The Auk 121(2):318-332, 2004
PHYLOGENY AND BIOGEOGRAPHY OF THE AMAZONA OCHROCEPHALA (AVES: PSITTACIDAE) COMPLEX
JESSICA R. EBERHARD^ AND ELDREDGE BERMINGHAM Smithsonian Tropical Research Institute, Apartado 2072, Balboa, República de Panamá
ABSTRACT.•We present a phylogenetic analysis of relationships among members of the Amazona ochrocephala species complex of parrots, a broadly distributed group in Middle and South America that has been a "taxonomic headache." Mitochondrial DNA sequence data are used to infer phylogenetic relationships among most of the named subspecies in the complex. Sequence-based phylogenies show that Middle American subspecies included in the analysis are reciprocally monophyletic, but subspecies described for South America do not reflect pat- terns of genetic variation. Samples from the lower Amazon cluster with samples collected in western Amazonia•not with samples from Colombia and Venezuela, as was predicted by subspecies classification. All subspecies of the complex are more closely related to one another than to other Amazona species, and division of the complex into three species {A. ochrocephala, A. auropalliata, and A. oratrix) is not supported by our data. Divergence-date estimates suggest that these parrots arrived in Middle America after the Panama land-bridge formed, and then expanded and diversified rapidly. As in Middle America, diversification of the group in South America occurred during the Pleistocene, possibly driven by changes in distribution of forest habitat. Received 20 January 2003, accepted 3 December 2003.
RESUMEN.•Presentamos un análisis de las relaciones filogenéticas entre miembros del complejo de loros Amazona ochrocephala, un grupo ampliamente distribuido en Mesoamérica y Suramérica, y que ha sido un "dolor de cabeza taxonómico." Utilizamos secuencias de ADN mitocondrial para reconstruir la relaciones filogenéticas entre la mayoría de las subespecies nombradas del complejo. Las filogenias basadas en estas secuencias muestran que las sub- especies mesoamericanas incluidas en el análisis son recíprocamente monofilétícas, pero las subespecies descritas para Suramérica no reflejan patrones de variación genética. Muestras de la baja Amazonia se agrupan con muestras de la Amazonia occidental, en vez de agruparse con las muestras de Colombia y Venezuela, como se esperaba con base en la clasificación actual de subespecies. Todas las subespecies del complejo están estrechamente relacionadas entre sí, separadas por distancias menores que las distancias entre miembros del complejo y otras especies de Amazona, y la división del complejo en tres especies (A. ochrocephala, A. auropalliata, y A. oratrix) no es apoyada por nuestros datos. Las fechas de divergencia estimadas con los da- tos moleculares sugieren que estos loros llegaron a Mesoamérica después de la formación del istmo de Panamá y luego expandieron su distribución y se diversificaron rápidamente. Como en Mesoamérica, la diversificación del grupo en Suramérica occurió durante el Pleistoceno, posiblemente como resultado de cambios en la distribución de habitats forestales.
BioGEOGRAPHicAL STUDIES OF Ncotropical of tropícal flora and fauna. Other important birds have been of central importance in devel- vicariant models of speciation that have been opment of models aimed at explaining the high supported by studies of the diversification of species diversity of the Neotropics (e.g. Cracraft Neotropical avifauna include the Andean up- 1985, Haffer 1985). An obvious example is the lift (Chapman 1917, Cracraft and Prum 1988), Forest Refuge hypothesis initially outlined by formation of river systems in the Amazon Haffer (1969, 1974), which suggests that isola- basin (Wallace 1853, Simpson and Haffer 1978, tion of remnant patches of rainforest during Capparella 1988), formation of the Panama land dry glacial periods fostered diversification bridge (Cracraff 1985, Cracraff and Prum 1988), and marine incursions (Nores 1999). It is within this rich ornithological tradi- Tresent address: Department of Biological Sciences tio^ that we present results of a phylogenetic and Museum of Natural Science, 202 Life Sciences, analysis of the Amazona ochrocephala complex Louisiana State University Baton Rouge, Louisiana of parrots, a group of biogeographic interest 70803, USA. E-mail: [email protected] because of its broad Neotropical distribution. 318 April 2004] Amazona ochrocephala Phytogeny 319
A historical perspective on relationships among Lousada and Howell 1996). Some taxonomists these parrots serves as a baseline from which to have considered the entire complex to con- study the interaction of history and ecology that stitute a single species, A. ochrocephala (e.g. has led to their contemporary diversity and dis- Monroe and Howell 1966, Forshaw 1989), but tribution. Furthermore, a molecular systematic others (e.g. Sibley and Monroe 1990, American analysis of the group is of taxonomic interest, Ornithologists' Union 1998, Juniper and Parr because classification of the complex using mor- 1998) divide the complex into three species: the phological characters has been a "taxonomic Yellow-crowned Amazon {A. ochrocephala [A. o. headache" (Howell and Webb 1995). Finally, ochrocephala, A. o. xantholaema, A. o. nattereri, and identification of conservation units, particularly A. 0. panamensis]); the Yellow-naped Amazon of the Mesoamerican subspecies, is important (A. auropalliata [A. a. auropalliata, A. a. parvipes, because their populations have suffered pre- and A. a. caribaea]); and the Yellow-headed cipitous declines due to habitat loss and the pet Amazon (A. oratrix [A. o. oratrix, A. o. tresmariae, trade (Collar et al. 1994). A. 0. belizensis, and A. o. hondurensis]). Two other Study species.•The A. ochrocephala complex races of oratrix are mentioned in the literature: includes eleven named subspecies that are "magna," from the Caribbean slope of Mexico, distributed from Mexico to the Amazon basin is not considered valid (Juniper and Parr 1998); (Fig. 1). Characters used to identify the various and "guatemalensis" has not been formally de- subspecies include plumage (in particular, ex- scribed and is included in belizensis by Juniper tent and position of yellow on head and thighs, and Parr (1998). and coloration at the bend of the wing), bill Parrots of the ochrocephala complex are gener- and foot pigmentation, and body size (Monroe ally found below 750 m (Forshaw 1989, Juniper and Howell 1966, Forshaw 1989, Juniper and and Parr 1998), inhabiting deciduous woodland, Parr 1998). However, those characters can vary gallery forest, savannah woodland, dry forest, significantly, even among individuals from the secondary growth along major rivers, and sea- same locality (Howell and Webb 1995, Lousada sonally flooded forests (Forshaw 1989, Juniper and Howell 1996, Juniper and Parr 1998, J. R. and Parr 1998). In Middle America, most of Eberhard pers. obs.), in part because of age- the subspecies appear to be allopatric, though related variation (Howell and Webb 1995, the biological barriers (if any) that separate the ranges are not obvious (Juniper and Parr 1998). A zone of contact among yellow-headed, yel- low-naped, and yellow-crowned forms may oc- belizensis cur along the Atlantic slope of Central America from Belize to Nicaragua (Lousada and Howell 1996). Unfortunately, free-flying birds are very rare in that region, and we were unable to secure representative samples for inclusion here. Given currently available information on distribution of ochrocephala subspecies in South America, no range discontinuity is known between A. o. ochrocephala (eastern Colombia, Venezuela, Trinidad, Guianas, and northern Brazil) and A. o. nattereri (southern Colombia, eastern Ecuador and Peru, western Brazil, and northern Bolivia) (Juniper and Parr 1998). The range of A. o. panamensis is mostly separated from other yellow-crowned forms by the Andes, though it may be continuous with A. o. ochro- FIG. 1. Distribution of the Amazona ochrocephala com- plex (after Juniper and Parr 1998); taxa sampled for the cephala in northwestern Venezuela (Juniper and present study are indicated in bold type. Distribution Parr 1998). of A. aestiva is outlined with the dashed line. Sample Four congeneric species were included as locations are indicated by points, and are numbered outgroups in our phylogenetic analyses: A. aes- to correspond with the sample listing in Table 1. tiva, A. amazónica, A. farinosa, and A. autumnalis. 320 EBERHARD AND BERMINGHAM [Auk, Vol. 121
Two of those species, A. aestiva and A. amazónica, strand, respectively. The primers C02GQL (GGACA- are consistently listed next to the A. ochrocephala ATGCTCAGAAATCTGCGG, L8929) and C03HMH complex in linear taxonomies (Forshaw 1989, (CATGGGCTGGGGTCRACTATGTG, H9947) were Sibley and Monroe 1990, Juniper and Parr 1998). used to amplify a 1,074-bp fragment that included the full ATPaseó and ATPase8 genes; along with those, The other two were included because samples the internal A6PWL (CCTGAACCTGACCATGAAC, were obtained opportunistically. An additional L9245) was used for sequencing the fragment. For outgroup species, A. barbadensis, was included one sample (a molted feather stored at room tem- in analysis of cytochrome oxidase I (COI) se- perature for about eight years), the ATPase region was quences. That species was of particular interest amplified in two overlapping pieces, using C02GQL because•like members of the ochrocephala com- with A6VALH (AGAATTAGGGCTCATTTGTGRC, plex, A. aestiva, and A. amazónica•it is character- H9436), and A6PWL with C03HMH. The ND2 gene ized by yellow plumage on parts of the head. was amplified with primer pairs METB (CGAAA- ATGATGGTTTAACCCCTTCC, L5233) and TRPC (CGGACTTTAGCAGAAACTAAGAG, H6343), and METHODS METB with ND2LSH(GGAGGTAGAAGAATAGGCY- TAG, H6102). The COI fragment was amplified using Samples.•Where possible, we used vouchered tis- primers COIa and COIf (Palumbi 1996), and the cyt-b sue samples from museum frozen-tissue collections. fragment was amplified and sequenced using primers However, because the ochrocephala complex is poorly CBl and CB3 (Palumbi 1996). Except for those involv- represented in those collections, many of the samples ing museum skins, PCRs were done using AmpIiTaq were obtained from field workers, captive breeding (Perkin-EImer, Wellesley, Massachusetts) and five cy- facilities, and pet owners (Table 1). Use of material cles with an annealing temperature of 50°C followed from captive birds was contingent on availability of in- by 30 cycles at 56°C. formation on the sampled individuals' geographic ori- A set of additional COI primers was designed gins. Source material included frozen tissues (muscle, to amplify and sequence a series of five overlap- liver), blood, feathers (both emerging and full-grown), ping fragments ranging in size from 106 to 196 bp. and small pieces of museum skins (toe and body skin). Sequences (5' to 3') of those primers, named ac- Three of the outgroup samples (representing A. aestiva cording to the position of the primer's 5' end, are and A. barbadensis) were from captive birds of unknown as follows: L7506 (TAGGGTTYATCGTATGGGCC), origin. Collection locations for the ochrocephala samples H7523 (ACTGTGAATATGTGGTGGGC), L7628 included here are shown in Figure 1. (GACTCGCCACACTACACGG), H7642 (CTCATTT- Laboratory procedures,•Toi A cellular DNA extrac- GATGGTCCCTCCG), L7773 (GTCTCACAGGRATC- tions from frozen tissue, blood, and feather samples GTCC), H7813 (GTATGTGTCGTGTAGGGCA), were done by incubating samples overnight in CTAB L7804 (AATAGGTGCCGTCTTTGCC), and H7879 buffer (Murray and Thompson 1980) and proteinase K, (G AATAGGGGG AATCAGTGGG). followed by a standard phenol-chloroform extraction That primer set was used in conjunction with prim- and dialysis. Extractions from museum skin samples ers COIa and COIf to amplify and sequence DNA ex- were done using the Qiamp kit (Qiagen, Valencia, tracted from museum skin samples. Polymerase chain California), following the protocol outlined by Mundy reaction amplifications of museum skin extracts were et al. (1997). Three mitochondrial DNA (mtDNA) done using AmpIiTaq Gold (Perkin-Elmer) in 25-|al re- fragments•the complete ATP synthase 6 and 8 genes actions and 40 cycles with an annealing temperature of (ATPase6,8), a 622-bp portion of COI, and the complete 60°C. Those reactions were set up in a UV hood to avoid NADH dehydrogenase 2 (ND2) gene•were amplified contamination. via the polymerase chain reaction (PCR) for most sam- Amplification products were visualized in agarose ples; only COI was analyzed for museum skin samples. gels, and then cleaned and purified using GELase In addition, a 694-bp fragment of the cytochrome-b (cyt (Epicentre Technologies, Madison, Wisconsin) fol- h) gene was sequenced for a subset of the samples (see lowing the manufacturer's protocol. PCR fragments Table 1) selected to include one member of each of the were then sequenced using either Dyedeoxy or dRho- clades identified using the other three coding regions. damine (Applied Biosystems, Foster City, California; To amplify and sequence the ATPaseó, ATPase8, Perkin-Elmer) cycle sequencing reactions and an ABI and ND2 genes, we used primers originally designed 377 automated sequencer. Amplification primers were by G. Seutin for studies of Neotropical passerine birds. used for sequencing both the heavy and light strands Primer sequences are given (5' to 3') followed by the of PCR fragments, and an additional internal primer, base position of the primer's 3' base relative to the A6PWL, was used to sequence the ATPase region. domestic chicken's (Gallus gallus) mtDNA sequence Three samples•two A. o. ochrocephala samples (Desjardins and Moráis 1990); the H or L indicates from Venezuela and the A. barbadensis sample•were whether the primer is located on the heavy or light sequenced by M. Rusello (Columbia University, New April 2004] Amazona ochrocephala Phytogeny 321
-Tí "C: f3 o ^-1 -ü iJ " 'S a * * ¿s. CJ CJ A ^ü * CJ (N CN CN (N CN CN CN CN CN CN CN CN (N CN Û Q Û O û Q ú û û Q û Q Q Q û û Q û û Q z z z z z Z A Z Z Z Z Z Z 2 2 Z 2 2 2 2 .S 3 .^ QJ K4" K-Í bc 3 O- O- O s f) O o o o o o () S S S 3 S S S S p¿ S u u u u u u u u u u u U u U U U U U U U cx5" ce 00" cxí CX) 00 00 00 00 00 oo oo"" oo"" 0(5" oo"" oo"" 0(5" 00" « c < < < « • < < < < < < < < < < < < < < < 3 C o < O OJ vo" ^íf ^"^ O O O ^i" ^0 \D \Û ^.O ^ ^O \0 '•^ \0 \0 '•^ \0 '^ "^ •_rt to Q 'S < < < U U U < < < <; <; < < <; < <<<<<<< u $ II o 'S Z 2 -^ C CD =i s¡ ,'S QJ QJ CO Si p c "DH Jí 3 T3 ^3 T3 T3 ^3 T3 T3 oí .y (O ~-^ •^3 3 0000 000 p O) CO -^ -HH 0000 000 QJ bcO QJ ^ J< eu ^^-.^^^-1 QJ QJ o c/) "^ PH CT) C/^ CO H S S S S 3 tu tu COCOCOCOCOfetiH 7 ^Ä S ^ o QJ 3 3 bo bO Z cu -ü o § 3 S O 01 QJ X x O o o c C/3 §< 2 o •^ 9 S -3 ^ Pi QJ S •T3 o .£ QJ 3 o •T3 II s 3 "S CT^i/^ en ^ >, s 3 cas - bO .2 0 iS H ni /rt DH JO en e/i CIJ ^-' O O 3 ^ C 3 • CN 3 o bo CO [Î3 CTÎ a z •T3 -a •o Ë « QJ N ,QJ QJ CO CO CO 1/1 c/) C¿ DHPH > QJ 3 U o 3 1/1 CO CO MJ QJ -^ rí ^ QJ O > 3 _o _o _o N -O H (u >- O o c •3 "o "o ^ ^ O QJ ' Ë ^;B Y; bo *H 3 .S CO CO 3 ^ 7^ u U u u U CO CO QJ ^ a > U HJ 3 s S) (N CO O Î! en en en en en O O K) bo VH O DH DH QJ m •3 P< CO N ^ 2á o O O O 3 QJ 0? > «3-0 j a a QJ QJ tó > B 2 CO CO QJ •^ QJ QJ m OH hJ CO PH r/î" r/ï~ r/î" c/i y ^ ^ O tH O ffl n3 rrî 03 cSt E^ P •fí CO ;3 y:i OH 0. 0- OH Í. rrí O O CO DH N tí Tí Z 3^ s '3 '3 "3 • f- IJ^ -H^ •^. •^. -H^ S 0 a er er er 3 S CO "S ^ ^ ^ DH DH CO o _ \0 Jn ^QJ 3 3 3 S 2 Ë " < QJ QJ Tí CO CO i|bC S •o a Ä QJ .;H- 3 ^ II U 3 2 o Ci tH« «VH 'n^ D D X -3 X a a a a 3 il S CO ^ Al >ä Z V-i CO s -^ CO ej: QJ U U U 2:a rö COCIS .3 u^ .a o" tS °o 1/1 DHCO H > > •a CO PL. QJ ^ -O n fö 03 tc QJ CO CO CCS ^ rë _eO 0000000 S g a a a •i; CJ CJ CJ CJ CJ CJ CJ CJ CO J¡ ;3 -S QJ c/l QJ .3 > ^ fö 03 tc QJ 3 O III 3 3 3 "o "o lo O U O TÍH O CO CO CO y ta U U > CO CO (^ PH PH S SS2SS2S ^1 QJ o o QJ QJ S c5 ^ 00 t3N a-c a ^0 Sí fo ó; c» ,a i I^> o X X X X X X X X X X X X c/3 a: QJ ;ií 3 a) c-i vt> K 3 c^ •tS QJ r-> c-i CO CO 1•lyiyiytiyuyi yjyiyiyDy^ytiyDy: Ç O) •^ bo ?3~ t! 'f. « 3 "O o QJ üT; >¡ 'S ^ o QJ c« O S S DH c)5 Pí VH o , 3 tj Z Z »-H y^ g Z QJ ^ Í3 P!Í P!Í P£ P¿ P!Í p!Í P¿ P Ü o -Q Ü (N CN (N (N CN (N ri CN (N (N oí CM" •T3 O Q Û O o Q û o Q Q û Q û û O C z z z z Z 2 Z Z 2 Z Z z o s 3 3 O o o o o 3 3 3 3 3 3 a- u u u u u u u u u u u u u u 00 00" 00 00 00 oo"" 00 00 oo"" oo" oo" oo" oo" oo" < < < < < < < < < < < < < ^ ^"" ^ ^ ^ ^ \£; ^ ^ \o vo" OJ QJ 0) OJ QJ 4= r^ 4= 4= ^3 13 T3 •a •73 -a O O O í 3 o o o 03 (Ö fö fö o O° ^fö QJQJQJQJ'•^'•^1•^1•P QI ^• I QJ.^,•I • QJ CO CO pH 3 s en fi o c QJ U S u 3 16 CT 1/3 d) QJ OJ (I) QJ -fi Tï cj y U 03 u U ^ Ö b -2 S S2 S S « o O fÖ (Ö fÖ fÖ (Ö Ë 1/3 ^ M g u ^i^ ''u ~i^ ^ÍH ^C < s fÖ (Ö fÖ fÖ (Ö S O 13 QJ o o o o .1-! ^ fÖ (í fÖ fÖ (í JD CO l/J W: 1/3 l/J W: QJ T3 T3 -a ,> _> ,> Q > 3 OJ Û, DH m S3 (C (Ö fÖ fÖ fÖ (Ö fÖ fÖ (Ö (T3 fö fö X^ o o o o o o o o Z Z Z Z Z c S - N N N N Ö Ö Ö Ö Ö o o o y u u y u g C o .S 03 03 < iï O Oí QJ OJ QJ MM>^MM c ^ "§3 o •? C 2 p^ QJ QJ •â s s Ë ï 'S 2 u 5 -a xxxx xxxxx 03 • .2 3 S 03" 1/JW31/31/) l/)W31/3l/)W3 W3p3l/) C S PLH K^ Z « ai p¿ Di ai Pi Pi Pi Pi Pi Cií 5 Pi ci Pi H H H H H H H H H H s H H H - i c/^ c/^ en c/^ en en en en en CO z cß en en ÎÉ fi :3 QJ [Q W3 1/3 1/3 S D ^ % _N _N _N _N T-H CN CD ':í^ m QJ QJ QJ l/J W3 1/3 l/J W3 (Î fö fö ° "^ QJ QJ QJ QJ QJ QJ QJ QJ QJ ^ Í TJ JD ^ ^ JD < < < •g O -° e 8- o u 03 •à 9- c •a í\) ÍIJ (y 3 c C C C C Ë e r• 1J g g S S £ 1-* r-# c M M f^ •2? o "^ "^ "^ 5^ S'' •S •5 u CM P P to !« i2 QJ . 'S 'S QJ •£ -5^ -< •^i li -îi •^i -s: -lî -s: -lî ^ ^ ^ ^ ^ ^ ^ II S _ë QJ- ^ ^ ^O ^ C^ K IN ir^ K April 2004] Amazona ochrocephala Phytogeny 323 York), using museum skin samples from the American and random branch addition. Neighbor-joining trees Museum of Natural History. Those sequences were were obtained using Tamura-Nei distances (Tamura obtained using the primer set listed above. and Nei 1993). To obtain an independent molecule-based estimate Substitution model parameters for ML analy- of the relationship between A. aestiva and the ochro- ses in PAUP* were found using MODELTEST 3.1 cephala complex (see below), a nuclear intron fragment (Posada and Crandall 1998), which uses hierarchical from the glyceraldehyde-3-phosphate dehydrogenase likelihood-ratio tests to compare the fit of different (Gapdh) gene was sequenced for a subset of samples. nested models of DNA substitution to the data ma- The nuclear sequence for A. aestiva was obtained us- trix. For the ATPase+COI+ND2 data set, MODELTEST ing a DNA extract taken by P. Wainright from a blood supported the Tamura-Nei model with I = 0.7666 and sample (A.aestS; attempts to obtain nuclear sequences equal rates at all variable sites. To reduce computing using extracts of A.aestl and A.aest2 were unsuccess- time, the ML analyses of ATPase+COI+ND2 in PAUP* ful). Primers GapdL890 and GapdH950 (Friesen et were done using a reduced data set of 18 taxa (see al. 1997) were used for amplification and sequencing Table 1) that included representatives of all major of the nuclear fragment. Polymerase chain reactions clades identified by MP and NJ analyses. For the COI were done using AmpliTaq or TaqCold (Perkin-Elmer), data set, the best fit found by MODELTEST was an beginning with 5 min at 94°C, and then five cycles with HKY model with a Ti:Tv ratio of 30.9904, the propor- an annealing temperature of 50°C followed by 30 cycles tion of invariable sites set to 0.6426, and a gamma at 56°C. In some cases, the initial PCR product had to be shape parameter of 0.4975. Those parameters were re-amplified (30 cycles at 56°C) prior to sequencing. specified in PAUP* for heuristic ML tree searches and All sequences have been deposited in GenBank, un- bootstrapping analyses. der accession numbers AY194295-AY194327 (ATPaseó), For the Bayesian Markov chain Monte Carlo AY194328-AY194360 (ATPase8), AY194367-AY194403 (MCMC) searches, a general time-reversible model (COI), AY194434-AY194466 (ND2), AY194404-AY194413 was specified, with site-specific variation partitioned (cyt h), and AY194425-AY194433 (Gapdh). by codon position. Four chains were run for 500,000 Sequence anaii/s/s. •Sequences generated by the generations and sampled every 1,000 generations. In automated sequencer were aligned and proofread the ATPase+COI+ND2 analysis, because stationarity using SEQUENCHER (version 3.1.1; GeneCodes, Ann was reached by 15,000 generations, the first 20,000 Arbor, Michigan). The ATPase6,8, COI, and ND2 se- generations were discarded, and the remaining trees quences were then concatenated for most subsequent were used to obtain a majority-rule consensus. For phylogenetic analyses, which were done using PAUP* analysis of the COI data set, trees from the first 35,000 (version 4.0b8; Swofford 1999). Sequences were com- generations were discarded prior to generating the bined because the mitochondrial gene regions are consensus tree. fully linked and thus represent a single phylogenetic Nodal support was assessed by bootstrap analysis marker (a partition-homogeneity test showed that the in the MP, NJ, and ML analyses (1,000, 1,000, and 125 gene regions were not significantly heterogeneous [P > replicates, respectively), and by posterior probabili- 0.50]). The PAUP* and SEQUENCER5.0 programs (see ties in the Bayesian analyses. Posterior probabilities Acknowledgments) were used to calculate descriptive indicate percentage of the time that a given clade statistics about nucleotide variation. Analyses that occurs among trees sampled in the Bayesian analyses included the museum skin specimens were based on (Huelsenbeck and Ronquist 2001). COI sequences, because we attempted amplifications Divergence times among clades in the ocfirocepfiata of that gene region only from the DNA extracted from complex were estimated using two different molecular skin samples. The cyt-b sequences, which were ob- clock calibrations (no specific calibration exists for the tained for a subset of the samples, were used only for Psittaciformes). A2% sequence divergence per million estimates of divergence dates (see below). years (my) calibration (based on restriction-site varia- Phylogenies were reconstructed using neigh- tion across the mitochondrial genome; see Shields and bor-joining (NJ), maximum-parsimony (MP), and Wilson 1987, Tarr and Fleischer 1993) was used with maximum-likelihood (ML) algorithms in PAUP*, and the ATPase+COI+ND2 data set. Another set of diver- a Bayesian approach as implemented in MRBAYES gence time estimates was calculated using the parrot (Huelsenbeck and Ronquist 2001). Outgroup root- cyt-b data and a fossil-based molecular clock calibra- ing was used to root trees. The MP and NJ analyses tion for cyt b in cranes (Krajewski and King 1996). That were done with all characters weighted equally; MP second calibration uses cyt-b maximum-likelihood searches were also done with weights of 1 and 18 distances (as calculated using the DNADIST pro- assigned to transitions (Ti) and transversions (Tv), gram in PHYLIP; Felsenstein 1995), which in cranes respectively. For analysis of COI sequences, the Ti: diverge by 0.7%-1.7% my^^. For both of those data Tv weighting was 1:17. Those weightings reflect the sets, the assumption of clock-like sequence change Ti:Tv ratios determined empirically from the data. was first tested by using a likelihood ratio test (LRT; Parsimony trees were found using heuristic searches Felsenstein 1988) to compare likelihood scores of ML 324 EBERHARD AND BERMINGHAM [Auk, Vol. 121 trees found by heuristic searches in PAUP* with a mo- TABLE 2. Nucleotide variability at different codon lecular clock enforced versus not enforced. The LRTs positions in the ATPaseö, ATPase8, COI, and ND2 for both the ATPase+COI+ND2 and the cyt-b data sets genes of Amazona. Both ingroup and outgroup showed no statistical difference between trees found species were included in the calculations. with or without a clock enforced (P = 0.2286 and P = 0.2032, respectively). Codon Number Percentage To determine whether our data support division of Region position Total bp variable variable the ochwcephala complex into three species, alternative ATPaseó all 684 100 14.6 tree topologies were compared using the nonparamet- 1 228 25 11.0 ric Shimodaira-Hasegawa (S-H) test (Shimodaira and 2 228 7 3.1 Hasegawa 1999) in PAUP*, using RELL bootstrapping. 3 228 68 29.9 As pointed out by Goldman et al. (2000), that test is ATPase8 all 168 29 17.3 applicable when one of the trees being compared is 1 56 7 12.5 one selected with reference to the same data being 2 56 5 8.9 used in the test. Using both the ATPase+COI+ND2 3 56 17 30.4 and the COI data sets, we compared the Bayesian tree COI all 622 95 15.3 with a test tree composed of three clades: an oratrix 1 208 7 7.4 clade (oratrix + belizensis + tresmariae), an auropalliata 2 207 0 0.0 clade, and an ochrocephala clade {panamensis + ochw- 3 207 88 92.6 cephala + nattereri + xantholaema). Within each of those ND2 all 1041 167 16.0 clades, relationships among subspecies were not 1 347 32 9.2 resolved; the A.ael and belizl samples were omit- 2 347 25 7.2 ted from the analysis (see below). The S-H test was 3 347 110 31.7 also used to compare the Gapdh tree obtained in the parsimony analysis with a tree in which the A. aestiva sample is forced within the ochrocephala clade. For changes occur at first position, and 11.6% at the that test, we used likelihood parameters suggested second codon position. Transitions outnumber by MODELTEST (Posada and Crandall 1998) for the transversions by 18.6 to 1, averaging over all Gapdh data (F81, with equal rates for all sites and no pairwise comparisons. In the COI data set, invariable sites). which includes the additional skin samples, 95 (15.3%) characters are variable, and 49 (7.9%) are parsimony informative. In that region, the RESULTS vast majority (92.6%) of changes occur at third position, and the remainder occur at first posi- A total of 2,515 bp of coding sequence tion. All base changes in COI are synonymous. (ATPaseó, 684 bp; ATPaseS, 168 bp; COI, 622 bp; In the cyt-b fragment, most changes (83.5%) oc- and ND2,1,041 bp) was obtained for each parrot cur at third position, with 15.2% at first position, individual, except museum skin specimens, for and the remaining 1.3% at second position. which only the COI fragment was sequenced. According to the phylogenies generated by We also sequenced a 694-bp fragment of cyt b all of the tree-reconstruction algorithms using for 1(] of the parrots (see Table 1). Overlap be- the ATPase-i-COI+ND2 sequences, the ochro- tween sequences generated using heavy- and cephala complex forms a well-supported clade light-strand primers averaged approximately (bootstrap values of 100% in MP, NJ, and ML 72% for the ATPase coding region, 86% for the analyses, and 100% posterior probability value COI fragment, 58% for the ND2 gene, and 90% in the Bayesian tree). Topology of the Bayesian for the cyt-b fragment; no nucleotide differences tree (Fig. 2) is nearly identical to that of the trees were found between overlapping complemen- found by PAUP* using parsimony, distance, tary sequences. and maximum-likelihood algorithms. The MP No indels or stop codons were observed, and NJ trees differ only slightly in the arrange- as expected for protein-coding mitochondrial ment of the South American clade that includes regions. In the ATPase+COI+ND2 data set, 378 samples A.aestl, ochrol, ochro2, and naterS. (15.0%) base positions are variable and 205 The Middle American subspecies {oratrix, tres- (8.2%) are parsimony-informative. Sequence mariae, belizensis, auropalliata, and panamensis) variability differed across codons (Table 2). consistently form reciprocally monophyletic Third-position changes are the most common clades that are strongly supported by bootstrap (49.5% of variable sites), whereas 38.9% of analysis, with bootstrap values >99% in the MP April 2004] Amazona ochrocephala Phytogeny 325 belizi (see Fig. 2). Individuals representing nattereri orate - orat7 (three localities in Bolivia), xantholaema (Ilha de 0 rats oratríx orat1 Marajó, at the mouth of the Amazon River), and ï orat2 ochrocephala (Xingu River, Brazil) intermingle to 1 oratS 1- orat4 form a well-supported clade that does not in- beliz2 beliz3 beiizensis clude the remaining ochrocephala sample, which beliz4 is from Colombia. The Colombian ochrocephala auropalliata - auro2 I sample falls at the base of the complex and is tresi - tres2 separated from other South American samples 1Q0| - tresS by almost 2% (uncorrected) sequence diver- - tres4 i-tresa gence. The separation of the Colombian ochro- 1% P panal •pana2 cephala sample from the other South American -10 changes pana3 samples was further investigated by sequencing pana4 octirol the COI fragment for three additional samples I ochrocephala 100 ochro2 taken from museum skins: one from a second A.aesti A.aestiva naterS i locality in Colombia, and two others from dif- \ naten nattereri lOQp nater2 ferent localities in Venezuela (see Table 1 and I - Yxanttil xantholaema -octiroS •* ochmcephala Fig. 1). The COI sequence was also obtained A.autumnalis for an A. barbadensis sample. That was done in A.farinosa - A. amazónica response to placement of A. aestiva within the ochrocephala complex (see below). Like A. aestiva, FIG. 2. Phylogram of the Amazona ochrocephala com- A. barbadensis has yellow plumage on the head plex obtained in a Bayesian analysis of the mitochon- that is similar in hue and extent to that found in drial ATPase6,8, COI, and ND2 sequences (2,514 bp) members of the ochrocephala complex. obtained in the present study. The tree represents a 50% majority-rule consensus of 479 trees generated Phylogenetic analysis of the COI sequences, by MRBAYES during an MCMC search (see text), and which includes the additional samples from numbers at nodes are posterior probabilities. The tree Colombia and Venezuela, again demonstrates was rooted using A, amazónica, A. autumnalis, and A. a clear phylogenetic separation of ochrocephala farinosa as outgroups. Current subspecies groupings samples from northern South America versus (e.g. Forshaw 1989, Juniper and Parr 1998) are indi- those from central South America (Fig. 3). The cated to the right of sample names. Colombian and Venezuelan samples form an mtDNA clade that is sister to the remaining and NJ trees, and >95% in the ML tree, and South American individuals from Brazil and with posterior probability values of 100% in Bolivia, with moderate (79%) support in the the Bayesian analysis. Down-weighting transi- Bayesian analysis (Fig. 3). Bootstrap values for tions in the MP and NJ analyses of the complete the same node were 88% in an NJ tree, 67% in data set resulted in trees with almost no reso- an MP tree, and 64% in an ML tree (trees not lution within the ochrocephala complex clade. shown). Overall, the COI tree (Fig. 3) is not as Transversion weighting of the COI data set also well resolved as one based on the full mitochon- resulted in trees with decreased resolution, but drial data set (Fig. 2), presumably because the no results were in conflict with tree topologies COI data set is much smaller. Together, the two found using the equally weighted data. A single trees indicate that parrots from the Amazon, beiizensis sample (belizi) consistently clustered northern South America, and Mesoamerica form with the oratrix samples, but because the beii- a polytomy uniting three lineages of equivalent zensis samples were from wild-caught captive evolutionary distinctiveness. The COI data also birds for which exact location data were not show that A. barbadensis is a distinct species, available, and because morphological differ- and not particularly closely related to the ochro- ences used to distinguish oratrix and beiizensis cephala complex. are fairly subtle, we hesitate to overinterpret The unexpected placement of A. aestiva that result, assuming instead that the bird was within the ochrocephala complex was confirmed misidentified. using a second extraction from a different The samples representing named subspe- feather of the same bird, and using a second cies from South America do not form clades sample (A.aest2) obtained from the NMNH (the 326 EBERHARD AND BERMINGHAM [Auk, Vol. 121 COI sequences of A.aestl and A.aest2 differed Gapdh data set, 28 are variable and only two are by a single base substitution). We are confident parsimony-informative; 2 bp changes separate that those results are not due to species mis- Amazona aestiva and the ochrocephala complex. identification, because A. aestiva can be unam- Parsimony and distance-based analyses of the biguously distinguished from A. ochrocephala by nuclear data produced trees with consistent to- the presence of blue plumage on the forehead pologies in which A. aestiva falls outside of the (species identity of A.aestl was confirmed with ochrocephala clade (parsimony tree shown in Fig. photographs, A.aest2 was identified by NMNH 4), but an S-H test shows that those topologies personnel, and A.aestS was identified by staff at are not significantly different from one in which the University of Georgia Veterinary School). A. aestiva is forced within the ochrocephala clade The relationship between A. aestiva and the (P = 0.15). ochrocephala group was explored further using Short internode distances uniting Middle the Gapdh nuclear sequences. A total of 404 bp American subspecies indicate a rapid geo- were sequenced for an A. aestiva sample, four graphic expansion across Mesoamerica and members of the ochrocephala complex (repre- recent diversification. The genetic distances senting belizensis, panamensis, ochrocephala, and (uncorrected p distance, calculated using the auropalliata), as well as four other Amazona ATPase+COI+ND2 data set) between individuals species (A. farinosa, A. amazónica, A. viridigena- of different named Mesoamerican subspecies in lis, and A. autumnalis). The Gapdh sequences the ochrocephala complex are small, ranging were easily aligned, with only a single one- from 0.007 {belizensis vs. auropalliata) to 0.016 nucleotide indel shared by the A. autumnalis and A. viridigenalis samples. Of the 404 bp in the beliz2 orat1 pana2 orat2 orat3 62 orat4 oratrix oratS orate ochrol arat7 belizi beliz2 beliz3 belizensis beliz4 auro1 61 'r-| tres1 72 tres2 tres3 tresmariae Islas Mara; A. farine (Mexico) - tres5 tres4 1 change panal pana2 panamensis Panama lom pana3 A.aestS pana4 p A.aestl A.aestiva ochrolI . . , ochrocephala I C Brazil r•^ I naterS A. amazónica • m L naterl "ittereri Bolivia [• nater2 L xanthi xanthoiaema "*• - E Brazil ?^ochro3 I Colombia A. vitidigenalis - ochro4 ocbrocephala 951ochroS I Venezuela - ochro6 - A.barbaäensis - A.farinosa A. autumnalis A. amazónica FIG. 4. Phylogeny showing the relationship between FIG. 3. Phylogram of the Amazona ochrocephala com- four members of the Amazona ochrocephala complex and plex obtained in a Bayesian analysis of the COI data- several Amazona species based on nuclear Gapdh se- set (622 bp). The tree represents a 50% majority-rule quences (404 bp). The tree shown is one of the two most consensus of 464 trees generated by MRBAYES during parsimonious trees found in an exhaustive maximum an MCMC search (see text), and numbers at nodes are parsimony search by PAUP*, and is identical to the posterior probabilities. The tree was rooted using A. strict consensus of the two most parsimonious trees. amazónica, A. autumnalis, A. barbadensis, and A.farinosa as The phylogeny was rooted using A. autumnalis as the outgroups. Current subspecies groupings (e.g. Forshaw outgroup; the same topology is obtained using mid- 1989, Jumper and Parr 1998) and geographic origin of point rooting. Numbers at the nodes indicate bootstrap samples are indicated to the right of sample names. values obtained in 1,000 bootstrap replicates. April 2004] Amazona ochrocephala Phytogeny 327 {panamensis vs. tresmariae). Genetic distances provide an informative counterpoint to the between Middle American and South American morphological characters that have previously ochrocephala subspecies average 0.014, whereas been used to classify members of the ochroceph- distances between members of the ochrocephala ala parrot complex. Consistent with taxonomic group and outgroup species range from 0.059 arrangements that group all of the subspecies {A. amazónica vs. nattereri) to 0.076 (A. farinosa under a single species name (e.g. Forshaw vs. nattereri). Assuming that Amazona mtDNA's 1989), sequence data indicate that members rate of sequence divergence is approximately of the complex are very closely related, and 2% Ma^^ as in a variety of other birds, such as much more closely related to one another than geese (Shields and Wilson 1987) and honey- to other surveyed Amazona species. Molecular creepers (Tarr and Fleischer 1993), the above phylogenetic analysis supports the monophyly distances indicate that lineages sampled within of named subspecies from Middle America the ochrocephala complex shared a common an- {oratrix, tresmariae, belizensis, auropalliata, and cestor 1.2 million years ago (mya). panamensis), but not of the South American Maximum-likelihood distances calculated ones {ochrocephala, nattereri, xantholaema). We using DNADIST and the cyt-b data range from note that our analysis did not include samples 0.003 {oratrix vs. auropalliata) to 0.039 (tresmariae from the Caribbean slope of Honduras and vs. xantholaema) within the ochrocephala com- Nicaragua, where there may be a contact zone plex, whereas distances to A. amazónica range between several different subspecies (Lousada from 0.088 {A. amazónica vs. belizensis) to 0.100 and Howell 1996). [A. amazónica vs. tresmariae). According to the Subdivision of the ochrocephala complex into crane calibration of Krajewski and King (1996), three species is not supported by phyloge- the cyt-b ML distances among taxa in the ochro- netic analysis of parrot mtDNA genes. In the cephala complex suggest that diversification of comparison of alternative tree topologies, the the group occurred during the past 0.2-5.6 Ma. ATPase+C01+ND2 and COI data sets rejected That interval is broad, but consistent with cal- the three-species topology that reflects division culations based on the ATPase+C01+ND2 data; of the complex into A. ochrocephala, A. auropal- taken together, estimates indicate that the ochro- liata, and A. oratrix. Although plumage char- cephala complex diversified recently, probably acters support subdivision of the ochrocephala within the past 2 Ma. complex into three species, plumage patterns The ATPase+C01+ND2 and COI data sets are quite variable and appear to be very labile. strongly support reclassification of the South For example, an examination of museum skins American ochrocephala subspecies. South at the American Museum of Natural History by American ochrocephala parrots separate along J.R.E. found an ochrocephala specimen (AMNH geographic rather than currently described sub- 133032) with a yellow feather on its nape, and species lines. The mean genetic distance between a nattereri skin (AMNH 255153) with a yellow the northern and central South American birds feather on its throat; in both of those subspe- that we have examined is 0.02 (uncorrected p dis- cies, yellow feathers are typically confined to tance), indicating a split 1.0 mya under the 2% di- the forehead and crown (Forshaw 1989, Juniper vergence per my calibration (the mean cyt-b ML and Parr 1998). Similarly, Monroe and Howell distance is 0.026, yielding an estimate of 1.5-3.7 (1966) note a well-documented instance of an mya, with the cyt b 0.7%-1.7% Ma^^ calibration). individual captive parrot that had a yellow- Results of the comparison of alternative tree crowned plumage pattern for 10 years, and topologies do not support division of the ochro- afterward developed a yellow nape in addition cephala complex into three species. According to to the crown. the S-H test, the three-species tree is rejected at Under the phylogenetic species concept the P < 0.001 level when compared to the log- (Cracraft 1983), the mtDNA sequence data likelihood of the Bayesian tree (Fig. 2). would support elevating the Middle American subspecies included here to species status, and DISCUSSION regrouping the South American taxa into two species, A. ochrocephala {ochrocephala from north- Molecular systematics of the ochrocephala ern South America) and A. nattereri {ochrocephala complex.•The molecular data presented here from Amazonia, plus the currently recognized 328 EBERHARD AND BERMINGHAM [Auk, Vol. 121 nattereri and xantholaema), representing the were analyzed separately, they recovered a mtDNA clades from northern and central South "yellow-headed" clade, which presumably in- America. Alternatively, the data support rec- cluded A. aestiva, though a tree is not shown. ognition of a single monophyletic species, A. The distributions of A. ochrocephala and A. aes- ochrocephala (with South American subspecies tiva are largely separate, although there is some revised as described here), which conforms to overlap in their ranges (see Fig. 1). That region the opinions of Monroe and Howell (1966) and of overlap could permit hybridization between Forshaw (1989). A. ochrocephala and A. aestiva, possibly leading Because of the low levels of nuclear sequence to differential mtDNA introgression, which divergence among members of the ochrocephala could explain the mitochondrial sequence complex, reciprocal monophyly of the mtDNA data. Such introgression has been reported in lineages that form the Middle American studies of a range of animal taxa (Ferris et al. subspecies included here is not sufficient to con- 1983, Powell 1983, Tegelstrom 1987, Dowling clusively identify them as evolutionarily signifi- et al. 1989, Lehman et al. 1991, Boyce et al. cant units (ESUs), following criteria proposed 1994, Quesada et al. 1995, Rohwer et al. 2001). by Moritz (1994). However, the morphological Although no hybridization between those spe- differences (e.g. yellow nape and lack of red at cies has been noted in the wild. Amazona species the bend of the wing in auropalliata, or extensive are known to hybridize in captivity (Nichols yellow on the head in oratrix and tresmariae) 1980). Another possible explanation is that mi- probably reflect meaningful divergences at nu- tochondrial primers preferentially amplified an clear loci, even if such variation alone is not di- ancestral mitochondrial pseudogene "frozen" agnostic for some of the subspecies. Regardless in the nuclear genome of the A. aestiva samples. of the taxonomic nomenclature adopted for However, that does not seem likely, given the the ochrocephala complex, the combination of size of the pseudogene•or the repeated nature mtDNA data and plumage variation among of translocation•required to explain the coin- the Mesoamerican members indicates that each cident pattern observed in the four different subspecies should be considered distinct units mitochondrial gene regions. for conservation purposes. Biogeography of the ochrocephala complex.• Analyses of the mitochondrial and nuclear Recent work by Rusello and Amato (2004) in- sequence data produce conflicting results with dicates that the ochrocephala complex arose from respect to the position of A. aestiva relative to a South American ancestor, and our phyloge- the ochrocephala complex. The Gapdh phylogeny netic hypothesis (Fig. 2) is consistent with their (Fig. 4), in which A. aestiva falls outside of the analysis. Two of the three principal ochrocephala ochrocephala clade, is in agreement with the mor- mtDNA clades are confined to South America, phological characters that can be used to distin- and the level of intraclade mtDNA divergence guish the two. However, low levels of variation observed in the Amazonian lineages (Brazilian in the Gapdh sequence data make it impossible and Bolivian parrots) is greater than the ge- to conclusively reject the hypothesis supported netic distances observed between individuals by mtDNA analysis, which places A. aestiva representing the relatively large and taxonomi- within the ochrocephala complex. Agreement be- cally diverse sample of Mesoamerican parrots. tween the Gapdh analysis and the morphologi- Failure to reject a molecular clock prompted cal distinctiveness of A. aestiva suggests to us our application of available avian mtDNA clock that the mtDNA data do not accurately reflect calibrations, which indicated that the three A. aestiva's phylogenetic history, but additional principal ochrocephala mtDNA clades formed nuclear sequence data are necessary to resolve contemporaneously in the Pliocene or early that issue. Our mtDNA data placing A. aestiva Pleistocene. within the ochrocephala clade are consistent with The short genetic distances among Middle those of Rusello and Amato (2004), who used American subspecies, and the close relationship both mitochondrial and nuclear sequences in between the Middle and South American lin- a phylogenetic analysis of Amazona. However, eages, suggests that Middle America was colo- they also found A. barbadensis to fall within nized by ochrocephala parrots well after the rise the ochrocephala complex, whereas we did not. of the isthmus of Panama 3.5 mya (Coates 1997). When Rusello and Amato's (2004) nuclear data Both the short internodes and the short terminal April 2004] Amazona ochrocephala Phytogeny 329 branch lengths of Middle American subspecies recognized as A. o. nattereri from western indicate that the area was colonized relatively Brazil, Bolivia, and Peru. Recent population- quickly and recently. A similar pattern of rapid genetic analysis of mahogany (Swietenia macro- expansion across the Mesoamerican landscape, phylla) distributed along the southern arc of the followed by in situ phylogenetic diversifica- Amazon also failed to demonstrate strong geo- tion, has been demonstrated for freshwater graphic subdivision in the region, in contrast to fish (Bermingham and Martin 1998, Perdices Mesoamerican Swietenia macrophylla, which was et al. 2002, G. Reeves and E. Bermingham un- strongly structured into four regional popula- publ. data) and howler monkeys (Cortes-Ortiz tion groups (Novick et al. 2003). The compelling et al. 2003). The branching order of the tree in data for geographically structured populations Figure 2 is consistent with a south-to-north step- of Middle American freshwater fish, howler ping-stone pattern of colonization for Middle monkeys, and mahogany may indicate a geo- America, a dispersal pattern that has been rec- graphic history of population expansion and ognized in a number of organisms (see Savage subdivision for many groups that is consider- 1982). ably more dynamic than that of Amazona par- The allopatry of Middle American subspe- rots. Differences in regional histories such as the cies, and their diversification, may be attribut- one documented here for South American ochro- able to habitat preferences. These parrots are cephala parrots versus their Middle American lowland birds, generally found in relatively counterparts might also provide a partial dry or deciduous forests, forest edges and gal- explanation for the reduction in beta diversity lery forest, and savannahs; in South America, (species turnover) that characterizes western they appear to avoid continuous moist forest, Amazonian rainforest tree communities in com- perhaps being replaced by A. amazónica in those parison to those in Panama (Condit et al. 2002). habitats (Juniper and Parr 1998). Expansion The phylogenetic break between ochroceph- of ochrocephala parrots across Middle America ala parrots from northern and central South may have occurred during glacial periods of America has not been suggested by previous the Pleistocene, when dry forest and savan- taxonomic work but is well supported by the nah vegetation were probably more continu- molecular data (see Figs. 2 and 3). The most ous over much of the region (Colinvaux 1997). obvious landscape feature that coincides with Subsequent warmer and wetter periods would divergence between the two South American have permitted an extension of wet forests lineages is the Amazon River, which runs be- (Colinvaux 1997), possibly leading to fragmen- tween the geographic areas represented by the tation of the drier habitat preferred by parrots samples in the two South American clades. This of the ochrocephala complex. In addition, G. is consistent with the riverine barrier hypothesis Reeves and E. Bermingham (unpubl. data) have (Wallace 1853, Capparella 1988), which argues produced a model suggesting that phylogenetic that large river courses impede gene flow be- breaks between lineages can be maintained in tween populations on opposite banks, leading the absence of discrete barriers to gene flow, to speciation. Nevertheless, our support for the owing to inertia resulting from behavioral inter- riverine barrier hypothesis is weak, and further actions (repulsion) or demographic interactions sampling on both the north and south banks resulting from differences in population sizes of along the Amazon River would permit a much resident and immigrant populations. stronger test of the hypothesis. Alternatively, The strong phylogeographic structure in the genetic break could reflect past habitat dis- Mesoamerican ochrocephala parrots stands in continuities, such as changes in forest cover re- contrast to the apparent lack of geographic sulting from climatological cycles (Haffer 1969) structure in parrots collected across a region or isolation of Guiana Shield populations due to extending from the mouth of the Amazon sea-level changes (Nores 1999). to Bolivia and Peru, a distance of >2,00ü km. Alternatively, the genetic break could reflect Contrary to current subspecies descriptions, past habitat discontinuities (e.g. changes in our results clearly show that parrots described forest cover resulting from Pleistocene glacial as A. 0. ochrocephala and A. o. xantholaema, from cycles; Haffer 1969), with the Amazon being Rio Xingu and Ilha de Marajó in the the lower a secondary barrier, halting the expansion of Amazon River, are closely allied with parrots lineages from their centers of origin. Observed 330 EBERHARD AND BERMINGHAM [Auk, Vol. 121 genetic distances between members of the (Belize Zoo); Loro Parque; S. Rodden; U.S. National two South American clades, which imply that Museum of Natural History; P. Wainright; M. J. West- divergences occurred during the Pleistocene, Eberhard; and T Wright. Tissue collection of A, [o.] would be consistent with that hypothesis. The ochrocephala samples on the Rio Xingu by G. R. Graves was supported by the Academia Brasileira de Ciencias, northern samples could also reflect isolation of through a grant from Electronorte administered by P. an ochrocephala lineage on the Guiana shield of E. Vanzolini, and by the Smithsonian's International northwestern South America, which may have Environmental Sciences Program Neotropical been isolated during ~100 m sea-level rises Lowland Research Program. Funding was provided by that occurred during the late Tertiary and the awards from the American Museum of Natural History Pleistocene (Nores 1999). Frank M. Chapman Memorial Fund and the American Of the three named South American subspe- Ornithologists' Union to J.R.E., a Smithsonian Tropical cies, xantholaema is the most different morpho- Research Institute postdoctoral fellowship to J.R.E., logically (more extensive yellow on the head), and the Smithsonian Institution's Molecular Evolution has vocalizations that differ from mainland Program. For information on the SEQUENCHER 5.0 program, see http://nmg.si.edu/Sequencer.html ochrocephala (C. Yamashita and P. Martuschelli pers. comm.), and is somewhat isolated on Ilha de Marajó. However, according to our sequence LITERATURE CITED data, xantholaema clusters closely with a nattereri AMERICAN ORNITHOLOGISTS' UNION. 1998. Check- sample from Bolivia. That lack of divergence list of North American Birds, 7th ed. American is also shown by phylogenetic analysis of se- Ornithologists' Union, Washington, D.C. quence data from the rapidly evolving control BERMINGHAM, E., AND A. P. MARTIN. 1998. region, using samples from the present study Comparative mtDNA phylogeography of (J. R. Eberhard and E. Bermingham unpubl. Neotropical freshwater fishes: Testing shared data) and additional samples from wild-caught history to infer the evolutionary landscape of xantholaema (C. Y. Miyaki pers. comm.). There lower Central America. Molecular Ecology 7: may be sufficient gene flow between island 499-517. and mainland populations to prevent genetic BOYCE, T. M., M. E. ZWICK, AND C. F. AQUADRO. divergence of xantholaema from its relatives. 1994. Mitochondrial DNA in the bark weevils: Phytogeny and evolution in the Pissodes strobi Alternatively, xantholaema may have diverged species group (Coleóptera: Curculionidae). too recently for mtDNA to be a useful marker, Molecular Biology and Evolution 11:183-194. but sufficiently long ago to permit divergence in CAPPARELLA, A. P. 1988. Genetic variation in plumage and vocalization patterns. Neotropical birds: Implications for the spe- ciation process. Pages 1658-1664 in Acta XIX ACKNOWLEDGMENTS Congressus Internationalis Ornithologici (H. Ouellet, Ed.). National Museum of Natural M. Rusello kindly provided three of the COI se- Sciences, Ottawa, Ontario. quences. We thank G. Amato, M. González, J.Groth, CHAPMAN, F. M. 1917. The distribution of bird-life J. Hunt, and I. Lovette for technical suggestions and in Colombia: A contribution to a biological sur- assistance, and M. Leone for assistance in process- vey of South America. Bulletin of the American ing our import permits. G. Seutin designed many Museum of Natural History, no. 36. of the primers used in the analysis of mitochondria! COATES, A. G. 1997. The forging of Central America. regions. We thank R. M. Zink and two anonymous Pages 1-37 in Central America: A Natural and reviewers for comments on an earlier version of this Cultural History (A. G. Coates, Ed.). Yale manuscript. Thanks also to the American Museum of University Press, New Haven, Connecticut. Natural History for permission to examine specimens CoLiNVAux, P. 1997. The history of forests on the in their collection, and to C. Miyaki for permission isthmus from the ice age to the present. Pages to cite unpublished data. This study would not have 123-136 in Central America: A Natural and been possible without the samples contributed by the Cultural History (A. G. Coates, Ed.). Yale following individuals and institutions: Academy of University Press, New Haven, Connecticut. Natural Sciences, Philadelphia; American Museum COLLAR, N. J., M. J. CROSBY, AND A. J. STATTERSFIELD. of Natural History; Collection of Genetic Resources 1994. Birds to Watch 2: The World List of at the Louisiana State University Museum of Natural Threatened Birds. BirdLife InternationaL Science; E. Enkerlin, J. J. Gonzalez, and Claudia Washington, D.C. Macias (Tecnológico de Monterrey, Mexico); H. de CoNDiT, R., N. PITMAN, E. G. LEIGH, J. CHAVE, J. Espinoza; Fundación ARA; O. Habet and S. Matóla TERBORGH, R. B. FOSTER, P. NUNEZ, S. AGUILAR, R. April 2004] Amazona ochrocephala Phytogeny 331 VALENCIA, G. VILLA, H. C. MULLER-LANDAU, E. Neotropical lowlands. Ornithological Mono- Losos, AND S. P. HUBBELL. 2002. Beta-diversity in graphs 36:113-146. tropical forest trees. Science 295:666-669. HowELL, S. N. G., AND S. WEBB. 1995. The Birds of CORTéS-ORTIZ, L., E. BERMINGHAM, C. RICO, E. Mexico and Northern Central America. Oxford RODRI'GUEZ-LUNA, I. SAMPAIO, AND M. RUIZ- University Press, New York. GARCI'A. 2003. Molecular systematics and HUELSENBECK, J. P., AND F. R. RONQUIST. 2001. biogeography of the Neotropical monkey MRBAYES: Bayesian inference of phylogeny. genus, Alouatta. Molecular Phylogenetics and Bioinformatics 17:754-755. Evolution 26:64-81. JUNIPER, T, AND M. PARR. 1998. Parrots: A Guide CRACRAFT, J. 1983. Species concepts and speciation to Parrots of the World. Yale University Press, analysis. Current Ornithology 1:159-187. New Haven, Connecticut. CRACRAFT, J. 1985. Historical biogeography and KRAJEWSKI, C, AND D. G. KING. 1996. Molecular patterns of differentiation within the South divergence and phylogeny: Rates and patterns American avifauna: Areas of endemism. of cytochrome b evolution in cranes. Molecular Ornithological Monographs 36:49-84. Biology and Evolution 13:21-30. CRACRAFT, J., AND R. O. PRUM. 1988. Patterns and LEHMAN, N., A. EISENHAWER, K. HANSEN, D. MECH, processes of diversification: Speciation and R. O. PETERSON, P. J. P. GOG AN, AND R. K. WAYNE. historical congruence in some Neotropical 1991. Introgression of coyote mitochondrial birds. Evolution 42:603-620. DNA into sympatric North American gray DESJARDINS, P., AND R. MORáIS. 1990. Sequence and wolf populations. Evolution 45:104-119. gene organization of the chicken mitochon- LOUSADA, S. A., AND S. N. G. HOWELL. 1996. drial genome. Journal of Molecular Biology Distribution, variation, and conservation of 212:599-634. Yellow-headed Parrots in northern Central DowLiNG, T. E., G. R. SMITH, AND W. M. BROWN. America. Cotinga 5:46-53. 1989. Reproductive isolation and introgression MONROE, B. L., JR., AND T. R. HOWELL. 1966. between Notropis cornutus and Notropis chryso- Geographic variation in Middle American cephalus (family Cyprinidae): Comparisons of parrots of the Amazona ochrocephala complex. morphology, allozymes, and mitochondrial Occasional Papers of the Museum of Zoology, DNA. Evolution 43:620-634. no. 34. Louisiana State University, Baton FELSENSTEIN, J. 1988. Phylogenies from molecular Rouge. sequences: Inference and reliability. Annual MoRiTz, C. 1994. Defining "evolutionary signifi- Review of Genetics 22:521-565. cant units" for conservation. Trends in Ecology FELSENSTEIN, J. 1995. PHYLIP (Pylogenetic and Evolution 9:373-375. Inference Package) 3.57c. Published by the MUNDY, N. I., P. UNITT, AND D. S. WOODRUFF. 1997. author. University of Washington, Seattle. Skin from feet of museum specimens as a non- Available for download at evolution.gs.washi destructive source of DNA for avian genotyp- ngton.edu/phylip.html. ing. Auk 114:126-129. FERRIS, S. D., R. D. SAGE, C.-M. HUANG, J. T. MURRAY, M. G., AND W. F. THOMPSON. 1980. Rapid NIELSEN, U. RITTE, AND A. C. WILSON. 1983. isolation of high molecular weight plant DNA. Flow of mitochondrial DNA across a spe- Nucleic Acids Research 8:4321-4325. cies boundary. Proceedings of the National NICHOLS, T. D. 1980. An unusual amazon hybrid. Academy of Sciences USA 80:2290-2294. ICBP Technical Publication 1:169-170. FoRSHAw, J. M. 1989. Parrots of the World, 3rd ed. NoRES, M. 1999. An alternative hypothesis for the Landsdowne Editions, Melbourne, Australia. origin of Amazonian bird diversity. Journal of FRIESEN, V. L., B. C. CONGDON, H. E. WALSH, AND Biogeography 26:475-485. T. P. BiRT. 1997. Intron variation in Marbled NoviCK, R. R., C. W. DICK, M. R. LEMES, C. NAVARRO, Murrelets detected using analyses of single- A. CACCONE, AND E. BERMINGHAM. 2003. Genetic stranded conformational polymorphisms. structure of Mesoamerican populations of Molecular Ecology 6:1047-1058. big-leaf mahogany {Swietenia macrophylla) in- GOLDMAN, N., J. P. ANDERSON, AND A. G. RODRIGO. ferred from microsatellite analysis. Molecular 2000. Likelihood-based tests of topologies in Ecology 12:2885-2893. phylogenetics. Systematic Biology 49:652-670. PALUMBI, S. R. 1996. Nucleic acids II: The poly- HAFFER, J. 1969. Speciation in Amazonian forest merase chain reaction. Pages 205-247 in birds. Science 165:131-137. Molecular Systematics, 2nd ed. (D. M. Hillis, HAFFER, J. 1974. Avian Speciation in Tropical C. Moritz, and B. K. Mable, Eds.). Sinauer South America. Publications of the Nuttall Associates, Sunderland, Massachusetts. Ornithological Club, no. 14. PERDICES, A., E. BERMINGHAM, A. MONTILLA, AND HAFFER, J. 1985. Avian zoogeography of the I. DoADRio. 2002. Evolutionary history of the 332 EBERHARD AND BERMINGHAM [Auk, Vol. 121 genus Rhamdia (Teleostei: Pimelodidae) in SiBLEY, C. G., AND B. L. MONROE, JR. 1990. Central America. Molecular Phylogenetics and Distribution and Taxonomy of Birds of the Evolution 25:172-189. World. Yale University Press, New Haven, POSADA, D., AND K. A. CRANDALL. 1998. Connecticut. MODELTEST: Testing the model of DNA sub- SHIMODAIRA, H., AND M. HASEGAWA. 1999. Multiple stitution. Bioinformatics 14:817-818. comparisons of log-likelihoods with applica- POWELL, J. R. 1983. Interspecific cytoplasmic gene tions to phylogenetic inference. Molecular flow in the absense of nuclear gene flow: Biology and Evolution 16:1114-1116. Evidence from Drosophila. Proceedings of SIMPSON, B. B., AND J. HAFFER. 1978. Speciation pat- the National Academy of Sciences USA 80: terns in the Amazonian forest biota. Annual 492-495. Review of Ecology and Systematics 9:497-518. QuESADA, H., R. WENNE, AND D. O. F. SKIBINSKL 1995. SwoFFORD, D. L. 1999. PAUP* Phylogenetic Differential introgression of mitochondrial DNA Analysis Using Parsimony (*and Other across species boundaries within the marine Methods), version 4. Sinauer Associates, mussel genus Mytilus. Proceedings of the Royal Sunderland, Massachusetts. Society of London, Series B 262:51-56. TAMURA, K., AND M. NEI. 1993. Estimation of the RoHWER, S., E. BERMINGHAM, AND C. WOOD. 2001. number of nucleotide substitutions in the con- Plumage and mitochondrial DNA haplo- trol region of mitochondrial DNA in humans type variation across a moving hybrid zone. and chimpanzees. Molecular Biology and Evolution 55:405-422. Evolution 10:512-526. RusELLO, M. A., AND G. AMATO. 2004. A molecu- TARR, C. L., AND R. C. FLEISCHER. 1993. Mitochondrial lar phylogeny of Amazona: Implications for DNA variation and evolutionary relationships Neotropical parrot biogeography, taxonomy in the Amakihi complex. Auk 110:825-831. and conservation. Molecular Phylogenetics TEGELSTROM, H. 1987. Transfer of mitochondrial and Evolution 30:421-437. DNA from the northern red-backed vole SAVAGE, J. M. 1982. The enigma of the Central (Clethrionomys rutilus) to the bank vole (C. American herpetofauna: Dispersals or vicari- glareolus). Journal of Molecular Evolution 24: ance? Annals of the Missouri Botanical Garden 218-227. 69:464-547. WALLACE, A. R. 1853. A Narrative of Travels on the SHIELDS, G. F., AND A. C. WILSON. 1987. Calibration Amazon and Rio Negro. Reeves, London. of mitochondrial DNA evolution in geese. Journal of Molecular Evolution 24:212-217. Associate Editor: R. Zinic