsed. Names in theM. tyrianthinaclade re- in, althoughplumage details and especiallythe fer to localitieswhere the samplesoriginated from strongsexual dimorphism, would suggestthat and are depicted in Figure 1; Ecuador refers to a it is nearestto M. tyrianthina. widespreadhaplotype found in all localitiesother Elevationaleffects.--Because of the linearityof than Carchiand Loja (seeResults). the distributionalranges along the eastern slopeof the Andes(Fig. 1), one couldassume Huancavelicaand Apurimac (the even larger that geographicneighbors were more closely distancebetween Pasco and Loja, 0.036,is less relatedthan non-neighboringspecies (see Fig. revealingbecause of the lack of samplesfrom 3). Amongthe six treelinespedes that replace northwesternPeru). eachother in a linear fashionalong the Andes Phylogeneticanalyses.--The topologies of the (M. williami, M. baroni,M. odomae,M. theresiae, treesbased on eachof the three fragmentsvar- M. eupogon,M. aeneocauda;Fig. 1), geneticdis- ied significantlyand were even contradictory. tancesbetween neighboring forms were highly However, in all phylogenies,the tyrianthina correlatedwith geographicdistances (r 2 = 0.78; birds were closelyrelated, as were most of the Fig. 4). This wasnot the caseamong samples of northernforms of the treelinesuperspecies (M. the mid-elevationspecies M. tyrianthinathat williami atrigularisand M. primolinus,M. baroni, covereda similar geographictransect (r 2 = M. odomae,and M. phoebe),whereas M. aeneo- 0.32). caudawas closeto the cladeformed by M. ther- esiaeand M. eupogon. DISCUSSION When using a combineddata set, all tree- constructionalgorithms yielded trees with Phylogeneticrelationships.---The DNA diver- Chalcostigmaclosely associatedto Metallura. gencedata (Fig. 2) highlight the closerelation- 708 GARCiA-MORENO,ARCTANDER, AND FJELDS,• [Auk,Vol. 116

Birds from Ecuador (M. williami, M. baroni, and M. odomae)form a close-knitgroup. Mar- ginal sympatrymay existbetween species with • 0.04 slightly different ecological requirements • • jJ o (Fjelds&and Krabbe 1990, Stotz et al. 1996).Me- 8 0.03! tallurabaroni and M. w. atrigulariswere found togetherat Rio Maz•n in Azuay, Ecuador (Gretton1986), with M. baronioccupying the o• 0.0• higher elevationalzone; similarly,M. odomae and M. williami primolinushave both been re- portedin the samelocality in PodocarpusNa- Linear GeographicDistance (kms). tional Park in Loja (Rasmussenand Rahbek 1994).However, this assumed sympatry has not FIG. 4. Relationshipbetween genetic distance been documentedby collectionor substantiat- (percentsequence divergence) between neighboring taxa and linear geographicdistance in Metallura ed by long-term studies (N. Krabbe pers. hummingbirds.Crosses denote treeline-bound spe- comm.)and could therefore represent marginal ciesreplacing each other across the eastern slope of or casual contacts. the tropicalAndes, from Ecuadorto Bolivia(M. wil- Thetwo subspeciesof M. williami(atrigularis liami,M. baroni,M. odomae,M. theresiae,M. eupogon, and primolinus)do not come out as sister taxa andM. aeneocauda),and open circles denote Metallura in anyof the phylogeneticanalyses performed. tyrianthina from different localities across a similar It would be temptingto suggestthat the two geographicrange (Carchi, Napo, Azuay, Loja, •asco, were separatespecies, but Moore (1940) stated Huancave]ica,Apurimac, La •az, and Cochabamba; seeTable i and Fig. 1 for localities).Continuous line thatbirds from Chimborazo show signs of in- is regressionfor the treelinespecies set (r• = 0.78), tergradation.Although one of ourM. m primo- and brokenline is regressionfor the lowerelevation linussamples was from the assumedarea of in- M. tyrianthina(r • = 0.32).Geographic distances are tergradation,its mtDNA wasidentical to that of approximatelinear distances (kin) betweensamples. samplesfrom Napo and Carchi.The individual variation of birds from Chimborazo and Mo- rona-Santiagoshould be carefullystudied for a ship between Metallura and Chalcostigma,better interpretation of what is happeningin this area. whichis alsoevident using different characters (Schuchmannand Heindl 1997).Chalcostigma Anotherproblem is how to interpretthe ruficepsis morphologicallyand ecologically in- markeddichotomy of morphologicaland mo- termediatebetween Chalcostigma and Metallura, leculardifferences exhibited by the sisterspe- andthe femalesare difficult to tell apart(even ciesM. theresiaeand M. eupogon.Differences in in thehand) from those of M. t. tyrianthinaand structuralcoloration (e.g. gorget)could be the M. aeneocaudamalagae (Fjelds& and Krabbe 1990, resultof minutechanges in the arrangementof Fjelds&1992). highlyrefractive melanin granules in thefeath- Within Metallura,two distinctgroups were er barbules(Dyck 1976, Greenwalt 1991) and identified: (1) one embracingall representa- thereforemay be of slighttaxonomic signifi- tivesof M. tyrianthinaand (2) one comprising cance.In and of itself,the lack of geneticdif- all membersof the treelinesuperspecies. This ferentiationbetween the two forms(Fig. 2) can- result confirms the idea that montane forest not immediatelybe usedto disprovetheir spe- and treeline forms are sister taxa in a strict ciesstatus, because the low geneticdistance sense.Metallura phoebe, although morphologi- couldindicate a veryrecent and shortperiod of cally very distinctive,clearly is related to the isolation between the two forms. Ecuadoriantreeline taxa and probablyorigi- Thegroupings within the Metallura tyrianthi- nated by successfuldispersal of an ancestral na cladereflect geographic proximity to some form west of the North Peru Low (where the extent(Figs. 2, 3). The relativelylarge genetic Andesreach their lowest elevation) to thewest- changebetween Huancavelica and Apurimac ern Andeanslope of Peru.Its morphological occursin an areawith severaldeep gaps in the distinctnessdoes not reflect a particularlyearly easternAndean ridge and cloud coverthat is isolation. quitevariable (Fjeldsa 1995), which could act as July1999] Phylogenyof MetalluraHummingbirds 709 a filter barrier Birds from Ecuador north of winter Servantet al. (1993) suggestthat the Lojashow no geneticdifferentiation regardless Pleistocenevegetational changes in tropical of whetherthey occur on the easternor western SouthAmerica were driven by increasedeffects slopeof the Andes.Several dispersal corridors of thesewinds. Accordingly,speciation along connectingeastern and westernavifaunas have muchof the Andescould be drivenby ecocli- beenpostulated in Ecuador(Poulsen and Krab- matic disturbancein which sensitivespecies be 1997),which might explainthe lack of ge- survived periods of stressas local isolatesin netic differentiationbetween the seemingly placeswhere the ecologicaldisturbance was isolatedeastern and westernpopulations. The moderated (on various time scales). relatively low variation in Ecuadorcould also Speciationscenario.--Using different substi- suggesta ratherrecent establishment, although tution rates (2% per million years [see Klicka samplesfrom northwesternPeru (M. t. septen- and Zink 1997];10% third-codon position sub- trionalis)and the extremenorthern end of the stitutionsor 0.5% third-codonposition trans- rangewould be neededto firmly establishthis. versionsper million years [Irwin et al. 1991]; Elevationaleffects.--The distribution of M. ty- but see Loewe and Scherer 1997, Hillis et al. rianthinabroadly overlapsthat of the treeline 1996),our moleculardata suggestthat Chalco- forms,with anelevational segregation in which stigmaand Metalluradiverged about4 million the formerspecies occupies lower elevations. yearsago. The early split betweenmid-eleva- Even thoughwe lack samplesof severalof the morphologicallymost divergent M. tyrianthina tion and treelineforms of Metallura(Fig. 3) ap- subspecies,and therefore cannot test this result pearsto haveoccurred soon afterward, also in rigorousl• a cleartrend emerges from our data. the Pliocene,2 to 4 millionyears ago, with sev- Within a similargeographic transect, stretch- eral speciationevents likely to have occurred ing from Ecuadorto Bolivia (Fig. 1), genetic duringthe Pleistocene. Figure 3 suggestsrather distancesbetween neighboring forms are high- similararea relationships in M. tyrianthinaand ly correlatedwith linear geographicdistances in the treelinegroup, with the mostrecent dif- in the treelinespecies group, but only weakly ferentiationin the north.Treeline populations in the mid-elevationM. tyrianthina(Fig. 4). A in the southerntropical Andes appear to have comparisonof geneticdifferentiation between separatedearly in the historyof the group (in neighboringtaxa suggeststhat the treeline Metalluraand Chalcostigma;Schuchmann and birdsshow not only greater morphological dif- Heindl 1997). It seemsthat early vicarlance ferentiation,but also strongergenetic differ- events in the south were followed, in both entiation than the lower-elevation forms. Ran- groups,by a relativelymore recentestablish- dom morphologicalvariation (Graves 1980, ment in the paramosnorth of the North Peru Remsen1984) suggests that stochastic process- Low. However, it is difficult to exclude the al- es and rapid divergencein small populations ternativeinterpretation that the most recent op- (Paterson1985, Carson 1990) are important el- portunitiesfor localdifferentiation were in the ements in the diversification of these birds. north.In thisregard, the lackof samplesof M. Treeline populationsmay disintegrateeasier t. septentrionalisand of Colombianpopulations into tiny isolatesbecause the narrowconfigu- is regrettable,although morphological varia- ration of the ecotonethat they inhabitmakes tion suggestsnorthward polarity and thatboth them extremely sensitiveto environmental groupsoriginated in the southernpart of their changes(Graves 1988); many potentialisolat- range (Heindl and Schuchmann1998). ing mechanisms,such as volcanism, glaciation, and forestfragmentation by climatechange are ACKNOWLEDGMENTS prominentin the Andes.Remote sensing data suggestthat reducedecological disturbance This work was part of a Ph.D. projectby the first occursin placeswith peak concentrationsof author.All threeauthors were supported by theDan- endemismalong the Andeantreeline (Fjelds• ish Natural History ResearchCouncil through a et al. unpubl.data). At leastin thesouthern part grant to the Centerfor TropicalBiodiversity (grant of the tropicalAndes, these places are charac- no. 11-0390).R. Bleiweiss,A. Helbig, B. Slikas, and terizedby localtopographic moderation of the four anonymousreferees read and commentedon influenceof cold south polar winds during earlierversions of the manuscript. 710 GARCiA-MORENO ARCTANDER, AND FJELDS,g• [Auk, Vol. 116

LJTERATURE CITED Balslev,E. Forero,and J. L. Luteyn, Eds.). New York Botanical Garden, New York. ARCTANDER,P. 1988. Comparativestudies of avtan FJELDS,g•,J., AND N. KRABBE.1990. Birds of the High DNA by restrictionfragment length polymor- Andes. Apollo Books,Svendborg, Denmark. phism analysis:Convenient procedures based FJELDS,g•,J., AND C. RAHBEK.1997. Speciesrichness on bloodsamples from live birds.Journal ffir Or- and endemismin SouthAmerican birds: Impli- nithologie129:205-216. cationsfor the designof networksof nature re- ARCTANDER,P.1995. Comparison of a mitochondrial serves. Pages466-482 in Tropical forest rem- gene and a correspondingnuclear pseudogene. nants:Ecology, management, and conservation Proceedingsof the RoyalSociety of LondonSe- of fragmented communities(W. E Lauranceand ries B 262:13-19. R. O. Bierregaard,Jr., Eds). University of Chica- ARCTANDER,P., AND J. FJELDS,i..1994. Andean tapa- go Press,Chicago. culosof the genus$cytalopus (Aves, Rhinocryp- GILBERT,D. 1992.SeqApp. A biosequenceeditor and tidae): A study of speciation using DNA se- applicationprogram, version1.9a157. Universi- quencedata. Pages205-226 in Conservationge- ty of Indiana, Bloomington. netics (V. Loeschke,J. Tomiuk, and S. K. Jain, GRAVES,G. R. 1980.A new speciesof metaltailhum- Eds.).Birkhauser Verlag, Basel, Switzerland. mingbird from northern Peru. Wilson Bulletin BLEIWEISS,R., J. A. W. KIRSCH, AND J. C. MATHEUS. 92:1-7. 1997.DNA hybridizationevidence for the prin- GRAVES,G. R. 1985.Elevational correlates of specia- cipal lineagesof hummingbirds(Aves: Trochil- tion and intraspecificgeographic variation in idae). MolecularBiology and Evolution14:325- plumage in Andean forest birds. Auk 102:556- 343. 579. BURNS,K. J. 1997.Molecular systematicsof tanagers GRAVES,G. R. 1988.Linearity of geographicrange (Thraupinae):Evolution and biogeographyof a and its possibleeffect on the populationstruc- diverseradiation of Neotropicalbirds. Molecu- ture of Andean birds. Auk 105:47-52. lar Phylogeneticsand Evolution8:334-348. GREENWALT,C. H. 1991.Hummingbirds. Dover Pub- CARSON,H. L. 1990.Increased genetic variance after lications, New York. a populationbottleneck. Trends in Ecologyand GRETTON,A. 1986. Birds. Pages33-57 in Rio Mazan Evolution 5:228-230. project 1986 report (E Robinson,Ed.). Oxford DESJARDINS,P., AND R. MORAIS.1990. Sequenceand University, Oxford. geneorganization of the chickenmitochondrial HACKETT,S. J. 1996. Molecular phylogeneticsand genome.A novel gene order in higher verte- biogeographyof tanagersin the genusRarnpho- brates. Journalof Molecular Biology 212:599- celus(Aves). Molecular Phylogenetics and Evo- 634. luton 5:368-382. DOYLE,J. J., ANDJ. L. DOYLE.1987. A rapid DNA iso- HEINDL,M., AND K. L. SCHUCHMANN.1998. Bioge- lation procedurefor small quantitiesof fresh ography,geographical variation, and leaf tissue.Phytochemical Bulletin 19:11-15. of the Andean hummingbirdgenus Metallura, DYCK,J. 1976. Structural colours.Pages 426-437 in Gould 1847. Journalof Biogeography139:425- Proceedingsof the XVI InternationalOrnitho- 473. logical Congress(H. J. Firth and J. H. Calaby, HILLIS, D. M., B. K. gABLE, AND C. MORITZ.1996. Ap- Eds.). Canberra, 1974. Australian Academy of plicationsof molecularsystematics: The stateof Science, Canberra. thefield and a lookto the future.Pages 515-543 EDWARDS,S. V., P. ARCTANDER,AND A. C. WILSON. in Molecularsystematics (D. M. Hillis, C. Moritz, 1991.Mitochondrial resolution of a deepbranch and B. K. gable, Eds.). SinauerAssociates, Sun- in the genealogicaltree for perchingbirds. Pro- derland, Massachusetts. ceedingsof the RoyalSociety of LondonSeries B KISHINO, H., AND M. HASEGAWA.1989. Evaluation of 243:99-107. the maximum likelihood estimate of the evolu- FELSENSTEIN,J. 1995. PHYLIP (PhylogeneticInfer- tionary tree topologies from DNA sequence ence Package),version 3.56. Department of Ge- data, and the branchingorder in Hominoidea. netics,University of Washington,Seattle. Journal of Molecular Evolution 29:170-179. FJELDS,i.,J. 1992. Biogeographic patterns and evolu- KLICKA,J., AND R. M. ZINK. 1997.The importanceof tion of the avifaunaof relicthigh-altitude wood- recentice agesin speciation:A failed paradigm. landsof the Andes.Steenstrupia 18:9-62. Science 277:1666-1669. FJELDS,i•,J. 1995. Geographicalpatterns of neoen- KOCHER,T. D., W. K. THOMAS, A. MEYER, S. V. ED- demicand older relict speciesof Andeanbirds: WARDS,S. PJkJkBO,S. X. VILLAFRANCA,AND A. C. The significanceof ecologicalstability areas. WILSON.1989. Dynamics of mitochondrialDNA Pages89-102 in Biodiversityand conservationof evolution in : Amplification and se- Neotropicalmontane forests (S. P. Churchill,H. quencingwith conservedprimers. Proceedings July1999] PhylogenyofMetallura Hummingbirds 711

of the National Academy of SciencesUSA 86: RoY, M. S., J. M. C. SILVA, P. ARCTANDER,J. GARCiA- 6196-6200. MORENO,AND J. FJELDS.&. 1997. The speciationof KUMAR, S., K. TAMURA, AND M. NEI. 1993. MEGA: South American and African birds in montane Molecular Evolutionary Genetic Analysis, ver- regions.Pages 321-339 in Avian molecularevo- sion 1.01. PennsylvaniaState University, Penn- lution and systematics(D. E Mindell, Ed.). Ac- sylvania. ademicPress, San Diego. LOEWE,L., AND S. SCHERER.1997. Mitochondrial Eve: SCHUCHMANN,K.-L., ANDM. HEINDL.1997. Biogeo- The plot thickens.Trends in Ecologyand Evo- graphie, geographischeVariation und Taxon- lution 12:422-423. omie der andinen KolibrigattungChalcostigma MADDISON, W. P., AND D. R. MADDISON. 1992. Reichenbach,1854. Mitteilungen Zoologischer MacClade:Analysis of phylogenyand character Museum Berlin 73 SupplementumAnnales Orn- evolution. Sinauer Associates, Sunderland, Mas- nitologici21:131-153. sachusetts. SERVANT, M., J. MALEY, B. TURCQ, M.-L. ABSY, P. MOORE,R. T. 1940. The nomenclature and habits of BRENAC,M. FOURNIER,AND M.-P. LEDRU. 1993. the Black-throatedCoppertailed . Tropicalforest changes during the LateQuater- Condor 42:251-254. nary in African and South American lowlands. NUNN, G. B., ANDJ. CRACRAFT.1996. Phylogenetic Globaland PlanetaryChange 7:25-40. relationshipsamong the major lineagesof the STOTZ,D. E, J. W. FITZPATRICK,T. A. PARKER,AND D. -of-Paradise (Paradisaeidae) using mito- K. MOSKOVITS.1996. Neotropical birds. Ecology chondrialDNA genesequences. Molecular Phy- and conservation.University of ChicagoPress, logeneticsand Evolution5:445-459. Chicago. O'NEILL,J.P. 1992.A generaloverview of the mon- SWOFFORD,D. L. 1991.PAUP: Phylogenetic analysis fane avifauna of Peru. Memorias del Museo de usingparsimony. Center for Biodiversityof the Historia Natural de la Universidad Nacional Illinois Natural History Survey,Champaign. Mayor de San Marcos (Lima) 21:47-55. SWOFFORD,D. L., G. J. OLSEN, P. J. WADELL, AND D. PATERSON,H. E. H. 1985.The recognitionconcept of M. HILLIS.1996. Phylogenetic inference. Pages species.Pages 21-29 in Speciesand speciation 407-514 in Molecular systematics(D. Hillis, C. (E. S. Vrba, Ed.). Transvaal Museum, Pretoria. Moritz, and B. K. Mable, Eds.). Sinauer Associ- POULSEN,B. O., ANDN. KRABBE.1997. The diversity ates, Sunderland, Massachusetts. of cloud forest birds on the eastern and western VRBA, E. S. 1993. Mammal evolution in the African slopesof the EcuadorianAndes: A latitudinal Neogeneand a new look at the greatAmerican and comparativeanalysis with implicationsfor interchange.Pages 393-432 in Biologicalrela- conservation.Ecography 20:475-482. tionshipsbetween Africa and SouthAmerica (P. RAHBEK,C. 1997.The relationshipamong area, ele- Goldblatt, Ed.). Yale University Press,New Ha- vation,and regional species richness in Neotrop- ven, Connecticut. ical birds. American Naturalist 149:875-902. YANG,Z. 1996.Among-site rate variationand its im- RASMUSSEN,J. F., AND C. RAHBEK. 1994. Aves del pact on phylogeneticanalyses. Trends in Ecol- ParqueNacional Podocarpus-Una Lista Anota- ogy and Evolution11:367-372. da. Birds of PodocarpusNational Park. An an- ZIMMER, J. T. 1952. Studies of Peruvian birds, no. 62. notated checklist. CECIA, Quito, Ecuador. The generaPatagona, Sappho, Polyonymus, Ram- REMSEN,J.V. 1984.High incidenceof "leapfrog"pat- phomicron,Metallura, Chalcostigma,Taphrolesbia tern of geographicvariation in Andean birds: and Aglaiocercus.American Museum Novitates No. 1595. Implicationsfor the speciationprocess. Science 224:171-173. AssociateEditor: A. J. Baker