The Auk 118(4):874-887, 2001

A MOLECULAR PHYLOGENY OF THE DOVE GENERA STREPTOPELIA AND COLUMBA

K ev in P. Jo h n so n ,1-3-5 S elv in o de K o r t ,2 K a ren D in w o o d ey ,3 A. C. M a tem a n ,4 C a r el ten C a t e ,2 C. M . L essells,4 a n d D a le H. C la y to n 3 1Illinois Natural History Survey, Champaign, Illinois 61820, USA; institute of Evolutionary and Ecological Sciences, Leiden University, Leiden, The Netherlands; departm en t o f Biology, University o f Utah, Salt Lake City, Utah 84112, USA; and Netherlands Institute of Ecology, Heteren, The Netherlands

A b s t r a c t . — Evolutionary history of the dove genus Streptopelia has not been examined with rigorous phylogenetic methods. We present a study of phylogenetic relationships of Streptopelia based on over 3,600 base pairs of nuclear and mitochondrial gene sequences. To test for monophyly of Streptopelia, we used several other columbiform taxa, including Colum- ba (Old and New World), Macropygia, Reinwardtoena, and the enigmatic Pink Pigeon (Nesoenas mayeri). On the basis of our analyses, Streptopelia (as currently defined) is not monophyletic; Nesoenas mayeri is the sister species to S. picturata, resulting in paraphyly of Streptopelia. Three main clades of Streptopelia are identified: (1) S. chinensis plus S. senegalensis, (2) S. picturata plus Nesoenas mayeri, and (3) all other species of Streptopelia. It is unclear whether those clades form a monophyletic group to the exclusion of Old World Columba, but several analyses pro­ duce that result. Species of Old World Columba are closely related to Streptopelia, with species of New World Columba clustering outside that group. Taxonomic changes suggested by our results include merging Nesoenas with Streptopelia and changing the generic name for New World Columba species to Patagioenas. Vocal similarities between S. picturata and N mayeri are striking, given the general diversity of vocalizations in other species. Received 20 September 2000, accepted 27 March 2001.

Species of doves in the genus Streptopelia are lationships within the genus are uncertain. No­ important model systems for studies of physi­ wak (1975) examined morphological character­ ology (Walker et al. 1983, Janik and Bun tin istics within Streptopelia and produced a 1985, Cheng 1986, Ramos and Silver 1992, ten classification that grouped species into several Cate et al. 1993, Lea et al. 1995, Georgiou et al. sub genera. However, he did not produce an ex­ 1995) and behavior (Lade and Thorpe 1964, plicit phylogenetic tree. Goodwin (1983) de­ Zenone et al. 1979, Cheng et al. 1981, Cheng picts a tree of "presumed relationships" 1992, Slabbekoorn and ten Cate 1998, Slabbe- among species of Streptopelia, and that tree dif­ koorn et al. 1999). The 16 species of Streptopelia fers from Nowak's classification, but Goodwin's historically had an African and Eurasian dis­ tree is not based on a rigorous phylogenetic tribution, but some species have been intro­ analysis. Johnson and Clayton (2000a) showed duced to the New World and Australia. Several that species of Old World Columba form the sis­ species of Streptopelia seem highly adaptable to ter group to Streptopelia, with species of New human-altered environments and have ex­ World Columba being more distantly related to panded their ranges considerably (e.g. S. de- both groups. However, the authors included caocto; Bijlsma 1988, Hengeveld and van den only single representatives of Streptopelia and Bosch 1991, Hengeveld 1993, Kasparek 1996). Old World Columba in their study. The goal of Other species have remarkably localized distri­ the present study is to assess monophyly of the butions (e.g. S. hypopyrrha). An understanding genus Streptopelia and identify relationships of historical relationships would provide an within the genus. We also use this study to test important context for work on physiology, be­ Nowak's (1975) classification and Goodwin's havior, and biogeography of this genus. (1983) proposed relationships of species within Even though species of Streptopelia have been the genus. We include several representatives well studied in many respects, phylogenetic re­ of New and Old World Columba, as well as the endangered, enigmatic Pink Pigeon (Nesoenas 5 E-mail: [email protected] mayeri) of Mauritius. We also include represen­

874 October 2001] Phylogeny of Streptopelia 875 tatives of Macropygia and Reinwardtoena, which of individual variation within each species, we se­ together with Columba and Streptopelia form a quenced (when available) multiple individuals for distinct clade within Columbiformes (Johnson the COI gene (see Table 3 for list of those multiple and Clayton 2000a). The entire phylogenetic individuals and GenBank accession numbers). Mi­ tochondrial genes tend to evolve at similar rates at analysis is rooted on Geotrygon, Leptotila, and low levels of sequence divergence in birds (Johnson Zenaida, which together consitute the sister and Sorenson 1998, Johnson and Lanyon 1999) and group to the Columba/ Streptopelia / Macropygia more specifically for cyt b, ND2, and COI in Col­ clade (Johnson and Clayton 2000a). umbiformes (Johnson and Clayton 2000b). Mito­ For our current study, we use sequences of chondrial divergences are also highly correlated both mitochondrial and nuclear genes to con­ with divergences in FIB7 in Columbiformes (Johnson struct a phylogeny for Streptopelia, Columba, and and Clayton 2000a, b). Thus, within-species varia­ related taxa. We compare relative usefulness of tion in COI sequences is likely to be representative of nuclear and mitochondrial genes for phyloge­ other mitochondrial and nuclear genes. In general, netic resolution at that level. Using the phylog­ individuals of the same species differed only slightly, if at all, in those COI sequences (see below), so single eny we recommend some changes in taxonomic exemplar individuals are reasonable for the multi­ classification, and compare phylogeny to vocal gene data set. similarities and diversity in Streptopelia. Where possible, we used tissue or feather samples to avoid risk of nuclear copies of mitochondrial M e t h o d s genes (Sorenson and Quinn 1998). In cases where blood samples were used, we verified sequences for Sequencing.—We obtained samples of muscle tis­ COI using multiple individuals. We also checked sues, feathers, or blood from representatives of 14 of chromatograms for signs of double peaks, as well as the 16 described species of Streptopelia (Table 1). We checking for indels and stop codons. By sequencing also sampled members of New and Old World Co­ several mitochondrial genes, we were also able to lumba, Macropygia, Reinwardtoena, Nesoenas, Geotry­ test for any incongruence in phylogenies resulting gon, Leptotila, and Zenaida. We extracted DNA from from those gene regions, which would likely occur if those samples using a Qiagen Tissue Kit (Valencia, some of the sequences represented nuclear copies. California) with the manufacturer's protocols. For We found no evidence of nuclear copies in our feather samples (~2 mm of the tip of the shaft) we sequences. also added 30 |jlL of 10% DTT to the digestion buffer. Phylogenetic analysis.—Relative rates of substitu­ Using PCR we amplified portions of three mitochon­ tion can be examined by plotting pairwise sequence drial genes: cytochrome-^ (cyt b), cytochrome oxi­ divergences for various substitution types and dase I (COI), and NADH dehydrogenase subunit 2 genes. For the nuclear gene (FIB7), we estimated the (ND2). We also amplified the nuclear gene p-fibrin- "native" transition:transversion ratio by plotting ogen intron 7 (FIB7). Table 2 lists primers used for pairwise transition differences against pairwise those reactions. Protocols for reactions follow John­ transversion differences. Slope of the linear portion son and Clayton (2000a). of that curve estimates the transition:transversion ra­ Direct sequencing of PCR products and determi­ tio (Sturmbauer and Meyer 1992). We do not fit re­ nation of sequences was performed as described by gressions through those points because they are non­ Johnson and Clayton (2000a). We aligned sequences independent, rather we use slopes as a rough across species using Sequencher (GeneCodes, Mad­ indicator of relative rates. For mitochondrial genes, ison, Wisconsin). The three mitochondrial genes to­ we plotted transition difference at third positions taled 2,520 bp and the nuclear intron included 1,150 against transversions at third positions to estimate aligned base pairs (GenBank accession numbers in that ratio. To estimate relative rates of mitochondrial Table 1). In the case of mitochondrial protein coding versus nuclear substitution, we plotted pairwise di­ genes, alignment was straightforward. We noted sev­ vergences for mitochondrial genes against those for eral insertion-deletion (indel) events in the nuclear FIB7. We also plotted pairwise divergences for in­ intron, but for those taxa, divergences were so low dividual mitochondrial genes against each other and that manual alignment was straightforward. Gaps against FIB7 to determine which genes are more sub­ were treated as missing data in phylogenetic analy­ ject to multiple substitution. ses. For Streptopelia tranquebarica and S. bitorquata, PAUP* (Swofford 2000) was used for all phyloge­ only partial FIB7 sequences were available because of netic analyses. We used the partition homogeneity amplification failures. Thus, we repeated analyses test (Farris et al. 1994, 1995; Swofford 2000) to ex­ involving the nuclear intron both with and without amine whether the gene regions could be considered those species. samples of the same underlying phylogeny, or For each species, we included single representa­ whether there was evidence for different phyloge­ tives for all sequences. However, to assess the level netic signal between gene regions. Even though that 876

T a b le 1. Samples sequenced for all genes.

GenBank accession numbers Species Locality Sample number Tissue type Cyt b ND2 COI FIB7 Streptopelia decaocto Netherlands SDK 4 Muscle AF353398 A F353418 A F353469 A F353449 Streptopelia roseogrisea Cam eroon SDK 1 Feather A F353399 A F353419 A F353470 A F353450 Streptopelia decipiens Cam eroon SDK 1 Feather A F353400 A F353420 AF353471 AF353451 Streptopelia semitorquata South Africa LSUMNS B34270a Muscle AF353401 AF353421 A F353472 A F353452 Streptopelia capicola South Africa LSUMNS B34271a Muscle A F279709 A F353422 A F279734 AF279719 Streptopelia vinacea Cent. Afr. Rep. AMNH PRS-2168a Muscle AF353402 A F353423 A F353473 A F353453 Streptopelia hypopyrrha Cam eroon SDK 4 Muscle AF353403 A F353424 A F353474 A F353454 Streptopelia turtur Kazakhastan UWBM DAB-2363 Muscle AF353404 A F353425 A F353475 AF353455 Streptopelia orientalis Russia UWBM 47282a Muscle A F353405 A F353426 A F353476 A F353456 Streptopelia bitorquata Captive SDK 1 Feather A F353406 A F353427 A F353477 A F353457 Streptopelia tranquebarica Captive SDK 2 Feather AF353407 AF353428 A F353478 AF353458 Nesoenas (Streptopelia) mayeri Captive Tracy Aviary Feather A F353408 A F353429 A F353479 AF353459 Johnson et a l. l. a et Johnson Streptopelia picturata M adagascar CC 1 Feather A F353409 A F353430 A F353480 A F353460 Streptopelia chinensis Philippines FMNH DW-47123 Muscle A F182695 AF353431 A F353481 AF182662 Streptopelia senegalensis South Africa LSUMNS B34209a Muscle A F279710 A F353432 A F279735 AF279720 Columba liva Utah 423 Muscle A F182694 A F353433 A F279733 A F182661 Columba rupestris Mongolia UWBM 59755a Muscle A F353410 A F353434 A F353482 A F353461 Columba guinea South Africa LSUMNS B34209a Muscle A F279708 A F353435 A F279732 A F279718 Columba palumbus Captive Tracy Aviary Muscle AF353411 A F353436 A F353483 A F353462 Columba arquatrix South Africa CC 2 Feather A F353412 A F353437 A F353484 A F353463 Columba pulchrichollis Captive SDK 1 Muscle A F353413 A F353438 A F353485 A F353464 Columba plumbea Brazil FMNH ATP86-1363 Muscle AF182691 AF251547 A F279736 A F182658 Columba subvinacea Peru FMNH SML-1045a Muscle A F182692 A F353439 A F353486 A F182659 Columba oenops Peru FMNH AJB-5563 Muscle A F182690 A F353440 A F353487 A F182657 Columba leucocephala Florida KUMNH B1798a Muscle A F182689 AF353441 A F353488 A F182656 Columba speciosa Mexico KUMNH B2096a Muscle AF279711 A F353442 A F279737 A F279721 Columba fasciata Utah DHC l a Muscle A F353414 A F353443 A F353489 A F353465 Macropygia mackinlayi Solomon Islands AMNH MKL-823 Muscle A F353415 A F353444 A F353490 A F353466 Macropygia tenurirostris Philippines FMNH SEA-0743 Muscle A F353416 A F353445 AF353491 A F353467 Reinwardtoena browni Captive NMNH B40243 Muscle A F353417 A F353446 A F353492 A F353468 Zenaida asiatica Arizona KPJ 5 Muscle A F251533 A F251543 AF279731 A F258324 Zenaida macroura Arizona KPJ 5 Muscle A F251530 A F251535 A F353493 AF258321 118 Vol. [Auk, Geotrygon montana Peru KUMNH B9953 Muscle AF182696 A F353447 A F279728 A F182663 Leptotila rufaxilla Peru KUMNH B793a Muscle A F182698 A F251546 A F353494 A F182665 Leptotila verreauxi Texas DHC 5a Muscle A F279705 A F353448 A F279725 A F279715 a Indicates samples that have a corresponding museum skin voucher. October 2001] Phylogeny of Streptopelia 877

T a b l e 2 . Primers used in study.

Gene Primer name Source cyt b L14841 Kocher et al. (1989) H 4a Harshman (1996) H 15299 Kocher et al. (1989) L15517 Johnson and Sorenson (1998) ND2 L5215 Hackett (1996) H 6313 Johnson and Sorenson (1998) L5758 Johnson and Sorenson (1998) COI L6625 Hafner et al. (1994) H 7005 Hafner et al. (1994) FIB7 FIB-BI7L Prychitko and Moore (1997) FIB-BI7U Prychitko and Moore (1997) FIB-DOVEF Johnson and Clayton (2000a) f i b -d o v e r Johnson and Clayton (2000a)

test indicated no significant incongruence between dall 1997). Using the estimated model parameters, any of the gene regions (see below), we also con­ we conducted 50 heuristic maximum-likelihood ducted several analyses with mitochondrial and nu­ searches with nearest neighbor interchange (NNI) clear genes separately. We did this because of the branch swapping. We used bootstrapping (100 rep­ slow rate and lack of informative sites in the nuclear licates with NNI branch swapping) to evaluate rela­ gene and because we wanted to identify what nodes tive support for nodes in maximum-likelihood were common to trees derived from independently topologies. sorting gene regions. We conducted parsimony analyses with a range of R e su l t s transversion weighting schemes from 1:1 (unordered parsimony) to 20:1. For each weighting scheme, we Sequence variation.—Mitochondrial genes used 50 random addition replicates with heuristic were similar in the fraction of sites that were searches. For each analysis, we also performed 1,000 variable and potentially phylogenetically infor­ bootstrap replicates (Felsenstein 1985) to assess rel­ mative, but ND2 showed the most variation, ative support for various branches in the phylogeny. followed by cyt b, then COI (Table 4). FIB7 We used one of the combined unordered parsi­ mony trees to estimate parameters for maximum- showed a much smaller fraction of variable and likelihood searches. We used likelihood ratio tests to phylogenetically informative sites (Table 4). estimate the best fit model that could not be rejected Within Streptopelia, mitochondrial sequence di­ in favor of a simpler model (Huelsenbeck and Cran­ vergences ranged from 1.9 to 12.4%. For those

T a b l e 3 . Samples sequenced for COI only.

GenBank Species Locality Sample number Tissue type number Streptopelia decaocta Netherlands SDK 1 Feather A F353495 Streptopelia decaocto Netherlands SDK 2 Feather A F353496 Streptopelia roseogrisea Cam eroon SDK 2 Blood A F353497 Streptopelia roseogrisea Cam eroon SDK 3 Feather A F353498 Streptopelia decipiens Cam eroon SDK 2 Feather A F353499 Streptopelia semitorquata Cam eroon SDK 1 Feather A F353500 Streptopelia capicola South Africa LSUMNS B342053 Muscle AF353501 Streptopelia vinacea Cam eroon SDK 1 Feather A F353502 Streptopelia hypopyrrha Cam eroon SDK 2 Feather A F353503 Streptopelia turtur Cam eroon SDK 1 Feather A F353504 Streptopelia turtur Cam eroon SDK 2 Feather A F353505 Streptopelia orientalis Captive SDK 2 Feather A F353506 Streptopelia tarnquebarica Captive SDK 3 Feather A F353507 Streptopelia picturata Captive SDK 1 Feather A F353508 Streptopelia chinensis Singapore AMNH PRS-6783 Fuscle A F353509 Streptopelia senegalensis Cam eroon SDK 1 Feather A F353510 a Indicates samples that have a corresponding museum skin voucher. 878 J o h n s o n e t a l . [Auk, Vol. 118

T a b le 4. Variable and phylogenetically informative 0.25 sites for each gene region.

Number Percent Percent 0.2- Gene of sites variable information

cyt b 1070 39.0% 33.1% ND2 1067 48.7% 39.8% 0.15 COI 383 34.2% 30.0% cI FIB7 1150 19.6% 7.9% 0.1 - Id •c same comparisons, FIB7 showed much less di­ 0 .0 5 - vergence (0.0 to 1.4%). Plots of third position transitions against third position transversions for mitochondrial genes (Fig. 1) indicated con­ 0 0.025 0.05 0.075 0.1 siderable multiple substitution of third posi­ tion transitions with an estimated transition: mitochondrial third position transversions trans version ratio of approximately 10:1. F ig . 1. Plot of pairwise percentage divergence for Divergences of mitochondrial genes were sim­ transitions against transversions for third positions ilar, but comparisons of those divergences in­ of all mitochondrial genes combined. dicated that ND2 was less subject to multiple substitution than was cyt b or COI, because plots of ND2 divergence against FIB7 or the drial genes indicate that those genes could be other mitochondrial genes indicated continu­ considered samples of the same underlying phy­ ing accumulation of substitutions at high di­ logeny (P = 0.44). Similarly, trees derived from vergences. The estimated transition: trans ver­ FIB7 alone compared to trees for mitochondrial sion ratio for FIB7 was approximately 1.5:1 genes combined were not significantly incon- (Fig. 3), with no evidence of multiple substi­ gruent (P = 1.0). tution of transitions compared to transver­ Unordered parsimony analysis of all three sions. Plots of divergences for the mitochon­ mitochondrial genes combined produced three drial genes against FIB7 divergence indicated a equally parsimonious trees (not shown, but see 5 to 10 X higher rate of substitution in the mi­ Fig. 5 for reference). Those trees indicated that tochondrial genes (Fig. 4). Streptopelia is not monophyletic. Nesoenas mayeri Several indels occurred in FIB7. One of those is the sister taxon to S. picturata with high boot­ is a 166 bp deletion shared by Streptopelia, Ne- strap support. In addition, two groups of Strep­ soenas, and Old World Columba. Other colum- topelia are sister to Old World Columba, but that biform taxa do not show that deletion. Other in­ result has only weak bootstrap support (49%). dels in FIB7 ranged from 1 to 125 bp, but only One large group of Streptopelia is monophyletic two others were potentially phylogenetically (clade A) with strong support (100%). informative (one uniting Streptopelia, Nesoenas, Trans version weighting of 2:1 to 5:1 pro­ Old World Columba, and Macropygia the other duced identical tree topologies (one tree) that uniting clade A of Streptopelia [Fig. 5]). were completely resolved (not shown). Analy­ Within species, little sequence variation was sis of mitochondrial genes with those low evident. For the COI gene, within-species vari­ transversion weights also produced a paraphy- ation ranged from 0.0% (for 6 of 13 species with letic Streptopelia, both with respect to Nesoenas multiple samples) to 1.0% (for comparisons of and with respect to Old World Columba. How­ Streptopelia semitorquata between South Africa ever, transversion weighting of 10:1 and higher and Cameroon). The low within-species varia­ produced a single tree showing monophyly of tion for mitochondrial genes suggests that single Streptopelia with respect to Old World Columba. species exemplars are suitable for species-level Nesoenas was still sister to S. picturata, again phylogenetic analysis, given that between-spe- with high bootstrap support. cies divergences all exceeded 2.5%. Unordered parsimony analysis of the nuclear Phylogeny.—Partition homogeneity tests com­ intron alone produced 32,020 trees (not paring congruence among the three mitochon- shown). The high number of trees was not sim- O ctober 2001] ctober O plots (A -C ) indicate expectation if rates are equal. are rates if expectation indicate ) -C (A plots cyt (D) COI, versus F ig . . ig F cyt b percent divergence ND2 percent divergence ND2 percent divergence 2. Plot of pairw ise percentage divergence for (A) ND2 versus cyt cyt versus ND2 (A) for divergence percentage ise pairw of Plot 2. COI percent divergence FIB7 percent divergencepercent FIB7 divergencepercent COI b cyt cyt versus FIB7, (E) ND2 versus FIB7, (F) COI versus FIB7. Lines for the mitochondrial gene mitochondrial the for Lines FIB7. versus COI (F) FIB7, versus ND2 (E) FIB7, versus b percent divergence FIB7 percent divergencepercent FIB7 divergencepercent Phylogeny of Phylogeny Streptopelia b , (B) ND2 versus COI, (C) cyt cyt (C) COI, versus ND2 (B) , 879 b

880 J o h n s o n e t a l . [Auk, Vol. 118

0.2

0 0.02 0.04 0.06 0.08

FIB7 transversions FIB7 percent divergence

F ig. 3. Plot of pairwise percentage divergence for F ig. 4. Plot of overall percentage divergence for transitions against transversions for all positions of all mitochondrial genes combined against percent­ nuclear FIB7. age divergence for nuclear FIB7. ply an artifact of incomplete sequences for els. Maximum-likelihood searches on mito­ Streptopelia bitorquata and S. tranquebarica. Ex­ chondrial genes alone produced a tree very clusion of those sequences from phylogenetic similar to parsimony trees, and that tree indi­ analyses produced 4,236 trees, the consensus of cated monophyly for Streptopelia with respect which was identical to the consensus of trees to Old World Columba. Likelihood analysis of with those taxa included. FIB7 trees did not re­ the intron alone identifies monophyly for the solve relationships among major Streptopelia three main clades of Streptopelia and Old World lineages plus Old World Columba. However, Columba, but does not resolve relationships New World Columba is placed outside Strepto­ among those groups. There is also little reso­ pelia and Old World Columba (bootstrap sup­ lution within the large Streptopelia clade A. port 100%). Monophyly of both clades B and C Maximum-likelihood trees for combined gene of Streptopelia (see Fig. 5 for reference) was regions (Fig. 6) are well resolved and indicate strongly supported in parsimony analyses of monophyly for Streptopelia (inclusive of Nesoen­ FIB7 alone, as was monophyly of Old World as) with moderate support (bootstrap 64%). Columba. The three main clades of Streptopelia identified Unordered parsimony analysis of combined by combined parsimony analysis are also re­ gene regions (mitochondrial and nuclear) pro­ covered in that tree with strong support (100% duced two trees (Fig. 5). Strict consensus of in each case), as is paraphyly of Columba (100% those trees, like unordered parsimony analysis bootstrap support for sister relationship be­ of mitochondrial genes alone, does not recover tween Old World Columba and Streptopelia). In a monophyletic Streptopelia with respect to Old fact, support for a sister relationship between World Columba. However, in most respects that New World Columba and Old World Columba + tree is very similar to trees produced by sepa­ Streptopelia is relatively poor (51%). Most other rate analysis. Increasing transversion weight­ branches are identical to those recovered by un­ ing of combined gene regions (in excess of 10: ordered parsimony analysis (Fig. 5). 1) produced trees with a monophyletic

Streptopelia with respect to Old World Columba D is c u s s io n (not shown, see Fig. 6 for reference). Likelihood ratio tests indicated that models Phylogeny.—A phylogeny based on 2,530 bp incorporating unequal base frequencies, six of mitochondrial and 1,150 bp of nuclear DNA substitution types, and rate heterogeneity were sequences is generally well resolved and well statistically better justified than simpler mod­ supported for the dove genus Streptopelia and October 2001] Phylogeny of Streptopelia 881

Streptopelia decaocto Streptopelia roseogrisea Streptopelia decipiens 59 100 ----- Streptopelia capicola — Streptopelia vinacea Streptopelia semitorquata Clade A

100 100 |— Streptopelia hypopyrrha 100 Streptopelia turtur ■cEStreptopelia orientalis Streptopelia bitorquata Streptopelia tranquebarica 100 Columba livia 99 69 - c Columba rupestris 74 Columba guinea Old World Columba palumbus Columba 88 Columba arquatrix Columba pulchricollis 50 100 Nesoenas mayeri Clade B - a — Streptopelia picturata 100 — Streptopelia chinensis ------Streptopelia senegalensis Clade C 86 -----Columba leucocephala 66 - Columba speciosa New World 100 - Columba plumbea 99 100 “Columba” Columba subvinacea (Patagioenas) 100 Columba oenops Columba fasciata 100 ------Macropygia mackinlayi 100 Macropygia tenuirostris Reinwardtoena browni 100 ------Leptotila rufaxilla 55 ------Leptotila verreauxi 100 Zenaida asiatica ------Zenaida macroura Geotrygon montana

F ig . 5. Strict consensus of two trees (length = 4470, RC = 0.200) from unordered parsimony analysis of combined mitochondrial (cyt b, ND2, COI) and nuclear (FIB7) genes. Numbers above branches indicate boot­ strap support from 1,000 full heuristic replicates. Unlabelled nodes received <50% bootstrap support. Branch lengths are proportional to reconstructed changes using parsimony. Indicated groups are clades referred to in text.

outgroup taxa. However, even given the rela­ better resolution for the phylogeny of Streptope­ tively large amount of DNA sequence data, lia. Mitochondrial genes have around a 5 to 10 X some aspects of the tree are still weakly sup­ higher substitution rate (Fig. 4), and thus vari­ ported. Primary among those uncertainties is ation can accumulate in mitochondrial DNA be­ whether or not Streptopelia is monophyletic tween closely related species. For several closely with respect to Old World Columba. related species of Streptopelia, sequences of FIB7 Despite the lower homoplasy present in the were identical, preventing resolution of species FIB7 sequences, mitochondrial genes provided level relationships. Importantly for this study, 882 J o h n s o n e t a l . [Auk, Vol. 118

100 ------Streptopelia decaocto ' 100 ------Streptopelia roseogrisea 70 — Streptopelia decipiens 82 — -Streptopelia semitorquata 100 ----- Streptopelia capicola — Streptopelia vinacea Clade A 100 100j------Streptopelia hypopyrrha 100 *----- Streptopelia turtur ------Streptopelia orientalis 64 92 ------Streptopelia bitorquata ------Streptopelia tranquebarica 100 Nesoenas mayeri 74 — Streptopelia picturata CladeB 100 “ Streptopelia chinensis 100 ------Streptopelia senegalensis Clade C 100 Columba livia 88 C Columba rupestris 100 ------Columba guinea Old World 85 ----- Columba palumbus Columba 100 51 ------Columba arquatrix ---- Columba pulchricollis 96 — Columba plumbea Columba subvinacea New World 100 100 ■ Columba oenops “Columba” 100 Columba leucocephala (Patagioenas) — Columba speciosa - Columba fasciata 100 Macropygia mackinlayi 100 Macropygia tenuirostris ------Reinwardtoena browni 100 ------Leptotila rufaxilla 54 ------Leptotila verreauxi 100 ------Zenaida asiatica ------Zenaida macroura ------Geotrygon montana -0.01 substitutions/site

Fig. 6. Tree resulting from maximum likelihood searches using combined mitochondrial (cyt b, ND2, COI) and nuclear (FIB7) genes (In likelihood = -25,202; model parameters: gamma shape parameter = 0.209 with eight rate categories, transition matrix: A-C = 1.41, A-G = 17.96, A-T = 1.22, C-G = 0.54, C-T = 16.77, G-T = 1.0). Numbers above branches indicate bootstrap support from 100 heuristic search replicates. Un­ labelled nodes received <50% bootstrap support. Branch lengths are proportional to likelihood estimated branch lengths (scale indicated). Indicated groups are clades referred to in text.

trees based on FIB7 identified the same major groups involve only two species: S. chinensis + clades that were identified by analyses of mito­ S. senegalensis (clade C) and S. picturata + Ne­ chondrial genes. In addition, there was no sig­ soenas (Streptopelia) mayeri (clade B). The other nificant conflict between nuclear and mitochon­ clade (A) includes all other species of Streptope­ drial data sets, suggesting that trees recovered lia. Monophyly of each of those clades is gener­ in our analyses estimate the species' phylogeny ally well supported, but relationships among rather than simply gene trees. those groups and Old World Columba is less All analyses identify three major clades with­ clear. Analyses that take into account large tran- in Streptopelia (Figs. 5 and 6). Two of those sition:transversion biases (transversion-weight­ October 2001] Phylogeny of Streptopelia 883 ed parsimony and maximum likelihood) tend to ranges from 11.1 to 11.9%, whereas the maxi­ recover monophyly for Streptopelia (inclusive of mum mitochondrial divergences within clade Nesoenas). A of Streptopelia range to 8.4%. Phylogenetic relationships within Streptope­ Streptopelia vinacea and S. capicola are a pair lia do not correspond in a strong way to either of allopatric sister taxa of particular interest. Nowak's (1975) classification or Goodwin's Those taxa are distributed in sub-Saharan (1983) hypothesized evolutionary relation­ northern Africa and in southeastern Africa, re­ ships. However, similarities with both are evi­ spectively. Another pair of dove taxa with a dent. Nowak (1975) unites S. decipiens, capicola, similar distribution is Turtur abyssinicus and T. vinacea, tranquebarica, semitorquata, roseogrisea, chalcospilos. Mitochondrial sequence diver­ and decaocto into a subgenus. We find support gence between S. vinacea and S. capicola is 2.5% for a similar clade (excluding tranquebarica). and that between T. abyssinicus and T. chalcos­ Goodwin (1983) outlines a similar group, but pilos is 1.7% (K. P. Johnson unpubl. data). Sim­ places tranquebarica outside of it and puts bitor­ ilarity in geographic distributions and genetic quata close to decaocto. In our phylogenetic anal­ divergences suggests some common biogeo­ yses, S. tranquebarica and S. bitorquata consis­ graphic factor underlying speciation between tently fall at the base of a clade containing those those species of Streptopelia and Turtur, perhaps two species plus S. orientalis, turtur, and hypo- ~1 Mya. pyrrha. Like many species of Streptopelia, tran­ Taxonomic implications.—The Pink Pigeon, in quebarica and bitorquata possess a "ring-neck," the monotypic genus Nesoenas Salvadori 1893, unlike orientalis, turtur, and hypopyrrha that is sister to Streptopelia picturata. That result is have a dark patch on the side of the neck. Our evidenced by two independent gene regions results suggest that the ring-neck (a dark bar and receives strong support in all analyses. The running around the back of the neck) is the an­ sister relationship between Nesoenas mayeri cestral condition in group A of Streptopelia, hav­ (Mauritius) and S. picturata (Madagascar) is not ing been lost in the ancestor of S. orientalis, tur­ surprising, given the geographic proximity of tur, and hypopyrrha. Species of Streptopelia those two islands. In light of those results, we outside group A tend to have spotting on the recommend transferring N. mayeri to the genus neck. The close relationship of S. chinensis and Streptopelia Bonaparte 1855. S. senegalensis was recognized by Goodwin Regarding the generic status of Streptopelia, (1983) and to some extent by Nowak (1975). our results do not conclusively demonstrate A novel finding of our study is the sister re­ monophyly of Streptopelia. However, in analy­ lationship between the Pink Pigeon (Nesoenas ses that take into account rate heterogeneity mayeri), which is endemic to Mauritius, and the and transition biases, Streptopelia (as redefined Madagascan Turtle Dove (Streptopelia pictura­ to include Nesoenas mayeri) is monophyletic. On ta), endemic to Madagascar. In most classifica­ the basis of those results, we suggest retaining tions, Nesoenas is united with species of Old the name Streptopelia for all species currently in World Columba whereas the Madagascan Turtle the genus. An alternative would be to recog­ Dove is considered to be an aberrant Strepto­ nize the three main clades (A-C) as separate pelia. On the basis of the close relationship of genera, but we feel a more conservative ap­ those two species, we suggest that the common proach, minimizing name changes, is prudent. ancestor of the Pink Pigeon and the Madagas­ All of our analyses indicated that Old World can Turtle Dove colonized Mauritius from Columba species are sister to, or imbedded Madagascar and subsequently speciated. Per­ within, Streptopelia. That phylogenetic position centage divergence for mitochondrial genes be­ results in paraphyly for Columba. Our current tween the Pink Pigeon and the Madagascan study confirms preliminary results of Johnson Turtle Dove is only 3.0%. On the basis of mi­ and Clayton (2000a), which also indicated that tochondrial molecular clock calibrations for relationship. Because Columba for Old World birds (Klicka and Zink 1997), that represents a species has priority, we recommend transfer­ colonization time of —1.5 Mya, much younger ring all species of New World Columba into the than the age of Mauritius (Proag 1995). Mito­ genus Patagioenas (as partially suggested by chondrial divergence between the Madagascan Johnston 1962). Johnston (1962) suggested that Turtle Dove and other species of Streptopelia the Band-tailed Pigeon (Columba fasciata) is a 884 J o h n s o n e t a l . [Auk, Vol. 118 close relative of Old World Columba; however, Streptopelia picturata our results strongly supported monophyly of New World " Columba," inclusive of fasciata. Further analysis is needed to determine if all New and Old World “Columba" species cluster by biogeographic distribution, but we suspect Streptopelia mayeri that will be the case. Vocal evolution.—Streptopelia vocalizations have been studied extensively (Lade and Thorpe 1964, Gaunt et al. 1982, Cheng 1992, Ballintijn and ten Cate 1997, Slabbekoorn and ten Cate t IM « 1998, Slabbekoorn et al. 1999), making that ge­ Streptopelia capicola nus an important group for comparative studies 1.2 of vocalizations. Slabbekoorn et al. (1999) com­ pared acoustic similarity of Streptopelia perch- 0.8 UN Al* coos with Goodwin's ideas on taxonomy (1983). 0.4 They concluded that, at the species level, there is Streptopelia vinacea —i------1------1------1------little congruence between similarity in perch- 1 2 3 4 (*) coos and phylogenetic relatedness, suggesting F ig . 7. Sonograms of selected Streptopelia perch- that environmental or social factors promote the coos. Note the similarity between S. picturata and S. relatively rapid diversification of perch-coo vo­ mayeri. Sonograms of S. capicola and S. vinacea repre­ calizations. Although a formal reanalysis is be­ sent two sister species of smaller genetic divergence, yond the scope of this paper, an informal com­ but greater divergence in perch-coo vocalizations. parison of acoustic similarity in relation to our phylogeny does not contradict Slabbekoorn et al.'s (1999) conclusions. Streptopelia are quite homogeneous in morpho­ The close phylogenetic relationship between logical characteristics. In fact, often the most re­ Streptopelia picturata and S. " Nesoenas" mayeri is liable way to discriminate species in the field is somewhat surprising on the basis of morpho­ by their distinctive perch-coos. Generally logical characters, but is supported by vocal speaking, vocalizations in Streptopelia seem to characteristics. First, the relationship in general be evolutionarily labile compared to morpho­ between Streptopelia and " Nesoenas" is sup­ logical characteristics. ported by similarities in the Excitement Cry Two hypotheses explain why Streptopelia pic­ (Baptista 1997). More dramatically, the perch- turata and S. mayeri show the greatest degree of coos of S. picturata and S. mayeri are very similar vocal similarity within Streptopelia (Slabbe­ (Fig. 7), unlike comparisons of many other sis­ koorn et al. 1999). First, selection may favor di­ ter pairs of Streptopelia. Two other sister species vergence between vocalizations of sympatric pairs also show some similarity in vocaliza­ species, because of a need to use a signal that tions as compared to other Streptopelia species: can be distinguished from that of closely relat­ vinacea-capicola and turtur-hypopyrrha. Al­ ed species (Nelson and Marler 1990). Unlike though those species pairs group together in a other Streptopelia species, S. picturata and S. similarity analysis, they can be separated un­ mayeri are allopatric with all other congeners, ambiguously on the basis of acoustic parame­ perhaps reducing need for divergence in signal ters (Slabbekoorn et al. 1999). Thus perch-coos structure. Second, environment and habitat of­ of other sister species pairs show structural ten play important roles in evolution of avian acoustic differences (e.g. Fig. 7), which is not signals (Morton 1975, Wiley and Richards 1978, the case between S. mayeri and picturata. The lat­ Marchetti 1993, Endler and Th£ry 1996, John­ ter instead show considerable divergence in son 2000, Johnson and Lanyon 2000). Both S. morphological features, including a lack of any picturata and S. mayeri predominantly inhabit neck pattern and a reddish bill in S. mayeri. densely forested habitats through which their Those features, among others, are the reasons low-pitched vocalizations should propagate that S. mayeri has often been placed in a mono­ further than high-pitched vocalizations (Mor­ typic genus. In contrast, most other species of ton 1975, Wiley and Richards 1978, Ryan and October 2001] Phylogeny of Streptopelia 885

Brenowitz 1985, McCracken and Sheldon 1997). mous reviewers provided helpful comments on this Further evidence for potential stabilizing selec­ manuscript. We thank the DNA Sequencing Facility tion on the perch-coos of those two species at the University of Utah, supported in part by Na­ comes from examination of the relationship be­ tional Cancer Institute grant #5p30CA42014. This work was supported by NSF-CAREER award DEB- tween body size and sound frequency. In New 9703003 to D.H.C. World pigeons, a negative correlation exists be­ tween body weight and perch-coo frequency L it e r a t u r e C it e d (Tubaro and Mahler 1998). That relationship is not apparent in the perch-coos of S. picturata Ballintijn, M. R., and C. ten Cate. 1997. Sex dif­ and S. mayeri, because S. mayeri is nearly twice ferences in vocalizations and syrinx of the Col­ as heavy as S. picturata, yet sound frequencies lared Dove (Streptopelia decaocto). Auk 114:22-39. of the perch-coos are very similar (Fig. 7). One Baptist a, L. 1997. Family Columbidae (Pigeons and or both of these factors may have caused vo­ Doves). Pages 60-243 in Handbook of the Birds calizations of those two species to remain rel­ of the World, vol. 4. Sandgrouse to Cuckoos (J. del Hoyo, A. Elliott, and J. Sargatal, Eds.). Lynx atively unchanged since speciation. Edicions, Barcelona, Spain. One other unexpected phylogenetic relation­ B ijls m a , R. G. 1988. Population growth in the Col­ ship is also supported by careful examination lared Dove Streptopelia decaocto in the Nether­ of vocalizations. Exclusion of Streptopelia bitor­ lands. Limosa 61:41-42. quata and S. tranquebarica, despite similarity in C h e n g , M. F. 1986. Female cooing promotes oviarian plumage characteristics, from the group of typ­ development in Ring Doves (Streptopelia risoria). ical ring-necked doves (decaocto through vinacea Physiology and Behavior 37:371-374. clade) is somewhat surprising. The Excitement C h e n g , M. F. 1992. For whom does the female dove or Flight Call is a conspicuous characteristic for coo? A case for the role of vocal self-stimulation. the true ring-necked species (as defined above). Animal Behaviour 43:1035-1044. C h e n g , M. F., M. Porter, and G. Ball. 1981. Do That display, in the same context, is absent in Ring Doves copulate more than necessary for S. bitorquata and S. tranquebarica. Absence of fertilization? Physiology and Behavior 27:659­ that display is shared with S. hypoppyrha, turtur, 662. and orientalis, the close relatives of bitorquata E n d l e r , J. A., a n d M. T h e r y . 1996. Interacting ef­ and tranquebarica. Likewise the bow-coo of fects of lek placement, display behavior, ambient those two species is very similar to that of S. light, and color patterns in three Neotropical hypoppyrha, turtur, and orientalis (S. de Kort un- forest-dwelling birds. American Naturalist 148: publ. data). In sum, despite the overall diver­ 421-452. sity and rapid evolution of vocalizations in F a r r i s , J. S., M. K allersjo, A. G. Kluge, and C. B u l t . 1994. Testing significance of congruence. Streptopelia, some behaviors reflect phylogenet­ Cladistics 10:315-320. ic relationships. F a r r i s , J. S., M. Kallersjo, A. G. Kluge, and C. B u l t . 1995. Constructing a significance test for A cknowledgments incongruence. Systematic Biology 44:570-572. Felsenstein, J. 1985. Confidence limits on phytoge­ We are grateful to the following individuals and nies: An approach using the bootstrap. Evolu­ institutions (and curators thereof) who were instru­ tion 39:783-791. mental in providing samples: American Museum of G a u n t , A. S., S. L. L. Gaunt, and R. M. C a s e y . 1982. Natural History, Field Museum of Natural History, Syringeal mechanisms reassessed: Evidence University of Kansas Museum of Natural History, from Streptopelia. Auk 99:474-494. Louisiana State University Museum of Natural Sci­ Georgiou, G. C., P. J. Sharp, and R. W. L e a . 1995. ence, U.S. National Museum, University of Washing­ (14C) 2-Deoxyglucose uptake in the brain of the ton Burke Museum, the Tracy Aviary, and Charlie Ring Dove (Streptopelia risoria). II. Differential Collins. The following individuals and organizations uptake at the onset of incubation. Brain Research provided field assistance in collecting specimens: Sa­ 700:137-141. rah Al-Tamimi, Jason Weckstein, Rob Moyle, Nigella G o o d w in , D. 1983. Pigeons and Doves of the World, Hilgarth, Tony Jones, Aldolfo Navarro, A. Townsend 3rd ed. Cornell University Press, Ithaca, New Peterson, Mark Robbins, Moussa Barka, Imar Mous- York. sa, Wil Silkens, Arizona Fish and Game, Texas Parks H ackett, S. J. 1996. Molecular phylogenetics and and Wildlife, and the U.S. Fish and Wildlife Service. biogeography of tanagers in the genus Rampho- M. Griekspoor provided sound recordings. Dave En- celus (Aves). Molecular Phylogenetics and Evo­ strom, Larry Page, Fred Sheldon, and two anony­ lution 5:368-382. 886 J o h n s o n e t a l . [Auk, Vol. 118

H a f n e r , M. S., P. D. Sudman, E X. Villablanca, T. W ils o n . 1989. Dynamics of mitochondrial DNA A. Spradling, J. W. Demastes, and S. A. N a d - evolution in animals: Amplification and se­ l e r . 1994. Disparate rates of molecular evolution quencing with conserved primers. Proceedings in cospeciating hosts and parasites. Science 365: of the National Academy of Sciences USA 86: 1087-1090. 6196-6200. Harshman, J. 1996. Phylogeny, evolutionary rates, L a d e , B. I., a n d W. H. T h o r p e . 1964. Dove songs are and ducks. Ph.D. dissertation. University of Chi­ innately coded patterns of specific behaviour. cago, Chicago. Nature 202:366-368. Hengeveld, R. 1993. What to do about the North L e a , R. W., G. C. G e o r g io u , Q. L i, R. J. S t e r l i n g , American invasion by the Collared Dove? Jour­ a n d P. J. S h a r p . 1995. (14C) 2-Deoxyglucose up­ nal of Field Ornithology 64:477-489. take in the brain of the Ring Dove (Streptopelia Hengeveld, R., a n d F. van den Bosch. 1991. The risoria). I. Prolactin-induced uptake. Brain Re­ expansion velocity of the Collared Dove Strep­ search 700:129-136. topelia decaocto population in Europe. Ardea 79: M archetti, K. 1993. Dark habitats and bright birds 67-72. illustrate the role of environment in species di­ Huelsenbeck, J. P., a n d K. A. C r a n d a l l . 1997. Phy­ vergence. Nature 362:149-152. logeny estimation and hypothesis testing using M cCracken, K. G., a n d F. H. S h e ld o n . 1997. Avian maximum likelihood. Annual Review of Ecolo­ vocalizations and phylogenetic signal. Proceed­ gy and Systematics 28:437-466. ings of the National Academy of Science USA 94: Ja n ik , D. S., a n d J. D. B u n t in . 1985. Behavioral and 3833-3836. physiological effects of prolactin in incubating M o r t o n , E. S. 1975. Ecological sources of selection doves ( Journal of Endocri­ Streptopelia risoria). on avian sounds. American Naturalist 108:17­ nology 105:201-210. 34. J o h n s o n , K. P. 2000. The evolution of courtship dis­ N e ls o n , D. A., a n d P. M a r l e r . 1990. The perception play repertoire size in the dabbling ducks (An- of bird song and an ecological concept of signal atini). Journal of Evolutionary Biology 13:636­ space. Pages 443-478 in Comparative Percep­ 644. tion, vol. 2: Complex Signals (W. C. Stebbins and Jo h n s o n , K. P., a n d D. H. C l a y t o n . 2000a. Nuclear M . A. Berkeley, Eds.). John Wiley and Sons, New and mitochondrial genes contain similar phy­ York. logenetic signal for pigeons and doves (Aves: N o w a k , E. 1975. Zur Systematik der Gattung Strep­ Columbiformes). Molecular Phylogenetics and topelia (Columbiformes, Aves). Bonner Zoologis- Evolution 14:141-151. che Beitraege 26:135-154. Jo h n s o n , K. P., a n d D. H. C l a y t o n . 2000b. A mo­ Pro AG, V. 1995. The Geology and Water Resources of lecular phylogeny of the dove genus Zenaida: Mi­ Mauritius. Mahatma Gandhi Institute, Moka, tochondrial and nuclear DNA sequences. Con­ Mauritius. dor 102:864-870. Prychitko, T. M., a n d W. S. Moore. 1997. The utility Jo h n s o n , K. P., a n d S. M. L a n y o n . 1999. Molecular systematics of the grackles and allies, and the ef­ of DNA sequences from an intron from the 0-fi­ fect of additional sequence (cyt b and ND2). Auk brinogen gene in phylogenetic analysis of wood­ 116:759-768. peckers (Aves: Picidae). Molecular Phylogenet­ J o h n s o n , K. P., a n d S. M. L a n y o n . 2000. Evolution­ ics and Evolution 8:193-204. ary change in color patches of blackbirds are as­ Ramos, C., and R. Silver. 1992. Gonadal hormones sociated with marsh nesting. Behavioral Ecolo­ determine sex differences in timing of incuba­ gy 11:515-519. tion by doves. Hormones and Behavior 26:586­ J o h n s o n , K. P., a n d M. D. S o r e n s o n . 1998. Com ­ 601. paring molecular evolution in two mitochondri­ R y a n , M. J., a n d E. A. Brenowitz. 1985. The role of al protein coding genes (cytochrome b and ND2) body size, phylogeny, and ambient noise in the in the dabbling ducks (Tribe: Anatini). Molecu­ evolution of bird song. American Naturalist 126: lar Phylogenetics and Evolution 10:82-94. 87-100. J o h n s t o n , R. F. 1962. The taxonomy of pigeons. Con­ Slabbekoorn, H ., S. de Kort, and C. ten Cate. dor 64:69-74. 1999. Comparative analysis of perch-coo vocal­ K a s p a r e k , M. 1996. Dispersal and range extension of izations in Streptopelia doves. Auk 116:737-748. the Collared Dove (Streptopelia decaocto) in Eu­ Slabbekoorn, H., a n d C. t e n C a t e . 1998. Multiple rope. Journal fur Ornithologie 137:1-33. parameters in the territorial coo of the Collared K l i c k a , J. T., a n d R. M. Z in k . 1997. The importance Dove: Interactions and meaning. Behaviour 135: of recent ice ages in speciation: A failed para­ 879-895. digm. Science 277:1666-1669. S o r e n s o n , M. D., a n d T. W. Q u in n . 1998. Numts: A Kocher, T. D., W. K. Thomas, A. Meyer, S. V. Ed­ challenge for avian systematics and population w ards, S. PAAbo, F. X. V illablanca, and A. C. biology. Auk 115:214-221. October 2001] Phylogeny of Streptopelia 887

Sturmbauer, C., a n d A. M e y e r. 1992. Genetic di­ W alker, L. E., J. M. W a l k e r , J. W. P alca, and R. J. vergence, speciation, and morphological stasis B e r g e r . 1983. A continuum of sleep and shallow in the lineage of African cichlid fishes. Nature torpor in fasting doves (Streptopelia risoria). Sci­ 358:578-581. ence 221:194-195. S w o f f o r d , D. L. 2000. PAUP*: Phylogenetic Analysis W ile y , R. H., a n d D. G. R ic h a r d s . 1978. Physical Using Parsimony, version 4.0. Sinauer, Sunder­ constraints on acoustic communication in the at­ land, Massachusetts. mosphere: Implications for the evolution of an­ ten Cate, C., R. W. L e a , M. R. Ballintijn, and P. J. imal vocalizations. Behavioral Ecology and So­ S h a r p . 1993. Brood size affects behavior, inter­ ciobiology 3:69-94. clutch interval, LH levels, and weight in Ring Z e n o n e , P. G., M. E. Sims, and C. J. E r ic k s o n . 1979. Dove (Streptopelia risoria) breeding pairs. Hor­ Male Ring Dove (Streptopelia risoria) behavior mones and Behavior 27:539-550. and the defense of genetic paternity. American T u b a r o , P. L., a n d B. M a h le r . 1998. Acoustic fre­ Naturalist 114:615-626. quencies and body mass in the New World doves. Condor 100:54-61. Associate Editor: F. Sheldon