© 2005 Society of Systematic Biologists
Is Homoplasy or Lineage Sorting the Source of Incongruent mtDNA and Nuclear Gene Trees in the Stiff-Tailed Ducks (Nomonyx-Oxyura)?
Edited by Allan Baker: Assiciate Editor
1 Institute of Arctic Biology, Department of Biology and Wildlife, and University of Alaska Museum, University of Alaska Fairbanks Fairbanks, Alaska 99775, USA fnkgm{at}uaf.edu
2 Department of Biology, Boston University Boston, Massachusetts 02215, USA
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
We evaluated the potential effects of homoplasy, ancestral polymorphism, and hybridization as obstacles to resolving phylogenetic relationships within Nomonyx-Oxyura stiff-tailed ducks (Oxyurinae; subtribe Oxyurina). Mitochondrial DNA (mtDNA) control region sequences from 94 individuals supported monophyly of mtDNA haplotypes for each of the six species and provided no evidence of extant incomplete lineage sorting or inter-specific hybridization. The ruddy ducks (O. j. jamaicensis, O. j. andina, O. j. ferruginea) are each others' closest relatives, but the lack of shared haplotypes between O. j. jamaicensis and O. j. ferruginea suggests long-standing historical isolation. In contrast, O. j. andina shares haplotypes with O. j. jamaicensis and O. j. ferruginea, which supports Todd's (1979) and Fjeldså's (1986) hypothesis that O. j. andina is an intergrade or hybrid subspecies of O. j. jamaicensis and O. j. ferruginea. Control region data and a much larger data set composed of
8800 base pairs of mitochondrial and nuclear sequence for each species indicate that the two New World species, O. vittata and O. jamaicensis, branch basally within Oxyura. A clade of three Old World species (O. australis, O. maccoa, O. leucocephala) is well supported, but different loci and also different characters within the mtDNA data support three different resolutions of the Old World clade, yielding an essentially unresolved trichotomy. Fundamentally different factors limited the resolution of the mtDNA and nuclear gene trees. Gene trees for most nuclear loci were unresolved due to slow rates of mutation and a lack of informative variation, whereas uncertain resolution of the mtDNA gene tree was due to homoplasy. Within the mtDNA, approximately equal numbers of characters supported each of three possible resolutions. Parametric and nonparametric bootstrap analyses suggest that resolution of the mtDNA tree based on 4300 bp per taxon is uncertain but that complete mtDNA sequences would yield a fully resolved gene tree. A short internode separating O. leucocephala from (O. australis, O. maccoa) in the best mtDNA tree combined with long terminal branches and substantial rate variation among nucleotide sites allowed the small number of changes occurring on the internode to be obscured by homoplasy in a significant portion of simulated data sets. Although most nuclear loci were uninformative, two loci supported a resolution of the Old World clade (O. maccoa, O. leucocephala) that is incongruent with the best mtDNA tree. Thus, incongruence between nuclear and mtDNA trees may be due to random sorting of ancestral lineages during the short internode, homoplasy in the mtDNA data, or both. The Oxyura trichotomy represents a difficult though likely common problem in molecular systematics. Given a short internode, the mtDNA tree has a greater chance of being congruent with the history of speciation because its effective population size (Ne) is one-quarter that of any nuclear locus, but its resolution is more likely to be obscured by homoplasy. In contrast, gene trees for more slowly evolving nuclear loci will be difficult to resolve due to a lack of substitutions during the internode, and when resolved are more likely to be incongruent with the species history due to the stochastic effects of lineage sorting. We suggest that researchers consider first whether independent gene trees are adequately resolved and then whether those trees are congruent with the species history. In the case of Oxyura, the answer to both questions may be no. Complete mtDNA sequences combined with data from a very large number of nuclear loci may be the only way to resolve such trichotomies.
Keywords: Ancestral polymorphism; homoplasy; lineage sorting; Nomonyx; Oxyura; Oxyurinae; stiff-tailed ducks; waterfowl
Received March 26, 2003; Revised December 4, 2003; Accepted September 10, 2004
Lack of resolution in a phylogenetic tree is represented as a polytomy in which three or more descendant lineages diverge from a single node. Systematists generally seek fully resolved trees, which may yield stronger inferences about character evolution or biogeographic history, and usually view polytomies as reflecting uncertainty about relationships. Such "soft" polytomies may be due to insufficient data and often can be resolved by adding more or different kinds of characters (Maddison, 1989; DeSalle et al., 1994). In contrast, a "hard" polytomy represents the simultaneous origin of three or more species (or gene lineages) from a common ancestor and has no bifurcating solution. Instances in which intervals between successive branching events are too short to accumulate informative variation also are effectively hard polytomies (Hoelzer and Melnick, 1994). Operationally, hard polytomies can be identified by internal branch lengths that do not differ significantly from zero (e.g., Walsh et al., 1999). The internal branch of a soft polytomy, however, might also meet this criterion if limited data are available. Thus, one test to discriminate between soft and hard polytomies is to add data and determine if significant internal branch lengths are obtained.
Resolution of a soft polytomy in a species-level phylogeny can be complicated further by incongruence of individual gene trees with the history of speciation (Tajima, 1983; Neigel and Avise, 1986; Avise et al., 1990). Consider, for example, a clade of three species (A, B, C) in which the divergence of B and C follows shortly after the divergence of A from the ancestor of B and C, such that ancestral polymorphism is retained in all three of the descendant lineages. If these events occurred very recently, incomplete lineage sorting may be detected with population level sampling. Over time, however, genetic drift will lead to the monophyly of alleles (or haplotypes) at each locus in each species, but due to the differential assortment of ancestral polymorphisms, species B and C will share closely related alleles at only some loci. By chance, species A and B will share alleles at another set of loci, and species A and C will share alleles at a third set of loci. If the internode (T) is very short (essentially a hard polytomy) and the effective population size (Ne) is large (e.g., 2Ne > T; Avise, 2000), and the speciation events occurred far enough in the past for lineage sorting to have completed, approximately 33% of gene trees from independent, polymorphic loci should support each of the three possible clades. As the length of the internode increases, however, an increasing proportion of gene trees should be congruent with the species history. Thus, a second test of soft versus hard polytomies is to survey multiple, unlinked loci.
Synapomorphies also can be obscured by homoplasy, which increases with time since speciation and rates of mutation. Homoplasy may be inconsequential for resolving polytomies in cases of very recent speciation, including those where retained ancestral polymorphism is still extant. As the time depth of a soft polytomy increases, however, homoplasy will tend to obscure relationships and increase the length of nucleotide sequence needed to resolve a given gene tree, regardless of whether that gene tree matches the species tree or not. For soft polytomies that occurred farther in the past, homoplasy may overwrite phylogenetic signal, such that each gene tree effectively becomes a hard polytomy with internal branch lengths that do not differ significantly from zero in a statistical framework.
Lineage sorting and homoplasy thus differ in their effects because the potential to resolve a polytomy depends not only on the length of the internode between speciation events but also on the overall time depth of the polytomy (e.g., O'hUigin et al., 2002). A soft polytomy that is one million years old now may be resolved with sufficient data, but 100 million years hence the same polytomy might meet all the criteria for a hard polytomy due to the cumulative effects of homoplasy. Ancestral polymorphism per se eventually disappears as each descendant species becomes fixed for a different set of alleles or haplotypes, but the mismatches between gene trees and species trees that ancestral polymorphism generates persist. We consider the above issues in attempting to resolve the species-level relationships of stiff-tailed ducks (subfamily Oxyurinae).
Stiff-Tailed Ducks
Stiff-tailed ducks traditionally have been classified as a monophyletic group of eight waterfowl species and four genera, three of which are monotypic: (1) black-headed duck (Heteronettaatricapilla), (2) masked duck (Nomonyxdominicus), (3) five or six Oxyura species, and (4) musk duck (Biziura lobata) (Delacour and Mayr, 1945; Raikow, 1970; Livezey, 1986, 1995; Johnsgard and Carbonell, 1996). Livezey (1997) restricted the term "stiff-tailed ducks" (subtribe Oxyurina) to Nomonyx, Oxyura, and Biziura, which along with Heteronetta compose the tribe Oxyurini ("stiff-tailed ducks and allies"). However, Harshman (1996), Sraml et al. (1996), and McCracken et al. (1999) demonstrated that Biziura is not a close relative of the other stiff-tailed ducks or their allies.
Excluding Biziura, six species of stiff-tailed ducks (Nomonyx, Oxyura) currently are recognized. Nomonyx dominicus inhabits tropical and subtropical wetlands of the New World and is the sister group of Oxyura (McCracken et al., 1999, Sorenson et al., in preparation; but see Livezey, 1995). Oxyura species are broadly distributed (Fig. 1), with Argentine lake duck (O. vittata) and ruddy duck (O. jamaicensis) in the New World; and white-headed duck (O. leucocephala), Australian blue-billed duck (O. australis), and maccoa duck (O. maccoa) in the Old World. The South American ruddy duck subspecies, O. j. andina and O. j. ferruginea, are endemic to the Andes, O. j. andina in Colombia and O. j. ferruginea from southern Colombia south to Tierra del Fuego.
|
Despite three previous morphological and molecular studies (Raikow, 1970; Livezey, 1995; McCracken et al., 1999), species-level relationships within Oxyura remain poorly resolved. McCracken et al. (1999) found strong support for Nomonyx as the sister group of Oxyura, but a small cytochrome b data set (1045 bp) and short internal branches (McCracken et al., 1999: Fig. 2, Fig. 3) yielded weak support for relationships within Oxyura. Behavioral and morphological character data (McCracken et al., 1999: Fig. 9) also provided no conclusive support. Parsimony analysis of geographical range data supported a clade of Old World species (O. australis, O. maccoa, O. leucocephala) with the New World species (Nomonyx, O. vittata,O. jamaicensis) branching basally, but McCracken et al. (1999) considered the short internodes within Oxyura essentially unresolved.
|
|
As outlined above, poor resolution of Oxyura phylogeny may stem from insufficient character data, the presence of hard polytomies, retained ancestral polymorphism, incongruence among data partitions due to the stochastic nature of the lineage sorting process, or homoplasy. Current or historical hybridization also might complicate resolution of the phylogeny. For example, four Nomonyx-Oxyura stiff-tailed ducks are sympatric in South America (Fig. 1). In Colombia, where the ranges of O. j. andina and O. j. ferruginea abut, O. j. andina shows variable amounts of white cheek spotting (Adams and Slavid, 1984; Fjeldså, 1986). Todd (1979) believed O. j. andina to be an intergrade, and Fjeldså (1986) suggested that it might represent a relict Pleistocene hybrid population of O. j. jamaicensis and O. j. ferruginea, with males approaching one or the other phenotype. Livezey (1995) classified O. j. andina as a subspecies of O. jamaicensis, but classified O. j. ferruginea as a separate species. This study is the first to assay genetic diversity of O. j. jamaicensis, O. j. andina, and O. j. ferruginea. Hybridization might also occur elsewhere in South America. O. j. ferruginea and O. vittata occur sympatrically in an ecotone at the base of the southern Andes, O. j. ferruginea and Nomonyx occur sympatrically in the central Andes, and Nomonyx and O. vittata occur sympatrically in the lowland subtropics of South America, as do Nomonyx and O. j. jamaicensis in Central America and the Caribbean (Johnsgard, 1965, 1978; Johnsgard and Carbonell, 1996). Elsewhere, geographic ranges of Oxyura species are widely separated, except in Europe where O. j. jamaicensis and O. leucocephala now hybridize in a human-mediated contact zone caused by the escape of captive O. j. jamaicensis from North America (Green and Hughes, 1996; Hughes et al., 1999).
To evaluate the potential effects of hybridization, homoplasy, ancestral polymorphism, and lineage sorting on our efforts to resolve the relationships of stiff-tailed ducks, we collected tissues from 94 individuals representing all Nomonyx and Oxyura species (including O. j. andina and O. j. ferruginea) and studied their phylogenetic relationships at two hierarchical levels. (1) We sequenced most of the mitochondrial (mtDNA) control region and part of the phenylalanine t-RNA gene from all 94 samples to test species monophyly and compare genetic divergence within and among species and among subspecies of O. jamaicensis. (2) We also sequenced 12S rRNA, ND2, part of ND5, cytochrome b, and a number of t-RNA genes from the mtDNA as well as introns and exons from 10 different nuclear loci from one representative of each Nomonyx-Oxyura species. We sought to determine if polytomies within Oxyura are soft or hard, whether character conflict occurs primarily within or between data partitions, and how much data would be required to resolve the relationships of this group. We also used parametric bootstrapping to evaluate the relationship between data set size and support for alternative arrangements within Oxyura.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
Collection, Amplification, and Sequencing
Collecting information for 94 Nomonyx-Oxyura genetic samples is provided in Appendix 1. DNA was isolated from muscle, blood, or the base of one or two feather quills using DNeasy Tissue Kits (QIAGEN, Valencia, California). Thirty microliters of 100 mg/mL DTT (dithiothreitol) was added to the digestion buffer to dissolve feather quills. We amplified and sequenced most of the mtDNA control region and part of the phenylalanine t-RNA gene (corresponding to bp 79 to 1250 in the chicken [Gallus gallus] mitochondrial genome; Desjardins and Morais, 1990). Our sequences span most of domain I (5' variable region) and all of domains II (conserved central domain) and III (3' variable region) of the avian control region (e.g., Marshall and Baker, 1997). Primer sequences used in this study are listed in Appendix 2. We amplified fresh tissues using the overlapping primer pairs L78–H774 and L736–H1251. We used L78–H493, L432–H774, and NODOCRL1–NODO12SH to amplify the same region from one Nomonyx museum skin (LSUMZ 123431) and L78–H493 (or L78–OXJACRH1 and OXJACRL1–H493) to amplify 5' control region sequences from three Nomonyx, one O. j. jamaicensis, five O. j. andina, and four O. j. ferruginea museum skins.
We also sequenced 12S rRNA, ND2, part of ND5, cytochrome b, and parts of the adjacent phenylalanine, valine, methionine, tryptophan, and threonine t-RNA genes (chicken mtDNA genome positions 1268 to 2293, 5217 to 6312, and 14771 to 16063; Desjardins and Morais, 1990) from one individual of each species included in the control region data set (but not including O. j. andina or O. j. ferruginea). Each region was amplified and sequenced in two or more overlapping fragments (see Appendix 2 for primers used). To this data set we also added introns and exons from ten nuclear loci: chromosome Z chromo-ATPase/helicase DNA-binding protein (CHD-Z), 
-enolase (-ENOL), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), hemoglobin -A (hemoglobin -A), lactate dehydrogenase-B (LDHB), lamin A (lamin-A), laminin receptor precursor/p40 (LRP/p40), myelin proteolipid protein (MPP), and phosphoenolpyruvate carboxykinase introns 3 and 9 (PEPCK3, PEPCK9). Hemoglobin -A sequences included three exons and two introns; all other nuclear sequences are introns with adjacent portions of the flanking exons. Our choice of nuclear loci follows previous studies (e.g., Friesen et al., 1997; Primmer et al., 2002; Pacheco et al., 2002), but we designed our own primers for each locus (Appendix 2) based on sequences for chicken and other vertebrates in GenBank. The LDHB sequence we obtained for Nomonyx was only 118 bp, and translation of the sequence indicated complete deletion of intron 3 in this species.
PCR reactions were carried out in GeneAmp PCR System 2400 and 9600 (Applied Biosystems, Foster City, California) and DNA Engine DYAD (MJ Research, Waltham, Massachusetts) thermalcyclers using 50-µL reactions containing 2 µ L template DNA, 2.5 µ L of each primer (10 µM), 5 µL of 10 µM dNTPs, 5 µL of 25 mM MgCl2, 5 µL of 10x PCR buffer, and 0.25 µ L Taq Polymerase. We used AmpliTaq Gold PCR Master Mix (Applied Biosystems, Foster City, California) for the 10 nuclear loci. Thermal cycling typically was as follows: 7 min preheat at 94°C, followed by 45 cycles of 20 s at 94°C, 20 s at 52°C, 1 min at 72°C, and a final extension of 7 min at 72°C. We used nested PCR for the PEPCK introns. Product from an initial 20 µ l PCR using the outer primers PEPCK3.F–PEPCK3.R and GTP1601.F–GTP1793.R was used as the template for a second PCR using the inner primers PEPCK3.F2–PEPCK3.R and PEPCK9.F–PEPCK9.R (Appendix 2). PEPCK3.F2 was paired with the same reverse primer as in the first PCR.
PCR products were electrophoresed in 1% to 2% agarose, excised from the gel, and purified using QIAquick Gel Extraction Kits (QIAGEN, Valencia, California). Both strands of each product were cycle-sequenced in 10-µL reactions using BigDye Terminator Cycle Sequencing Kits diluted fourfold, followed by electrophoresis on ABI 377 or ABI 3100 automated DNA-sequencers (Applied Biosystems, Foster City, California). Sequences from opposite strands were reconciled and verified for accuracy using Sequencher 3.1 (Gene Codes, Ann Arbor, Michigan) and Sequence Navigator 1.0.1 (Applied Biosystems, Foster City, California). Sequences are archived in GenBank: AY747698 [GenBank] –AY747869.
Phylogenetic Analysis of Control Region Data
Control region sequences varied in length among species due to insertions and deletions of nucleotides. We aligned control region sequences using direct optimization as implemented in POY 2.7 (Gladstein and Wheeler, 2000). Unlike other alignment algorithms such as Clustal W (Thompson et al., 1994), direct optimization considers sequence alignment and tree topology using a single optimality criterion (i.e., minimizing the number of steps on a tree), and, as such, is a logically consistent approach to phylogenetic analysis of variable length sequences (Wheeler, 1996).
Alignment of control region sequences was performed in two steps. We first used POY 2.7 in an analysis including each unique Oxyura haplotype using 100 random addition replicates (each limited to five trees), equal weights for all changes, tree bisection and reconnection (TBR) branch-swapping, and an insertion-deletion cost equal to one. Using the implied alignment generated by POY as a guide, we next optimized the alignment for the shortest tree found by POY using methods described by Sorenson and Payne (2001). Two long and unambiguous insertions observed in Nomonyx (see results) were uninformative with respect to relationships within Oxyura and were therefore removed from all subsequent analyses. Remaining, generally single-base indels were in included in parsimony analyses by treating gaps as a fifth character state. We used heuristic tree searches with random addition of taxa in PAUP* 4.0 (Swofford, 1998) to find all equally parsimonious trees for the optimized static alignment. Given uncertain resolution of the many closely related haplotypes within species, we reduced the number of trees found by collapsing all branches with a minimum length of zero. Decay indices (Bremer, 1988) were determined using TreeRot (Sorenson, 1999). The control region alignment is available for download in nexus format online at http://mercury.bio.uaf.edu/~kevin_mccracken/alignments/ and http://systematicbiology.org
We used hierarchical likelihood ratio tests as implemented in Modeltest 3.06 (Posada and Crandall, 1998) to determine the appropriate model of sequence evolution, which for the control region was the Hasegawa et al. (1985) model with gamma-distributed rate variation among sites ( = 0.20, ti/tv ratio = 3.41, A = 0.2781, C = 0.3111, G = 0.1513). With these parameter values fixed, we completed 100 full heuristic searches to find maximum likelihood tree(s) for 39 unique control region haplotypes. We used bootstraping to measure support for clades (Felsenstein, 1985) under both parsimony (1000 replicates, each with 100 random addition replicates limited to 5 trees each) and maximum likelihood (100 replicates with TBR branch-swapping) criteria. Posterior branch probabilities were obtained using MrBayes 3.0 (Huelsenbeck and Ronquist, 2001). We used the HKY+G model and completed two replicate analyses each with four Markov chains of 2.5 million generations. Trees were sampled every 1000 generations, and posterior branch probabilities were calculated after excluding the first 0.5 million generations. Finally, we used statistical parsimony as implemented in TCS (Clement et al., 2000) to illustrate the 95% set of unrooted trees for O. j. jamaicensis, O. j. andina, and O. ferruginea. Maximum likelihood pairwise genetic distances within and between each species were calculated using PAUP* 4.0 (Swofford, 1998).
Phylogenetic Analysis of Additional mtDNA and Nuclear Data
Sequence data from one representative of each Nomonyx-Oxyura species were assembled to generate a larger mtDNA plus nuclear data set that included portions of twenty genes from ten different linkage groups: mtDNA (control region, tRNA-Phe, 12S rRNA, tRNA-Val, tRNA-Met, ND2, tRNA-Trp, ND5, cytochrome b, tRNA-Thr), CHD-Z, -ENOL, GAPDH, hemoglobin -A, LDHB, lamin-A, LRP/p40, MPP, and PEPCK introns 3 and 9. Alignments of all gene regions except the control region (see above) were unambiguous and made by eye. Multiple base indels in the control region (69 bp and 123 bp), LDHB (466 bp, 8 bp, and 3 bp), LRP/p40 (4 bp), and MPP (8 bp) were coded as missing data, and new binary characters for each unique gap (0 = absent, 1 = present) were added to the end of the data matrix. All other gaps (1 to 3 bp in length) were coded as a fifth character state. A total of 8608 characters were derived from 8801 aligned positions. Double peaks in nuclear gene sequences, reflecting heterozygous positions, were coded with IUPAC degeneracy codes and were treated as polymorphisms. All trees were rooted with Nomonyx as the outgroup.
We used an exhaustive tree search and unweighted parsimony to find the most parsimonious tree(s) for the combined data set and explore suboptimal trees. We identified characters supporting three possible branching patterns among the Old World Oxyura species, O. australis, O. maccoa, and O. leucocephala, and compared these characters in terms of sequence position, substitution types, numbers of steps, and consistency indices.
We also completed maximum likelihood and Bayesian analyses on the combined mitochondrial data set (4335 characters) using both a single model of sequence evolution fitted to the entire data set or mixed models in which separate parameters were estimated for each of several different data partitions (see results). The fit of mixed models was assessed using the Akaike information criterion (AIC) (Akaike, 1973), and alternative tree topologies were compared using the approximately unbiased test of Shimodaira (2002). As described for the control region analysis above, the Bayesian analyses were replicated and each was run for 2.5 million generations, of which the first 0.5 million were discarded before determining posterior branch probabilities.
Finally, we used parametric bootstrapping (e.g., Saitou and Nei, 1986) to measure the power of the mtDNA data set to resolve relationships among three Old World Oxyura species in the mtDNA haplotype tree if, as indicated by our analysis, a short internode separates O. leucocephala from O. maccoa and O. australis. Using 4070 bp of aligned mtDNA sequence for which there were no gaps or missing data in any taxa, we estimated maximum likelihood parameters for the GTR+I+G model on the most parsimonious mtDNA tree using PAUP* 4.0 (Swofford, 1998). We then used Seq-Gen 1.2.6 (Rambaut and Grassly, 1997) to simulate the evolution of sequences on the same tree with branch lengths and model parameters estimated from the empirical data. We generated 500 data sets for each of four sequence lengths: 2 kb, 4 kb, 8 kb, and 16 kb. Simulated data were analyzed using maximum likelihood under the same model of sequence evolution, and the proportion of replicate data sets yielding each of the nodes in the original tree was determined. We also determined the "expected" distribution of parsimony bootstrap values for each of the three possible resolutions of the trichotomy by conducting a complete bootstrap analysis with 100 replicates and branch and bound tree searches for each of the simulated data sets.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
Control Region Sequence Diversity
Excluding within-individual length variation in Nomonyx, we found 39 unique haplotypes composed of 900 to 1097 nucleotides in 94 individuals. Two large nucleotide insertions (69 bp at positions 152 and 15 to 123 bp at positions 951) occurred in Nomonyx. Four Nomonyx individuals collected in Argentina apparently possessed multiple control region copies that varied in the length of the insertion at position 951 ranging from 15 to 73 bp. Multiple bands generated by PCR amplification of this region were isolated from agarose gels and sequenced, yielding six clean sequences from four individuals; three had 15-bp insertions at position 951, and three had 73-bp insertions. Although PCR artifact is a possible explanation for this length variation, the same results were produced repeatedly using four different overlapping primer combinations (L78–NODO12SH, L78–H2294, NODOCRL1–NODO12SH, NODOCRL1–H2294). Additional length heteroplasmy near position 951 might very well exist in some Nomonyx as we were not able to sequence all of the bands generated by PCR amplification of this region. Excluding these two insertions in Nomonyx, the alignment was 924 positions, of which 207 (22.4%) varied among haplotypes and 188 (20.3%) were parsimony informative. Among these, transitions occurred at 131 (63.3%) positions, transversions occurred at 57 (27.7%) positions, and indels occurred at 37 (17.9%) positions. As in other birds, the distribution of variable sites across the stiff-tailed duck control region showed considerable heterogeneity: 35.6% of substitutions and indels occurred within the first 341 base pairs of the 5' end of the alignment, and 61.2% occurred within 558 base pairs of the 3' end. Only 3.2% of substitutions and indels occurred in the intervening 217 base pairs. These divisions correspond to domains I, II, and III of Marshall and Baker (1997). Four conserved sequenced motifs reported for birds (Desjardins and Morais, 1990; Quinn and Wilson, 1993; Marshall and Baker, 1997; Crochet and Desmerais, 2000) also were observed in our sequences: F box at positions 227 to 254; D box at positions 333 to 357; C box at positions 382 to 408; and CSB-1 at positions 690 to 716 in the 924-position alignment excluding the two large gap regions in Nomonyx. One to 10 unique haplotypes were observed within each species or subspecies. Intra- and interspecific pairwise genetic distances are summarized in Table 1. Variation within species, identical sequences in overlapping regions amplified using different combinations of primers, and matching sequences from muscle, feather, and blood samples support the conclusion that our sequences are not nuclear copies (see Sorenson and Quinn, 1998).
|
Control Region Haplotype Relationships
Optimization alignment and unweighted parsimony analysis of 39 unique Nomonyx-Oxyura haplotypes produced 40 most parsimonious trees (length = 279, CI = 0.81, RI = 0.95). Interspecific relationships were as follows in 100% of the 40 most parsimonious trees as well as in two maximum likelihood trees and the Bayesian consensus tree: Nomonyx (O. vittata (O. jamaicensis (O. leucocephala (O. australis, O. maccoa)))). One of two equally likely trees (lnL = –2510.42) is illustrated in Figure 2 (the two maximum likelihood trees were identical, respectively, to 2 of the 40 most parsimonious trees). Monophyly of Nomonyx and monophyly of each Oxyura species was strongly supported (97% to 100% bootstrap values or posterior branch probabilities). There is no evidence of incomplete lineage sorting or hybridization among species: internal branches between species were long compared to branches within species. Branch support for interspecific relationships was more varied, but the control region data yielded generally higher support indices than cytochrome b (see McCracken et al., 1999: Fig. 2) as well as a more stable topology. In the present analysis, three Old World species (O. maccoa, O. australis, O. leucocephala) form a clade nested within New World lineages (Nomonyx, O. vittata, O. jamaicensis), reconfirming a New World origin for this group.
Oxyura j. jamaicensis, O. j. andina, and O. j. ferruginea
Two different patterns of O. jamaicensis haplotype relationships occurred in the 40 most parsimonious trees. Twenty of 40 trees (50%), including both maximum likelihood trees (Fig. 2), included a monophyletic O. j. ferruginea nested within O. jamaicensis. Bootstrap support for O. j. ferruginea monophyly was low (29%). The remaining 20 trees (50%) had one O. j. jamaicensis haplotype (OXJA-02) and one O. j. andina haplotype (OXAN-01) nested within the O. j. ferruginea clade. O. j. jamaicensis was paraphyletic in all 40 trees. An unrooted parsimony tree showing the 95% statistical parsimony set (Clement et al., 2000) of possible haplotype relationships within O. jamaicensis is illustrated in Figure 3. Uncertainty about relationships, reflected by multiple connections between six haplotypes in this network, is due in part to partial sequence data for museum specimens. Two results are worth noting: (1) Four O. j. ferruginea from southern Argentina are identical and differ by two nucleotide substitutions from eleven O. j. ferruginea collected in Ecuador, Perú, and Bolivia; and (2) five O. j. andina collected from Colombia show four different haplotypes. Three O. j. andina haplotypes are identical to O. j. jamaicensis haplotypes, one is identical to an O. j. ferruginea haplotype found in Perú and Bolivia, and one is unique.
Combined Analyses of Additional mtDNA and Nuclear Data
To further test interspecific relationships within Oxyura, we combined control region data with three additional mtDNA regions. Sequence variation in cytochrome b and ND2 were comparable to the control region in terms of number of variable and informative sites and average genetic distances between species, whereas 12S rRNA showed lower levels of variation (Table 1). Consistency and rescaled consistency indices were highest for the control region. This may be due in part to somewhat lower rate variation among sites in the control region but perhaps more significantly due to a lower transition:transversion ratio than in the other genes (Table 1). With higher transition:transversion ratios, multiple substitutions at a site are more likely to result in homoplasy.
Combined analysis of the entire mtDNA data set resulted in the same species-level topology as in the control region analysis but with increased support indices at most nodes, suggesting generally congruent phylogenetic signal among data partitions (compare Fig. 2 and Fig. 4a; lnL = –9600.54; Table 2, Table 3). Decay indices increased from 2, 4, and 1 to 11, 10, and 2, respectively, for nodes within Oxyura. The maximum likelihood bootstrap value and Bayesian posterior branch probability were slightly lower in the combined analysis only for the node uniting O. australis and O. maccoa. This sister relationship and the same overall tree topology was supported in a combined parsimony analysis of the three protein-coding genes (ND2 and ND5/cytochrome b), whereas the 12S rRNA data partition supported a sister relationship between O. leucocephala and O. australis. Mixed models with parameters estimated independently for different data partitions provided a better fit to the data as measured by the AIC but did not change phylogenetic inferences. The same topology found in the combined parsimony analysis was found in both maximum likelihood and Bayesian analyses using mixed models (Fig. 4, Fig. 5). Although all mtDNA data partitions contributed to the two basal nodes in the Oxyura phylogeny, likelihood scores suggest that the control region data were most discriminating with respect to relationships of the three Old World Oxyura. The approximately unbiased test of Shimodaira (2002) indicated that no data partitions or mixed models discriminated significantly among the three possible resolutions of the trichotomy, except for a marginally significant rejection of (O. maccoa, O. leucocephala) for the control region data (Table 2).
|
|
|
|
Combined data analysis of 8608 characters from 20 mtDNA and nuclear genes yielded 196 informative characters and two equally parsimonious trees (length = 1013, CI = 0.84, RI = 0.34; Fig. 5). In one tree, O. australis is sister to a clade composed of O. maccoa and O. leucocephala, whereas the other tree has O. leucocephala sister to O. maccoa plus O. australis. The next most parsimonious tree was two steps longer (length = 1015, CI = 0.84, RI =0.33; tree not shown) and included the third possible resolution of the above trichotomy. All three trees had the same basal relationships: (Nomonyx (O. vittata, (O. jamaicensis, (O. australis, O. maccoa, O. leucocephala)))).
Bootstrap support for the three possible resolutions of the Old World clade differed between the combined analysis and separate analyses of mtDNA and nuclear data (Table 3). In the combined analysis, none of the three clades was well supported: 41% for (O. maccoa, O. leucocephala), 35% for (O. australis, O. maccoa), and 20% for (O. australis, O. leucocephala). Based on mtDNA only, bootstrap support for these three alternative clades was 57% for (O. australis, O. maccoa), 35% for (O. australis, O. leucocephala), and 0% for (O. maccoa, O. leucocephala). In contrast, the nuclear data set yielded strong bootstrap support (92%) for (O. maccoa, O. leucocephala) and essentially zero support for the two alternative clades. This result reflects primarily the lack of characters supporting the alternative clades.
In parsimony analyses with heterozygous positions treated as polymorphisms, five characters in three different introns supported (O. maccoa, O. leucocephala), whereas no nuclear characters supported either of the alternative resolutions of the Old World trichotomy. For the -ENOL intron, however, the inference changes when the data for O. jamaicensis are coded as two separate sequences that differ at a single position in the diploid sequence. When this is done, O. jamaicensis shares one allele with O. vittata and O. australis and another allele with O. maccoa. The sequence in O. leucocephala is unique but shares a derived change with O. maccoa and one of the two O. jamaicensis alleles. In total, only 14 parsimony informative characters were obtained from 4274 aligned positions of nuclear sequence, and five loci (LDHB, LRP/p40, lamin-A, GAPDH, PEPCK3) included no variation that was informative about relationships at any level within Oxyura. At 8 of 10 nuclear loci, two or more Oxyura species shared identical sequences, suggesting that retention of ancestral sequences was due primarily to low mutation rates. Only PEPCK9 and MPP yielded fully resolved gene trees within Oxyura. These trees agreed in supporting (O. maccoa, O. leucocephala) but differed in other respects (Fig. 6).
|
Table 4 lists individual characters that support each of the three possible Old World clades. Interestingly, character support for (O. maccoa, O. leucocephala) was distributed among eight genes, including three nuclear introns, whereas character support for the other two clades was distributed among three and four genes, all mitochondrial. Thirteen of 29 characters (45%) supporting (O. australis, O. maccoa) or (O. australis, O. leucocephala) were in the control region, whereas only 1 character supporting (O. maccoa, O. leucocephala) was in the control region.
|
Nearly equal numbers of mitochondrial characters supported (O. australis, O. maccoa) (n = 15) and (O. australis, O. leucocephala) (n = 14), whereas somewhat fewer characters (n = 10) supported (O. maccoa, O. leucocephala). A single transversion in the control region supported (O. australis, O. maccoa); all other changes were transitions. In protein-coding genes, all but two characters were in third codon positions. The overall consistency of characters supporting the three alternative clades was on average slightly lower for the set of characters supporting (O. australis, O. maccoa). This issue was somewhat difficult to evaluate, however, in that five of the seven control region characters supporting (O. australis, O. leucocephala) were in portions of the control region that vary in length among taxa, such that homologous nucleotide positions in somewhat more divergent taxa (e.g., Stictonetta, Heteronetta) could not be identified. This prevented us from obtaining better estimates of character consistency by evaluating each character across a broader sample of waterfowl taxa.
Parametric Bootstrap Analysis of the mtDNA Data
The mtDNA maximum likelihood tree used to estimate branch lengths for the parametric bootstrap analyses is shown in Figure 4b. If this is the true haplotype tree and the actual branch lengths were as reconstructed, there is only a 72% chance of recovering the correct topology for the three Old World taxa with 4 kb of sequence data (Table 5). The proportion of data sets in which incorrect nodes are recovered declines with the size of the simulated data set until with 16-kb sequences less than 3% of data sets yield an incorrect resolution of the Old World trichotomy (Table 5). Although the (O. australis, O. maccoa) node is present in most trees derived from the simulated data sets, standard bootstrap values for this node vary greatly among data sets (Fig. 7). Strong bootstrap support for the correct node (
70%) increases steadily with data set size (9%, 20%, 33%, and 64% of 2-, 4-, 8-, and 16-kb data sets, respectively). Greater than 50% bootstrap support for an incorrect resolution of the trichotomy, however, is obtained in about 15% of the 2- to 8-kb data sets and exceeds 70% in 3% to 4% of data sets. With 16 kb, however, less than 1% of data sets yield 70% support for an incorrect node.
|
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
Systematic Relationships and Historical Biogeography
Monophyly of all Nomonyx-Oxyura species was robustly supported by control region data with bootstrap values ranging from 97% to 100% (Fig. 2). Internal branches between species and branches leading to each species clade were long compared to branches within each species. We found no evidence in the mtDNA data of incomplete lineage sorting or inter-specific hybridization based on the individuals sampled in this study.
Analyses of the control region and combined data sets (Fig. 2, Fig. 4, Fig. 5) agree on two points: Nomonyx is substantially diverged from Oxyura, and the two New World Oxyura species, O. vittata and O. jamaicensis, are basal to the three Old World species, O. australis, O. maccoa, and O. leucocephala. The ancestral geographic range of Nomonyx-Oxyura probably was lowland South America because most basal stiff-tailed ducks, Nomonyx and O. vittata (and also Heteronetta), are endemic to lowland South America, Central America, or the Caribbean.
The ruddy ducks, O. j. jamaicensis,O. j. andina, and O. j. ferruginea, are each others' closest relatives (Fig. 2, Fig. 3). Like most authors, we consider these three taxa subspecies based on their close genetic, morphological, and behavioral similarities (Johnsgard, 1965; Todd, 1979; Adams and Slavid, 1984; Fjeldså, 1986; Johnsgard and Carbonell, 1996; McCracken et al., 1999). In contrast to Livezey (1995), who placed O. ferruginea as the sister group of O. vittata and O. australis, our analyses show clearly that it is very closely related to other ruddy ducks. Monophyly of O. j. ferruginea mtDNA haplotypes also was supported, albeit with weak support. The lack of shared haplotypes between O. j. jamaicensis and O. j. ferruginea, however, suggests a strong degree of historical isolation between these subspecies. The O. j. ferruginea haplotype group was either nested within a larger clade of O. j. jamaicensis haplotypes or together with one O. j. jamaicensis and O. j. andina haplotype formed a sister group to all other O. jamaicensis haplotypes—both results were found among the set of 40 most parsimonious control region trees (Fig. 2, Fig. 3). Three O. j. andina had haplotypes also found in O. j. jamaicensis, whereas one was identical to a O. j. ferruginea haplotype, and one was unique, supporting Todd's (1979) and Fjeldså's (1986) hypothesis that O. j. andina is either an intergrade subspecies or hybrid population of O. j. jamaicensis and O. j. ferruginea, with males approaching one or the other in phenotype and genotype. A further test of this hypothesis would be to genotype O. j. andina samples for single nucleotide polymorphisms (SNPs) or other nuclear markers that distinguish O. j. jamaicensis and O. j. ferruginea to determine if the O. j. andina genome represents the admixture of O. j. jamaicensis and O. j. ferruginea genes. Likewise, more O. j. ferruginea samples are required from Patagonia and Tierra del Fuego to determine if southern populations of O. j. ferruginea are distinct from populations inhabiting the northern Andes. Our data indicate that the four individuals we sampled from Argentina's Patagonian provinces of Neuquén and Chubut differ from other O. j. ferruginea by two nucleotide differences (Fig. 3).
Two biogeographic hypotheses for O. jamaicensis are potentially consistent with our results. First, the evolution of O. jamaicensis from ancestral Oxyura might have followed the colonization of the Northern Hemisphere and establishment of a migratory population. Much more recently (given the limited divergence among O. jamaicensis subspecies) and perhaps in response to Pleistocene climate change, wintering O. j. jamaicensis from North America might have established local breeding populations in the Northern Central Andes, extended their range southwards, and subsequently colonized most of the Andean Cordillera and adjacent steppe and coastal habitat "top-down" from north to south and from high elevation to low elevation. Southern populations of O. j. ferruginea subsequently diverged as O. j. jamaicensis resident in the northern Andes expanded their range southwards and evolved phenotypic differences such as the all-black head, larger body size, and reduced spotting on the breast (Johnsgard, 1965; Fjeldså, 1986; Johnsgard and Carbonell, 1996). This hypothesis is consistent with the somewhat greater haplotype diversity found in O. j. jamaicensis and the nesting of O. j. ferruginea haplotypes within a larger clade of O. jamaicensis haplotypes (Fig. 2, Fig. 3), the result found in 50% of the most parsimonious trees.
Alternatively, ancestral O. jamaicensis might have evolved in the Southern Hemisphere and only recently colonized the Northern Hemisphere. The discovery of additional haplotype diversity and deeper divergences among O. j. ferruginea populations in South America would support this hypothesis. In either model, subsequent hybridization of the two subspecies in Colombia, where the winter range of O. j. jamaicensis abuts the range of O. j. ferruginea, likely explains the intermediate phenotypic traits (Adams and Slavid, 1984) and divergent haplotypes present in a hybrid O. j. andina population.
The above models have different implications for the route by which ancestral Oxyura reached the Old World. If O. jamaicensis colonized the Northern Hemisphere early in its history, then the common ancestor of O. leucocephala, O. maccoa, and O. australis might also have been a migratory species in Eurasia. If O. jamaicensis was restricted to the Southern Hemisphere until recently, as in the alternative model, then the common ancestor of Old World Oxyura might have originated in Africa or perhaps Australia. The almost contemporaneous divergence of the three Old World species from a common ancestor is consistent with either an "out-of-Africa" hypothesis, in which an ancestral O. maccoa population colonizes both Eurasia and Australia at about the same time, or a "north-to-south" hypothesis similar to that proposed above for O. jamaicensis, in which a historically widespread and migratory O. leucocephala ancestor in Eurasia colonizes Africa and Australia at about the same time. Distinguishing between these scenarios is complicated by the difficulty of resolving the relationships among these three species (see below). Nonetheless, we can conclude unambiguously that the divergence of O. j. jamaicensis and O. j. ferruginea is far more recent than the divergence of ancestral O. jamaicensis from ancestral Oxyura.
Conflicting mtDNA and Nuclear Characters
Control region data suggest that Eurasian O. leucocephala is basal to an Australian–African clade (O. australis, O. maccoa), but this clade is not well supported (46% to 81% parsimony and likelihood bootstrap; Fig. 2, Tables 2, Tables 3). Increasing the size of the data set to 8801 positions from 20 different mtDNA and nuclear loci (including the mtDNA control region) yielded only 196 informative characters and also failed to resolve the Oxyura trichotomy (Fig. 5, Table 3). Two alternative rearrangements, (O. maccoa, O. leucocephala) and (O. australis, O. maccoa) were equally parsimonious and had similar parsimony bootstrap support, 41% and 35%. The mtDNA data alone yielded 57% to 64% parsimony and maximum likelihood bootstrap support for (O. australis, O. maccoa), and 31% to 35% for (O. australis, O. leucocephala), and essentially no support for (O. maccoa, O. leucocephala). In contrast, nuclear data yielded 92% parsimony bootstrap support for (O. maccoa, O. leucocephala) but no support for the other two possible resolutions of the trichotomy.
Analysis of characters supporting the three possible resolutions of this trichotomy indicate that different loci and different characters within the same loci support different clades (Table 4). Within the mtDNA control region, six characters support (O. australis, O. maccoa) and seven characters support (O. australis, O. leucocephala), but only one character supports (O. maccoa, O. leucocephala). The protein-coding genes ND2, ND5, and cytochrome b, also show similar patterns of conflicting character support, with eight, nine, and five characters supporting the three clades. One character from the noncoding 12S rRNA gene supports (O. maccoa, O. leucocephala), and two characters support (O. australis, O. leucocephala). In contrast, the two introns (MPP, PEPCK9; Fig. 6) with synapomorphies relevant to the Oxyura trichotomy support (O. maccoa, O. leucocephala). Given limited homoplasy in the nuclear data, the nuclear topologies can probably be regarded as reasonable accurate gene trees that are in conflict with the results of the mtDNA analysis, particularly with respect to the relationships of these three species. Combining these conflicting data from different linkage groups into a single analysis resulted in reduced levels of bootstrap support.
The Old World Oxyura Trichotomy
Despite collecting more than 8000 base pairs of sequence data per taxon, we are left with an unresolved trichotomy among the three Old World Oxyura species. Two distinct questions can be considered in attempting to resolve this kind of trichotomy, and both appear to be relevant to the present example: (1) Have individual gene trees been adequately resolved? (2) Are the individual gene trees congruent with the species history?
In the present example, different factors limited the resolution of the mtDNA and nuclear gene trees, respectively. Gene trees for most nuclear loci were unresolved primarily due to a lack of informative variation. Two or more species shared identical sequences at eight of ten nuclear loci. For many loci, retention of ancestral sequences does not necessarily reflect incomplete lineage sorting during the internodes of the Oxyura phylogeny, but may simply result from a lack of informative substitutions. If the mutation rate is sufficiently low, such that no substitutions accumulated during one or more internodes, an unresolved gene tree and shared ancestral sequences may be observed even if the lineage sorting process was completed during each internode in a phylogeny. The lack of variation, however, makes it impossible to recover the genealogical history. For three loci, however, we observed patterns consistent with retained ancestral polymorphism or the persistent effects of differential lineage sorting in the past. For example, O. jamaicensis shares one allele for -ENOL with O. leucocephala and another with O. vittata and O. australis. All species have unique alleles for MPP, but the MPP gene tree is incongruent with the clade of three Old World species that is strongly supported by the mtDNA data and two nuclear loci (hemoglobin -A, PEPCK9; Fig. 6).
In contrast to the nuclear DNA results, uncertain resolution of the mtDNA gene tree is due primarily to homoplasy. Nearly equal numbers of characters support each of three possible resolutions of a trichotomy among three Old World species (Table 4). A very short internode separating O. leucocephala from (O. australis, O. maccoa) combined with relatively long terminal branches apparently allowed the small number of changes occurring on the internode to be largely obscured by homoplasy, with conflicting characters generated either by parallel changes on two of the three terminal branches or by reversals of changes occurring on the longer branch leading to the most recent common ancestor of these three species.
The parametric bootstrap analyses provide an estimate of the probability of recovering the (O. australis, O. maccoa) clade if these two species are indeed sister taxa in the mtDNA haplotype tree and if the internode separating them from O. leucocephala has the length estimated from the empirical data. The analysis is conducted under the assumption that sequences evolve perfectly, but stochastically, under a known model of sequence evolution that is also used to reconstruct the phylogeny, providing an estimate of the power of the analysis. Failure to reconstruct the expected tree even under these conditions stems from the stochasticity of sequence evolution. Given the short internode and relatively long terminal branches, some replicate data sets by chance accumulate few if any changes on the internode and some of these are overwritten by homoplasy. Given the branch lengths estimated from the empirical data, there is only a 72% chance of recovering the correct resolution of the trichotomy in a maximum likelihood analysis of 4-kb sequences (Table 5), whereas parsimony bootstrap support for the correct resolution of the trichotomy ranges almost uniformly from 10 to 100% (Fig. 7). Expected bootstrap values for the incorrect resolutions of the trichotomy are substantially lower, but values greater than 50% occur in up to 15% of replicates for data sets up to 8 kb. The probability of finding bootstrap support greater than 70% for an incorrect resolution of the trichotomy, however, is low even for data sets of only 2 kb. With 16-kb sequences, much better discrimination between alternative hypotheses is achieved in the simulated data (Fig. 7), suggesting that complete mtDNA sequences for these taxa might allow a more definitive resolution of the Old World Oxyura trichotomy. This assumes, however, that the true length of the internode is greater than or equal to that modeled in our analyses.
In addition, it should be noted that the parametric bootstrapping analysis uses a single model of sequence evolution to both simulate and analyze data. Additional complexities affect the evolution of real sequence data, including heterogeneity in the substitution process among genes and individual nucleotide positions as well as shifts in base composition among taxa. Considering parsimony informative sites in the mtDNA only, Oxyura species differed slightly in base composition, although O. maccoa and O. leucocephala are more similar to each other in base composition than either is to O. australis, suggesting that directional evolution of base composition is not responsible for the mtDNA result (i.e., sister relationship of O. australis and O. maccoa).
We conclude that resolution of the mtDNA tree in favor of (O. australis, O. maccoa) and 64% bootstrap support for this clade (based on the 4070 bp used to estimate parameters for the parametric bootstrap analysis) could occur by chance even if the true mtDNA tree included either a (O. australis, O. leucocephala) or (O. maccoa, O. leucocephala) clade. Nonetheless, (O. australis, O. maccoa) is the clade most strongly supported by the empirical data and must be viewed as the best current hypothesis for mtDNA haplotype relationships. Additional mtDNA sequence data are needed to determine if this trichotomy in the Oxyura mtDNA tree is a hard or soft polytomy.
The above discussion considers only the resolution of the mtDNA gene tree. In the case of Oxyura, there is uncertainty in both the mtDNA tree and gene trees for most nuclear loci, but for different reasons. Because each gene tree is but one manifestation of the lineage sorting process through a single history of population divergence, we must also consider whether individual gene trees are congruent with the species history. If branching events occur in rapid succession relative to a large Ne, then trees derived from independent loci are likely to be incongruent, due to the lasting effects of ancestral polymorphism and differential lineage sorting. Thus, even if the mtDNA tree is resolved in favor of (O. australis, O. maccoa), it may be difficult to say with any confidence that the O. australis and O. maccoa species lineages share a more recent common ancestor than either species lineage does with O. leucocephala.
In other words, congruence of the mtDNA gene tree and the species history depends on both internode length and Ne during this interval. In contrast, recovery of the correct mtDNA gene tree depends only on internode length. However, if the internode separating the three Old World Oxyura was not quite zero, the probability that the mtDNA gene tree matches the species history should be greater than for nuclear loci becaue mtDNA Ne is one quarter the size of nuclear Ne and therefore completes the lineage sorting process more rapidly (e.g., Moore 1995).
Old Ancestral Polymorphism or Homoplasy?
At the time the Old World stiff-tailed ducks diverged, ancestral lineages probably showed substantial mtDNA polymorphism as extant lineages do today (Fig. 2, Fig. 3), and for some time afterwards, ancestral lineages likely persisted in a state of incomplete lineage sorting (Avise et al., 1990; Avise, 1994; Moore, 1995). However, our data indicate that extant Oxyura lineages no longer show incomplete lineage sorting in the mtDNA (Fig. 2). Moreover, conflicting characters occur within individual mtDNA genes, in addition to between different linkage groups (mtDNA versus nuclear DNA). Ancestral polymorphism per se, if it occurred, no longer exists in the Oxyura mtDNA, but it may have lasting effects through the stochastic outcome of the lineage sorting process in independent genetic loci. If ancestral polymorphism generated mismatches between different gene trees (mtDNA versus nuclear gene trees), which gene tree correctly reflects the species tree?
If the true species tree is (O. australis, O. maccoa), then the mtDNA data yielded the correct tree, and two introns are incongruent (presumably due to the differential sorting of ancestral polymorphisms). If (O. maccoa, O. leucocephala) is the true species tree, then we are faced with an incongruent mtDNA tree (either because of old ancestral polymorphism and stochastic lineage sorting or incorrect resolution of the mtDNA tree due to homoplasy) plus two nuclear loci congruent with the species history. Alternatively, the Old World Oxyura trichotomy might very well be a hard polytomy in the species history. If so, any particular gene tree for a given linkage group should have an equal chance of supporting any of three possible resolutions. There also should be no a priori reason to expect difficulty in resolving each gene tree, unless the rate of evolution is so low that few changes have accumulated, as is generally the case for the nuclear introns and exons we sequenced. Our difficulty in resolving the mtDNA gene tree is clearly related to homoplasy. However, parametric bootstrapping suggests that complete mtDNA sequences would yield a fully resolved tree. It is the addition of nuclear data, which strongly support (O. maccoa, O. leucocephala; 92% bootstrap support), that turns the combined mtDNA and nuclear data tree into a trichotomy.
By concatenating mtDNA and nuclear data, we combined independent markers that potentially sorted in different patterns through the species history. Thus, one point of view is that each linkage group effectively provides a single measure of species history and what is important is how many loci support each resolution. If Oxyura is indeed a hard polytomy, 33% of gene trees based on independent, polymorphic loci should support each of the three possible clades. Thus, one test would be to sequence many additional polymorphic introns or exons. However, our sequencing effort yielded little more than one informative single nucleotide polymorphism (SNP) for every other locus. Several tens of kb of nuclear sequence from 50 to 100 loci might be required to perform such a test.
Stochastic homoplasy generally is expected to increase cumulatively with elapsed time and mutation rate (Takahata, 1995; O'hUigin et al., 2002). Bootstrap support, on the other hand, results from the interaction between depth of divergence and internode length. For Oxyura, this pattern is evident in our data. Within the Old World Oxyura clade, terminal branches are long, but the internodes separating the lineages are short (Fig. 2, Fig. 4, Fig. 5). Parametric bootstrap analysis, likewise, suggests that the levels of homoplasy we observed in the mtDNA data partition do not differ from those that would be expected given the short internodes and long terminal branches we observed in Oxyura. For example, the average number characters supporting the three possible resolutions of the trichotmy in 500 simulated data sets (14, 10, 10) are strikingly similar to the numbers of conflicting characters observed in the empirical data (15, 9, 11; numbers differ slightly from Table 4 due to exclusion from this analysis of positions with gaps in one or more taxa).
Similar patterns of conflicting characters also might have resulted if recombination occurred between sites as polymorphisms sorted differentially in related species. However, recombination in mtDNA is rare, and our mtDNA data show no apparent recombination axes. Nucleotide positions supporting the three possible resolutions of the trichotomy are distributed relatively uniformly throughout the control region, 12S rRNA, ND2, ND5, and cytochrome b genes, and the same pattern also occurs within genes (Table 4). No conflicting characters, likewise, were found within individual nuclear loci (Table 4), suggesting that the small size of the products we amplified (mean = 427 bp) and the recent history of Oxyura limited the potential influence of recombination at these loci.
Our results lead to the conclusion that sequences of tens of thousands of base pairs probably would be needed to adequately resolve the Oxyura phylogeny. Complete mtDNA genomes probably would produce a resolved mtDNA gene tree, but the mtDNA gene tree might not reflect the species tree. The root cause of our inability to fully resolve the mtDNA tree with 4 kb clearly is homoplasy. The effects of homoplasy we identified in our study apply generally and ultimately explain why many trees derived from mtDNA data show poor resolution and low bootstrap support. Although the problems of reconstructing short internodes for taxa with relatively long terminal branches are well known, these issues are more often considered in studies of highly divergent taxa. Our findings suggest that even among closely related species, substantial rate variation among sites can generate enough homoplasy to obscure relatively recent speciation events. Future efforts to resolve short internodes separating taxa with long terminal braches, such as those observed in Oxyura, might be better served by sequencing and scoring dozens (if not hundreds) of nuclear loci from a variety of different polymorphic linkage groups. Generally slower rates of mutation and fixation, however, will make the identification of sufficiently variable nuclear loci challenging.
|
|
|
Acknowledgments |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
We thank Yanina Arzamendia, Daniel Blanco, Bob Brua, Guillermo Cao, Raúl Cardón, Tamar Cassidy, Sonia Chavarra, Claudio Chehébar, Les Christidis, Raúl Clarke, Mike Christie, Donna Dittmann, Martin Funes, Sergio Goldfeder, Andy Green, Baz Hughes, Anthony MaGuire, Rodolfo Miatello, Manuel Nores, David Paton, Carmen Quiroga, Van Remsen, Kevin Shaw, Esther Signer, Alejandro del Valle, and Heidi Weingartz for helping to locate or provide stiff-tailed duck tissues. Melissa Cunningham, Sergio Goldfeder, Mike Grogan, Jess Hemmings, Gordon Jarrell, Bill Johnson, Kevin Johnson, Pamela McCracken, Rob Wilson, and Jason Weckstein assisted in the field, and Rich Olsen loaned a shotgun. Collections were made possible by Cape Nature Conservation, Centro de Ecología Aplicada y Dirección Provincial Recursos Faunisticos y Areas Naturales Protegidas Neuquén, Colección Boliviana de Fauna, Delegación Regional Patagonia Administración de Parques Nacionales, Dirección de Fauna y Flora Silvestre Argentina, Dirección de Fauna Santa Cruz, Dirección de Ordenamiento Ambiental–Área Técnica de Fauna Cordoba, Dirección Provincial de Recursos Naturales y Medioambiente Jujuy, Dirección de Recursos Naturales y Gestion Ambiental Corrientes, Ministerio de la Producción Chubut, Secretaría de Estado de Producción Río Negro, Secretaría de Medioambiente y Desarrollo Sustentable Salta, Estacion Biologica de Doñana, Field Museum of Natural History, Louisiana Cooperative Fish and Wildlife Research Unit, Louisiana State University Agricultural Center, College of Agriculture, Museum of Natural Science, and School of Forestry, Wildlife, and Fisheries, South Australia National Parks and Wildlife, Texas Parks and Wildlife, Transvaal Museum, Treehaven Wildfowl Trust, and Wildfowl and Wetlands Trust. Allan Baker, George Barrowclough, Michael Braun, and Chris Simon provided helpful comments on the manuscript. Laboratory costs were funded by the Institute of Arctic Biology at the University of Alaska Fairbanks, Alaska EPSCoR (NSF EPS-0092040), a grant from the Frank M. Chapman Fund at the American Museum of Natural History to K.G.M., and an NSF grant to M.D.S.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionAcknowledgmentsReferences |
|---|
-
Adams J., Slavid E. R. Cheek plumage in the Colombian ruddy duck Oxyura jamaicensis. Ibis (1984) 126:405–407.[CrossRef]
Akaike H. Information theory and an extension of the maximum likelihood principle. In: Proceedings of Second International Symposium on Information Theory—Petrov B. N., Csaki F., eds. (1973) Budapest: Akademia Kiado. 267–281.
Avise J. C. Molecular markers, natural history, and evolution (1994) New York: Chapman and Hall.
Avise J. C. Phylogeography: The history and formation of species (2000) Cambridge, Massachusetts: Harvard University Press.
Avise J. C., Ankney C. D., Nelson W. S. Mitochondrial gene trees and the evolutionary relationship between mallard and black ducks. Evolution (1990) 44:109–1119.
Bremer K. The limits of amino-acid sequence data in angiosperm phylogenetic reconstruction. Evolution (1988) 42:795–803.[CrossRef][Web of Science]
Clement M, Posada D., Crandall K. A. TCS: A computer program to estimate gene genealogies. Mol. Ecol. (2000) 9:1657–1660.[CrossRef][Medline]
Crochet P. A., Desmerais E. Slow rate of evolution in the mitochondrial control region of gulls (Aves: Laridae). Mol. Biol. Evol. (2000) 17:1797–1806.
Delacour J., Mayr E. The famil. Anatidae. Wilso. Bull. (1945) 56:3–55.
DeSalle R., Absher R., Amato G. Speciation and phylogenetic resolution. TREE (1994) 9:297–298.
Desjardins P., Morais R. Sequence and gene organization of the chicken mitochondrial genome: A novel gene order in higher vertebrates. J. Mol. Biol. (1990) 212:599–634.[CrossRef][Web of Science][Medline]
Felsenstein J. Confidence limits on phylogenies: An approach using the bootstrap. Evolution (1985) 39:783–791.[CrossRef][Web of Science]
Fieldså J. Color variation in the ruddy duck (Oxyura jamaicensis andina). Wilso. Bull. (1986) 98:592–594.
Friesen V. L., Congdon B. C., Walsh H. E., Birt T. P. Intron variation in marbled murrelets detected using analyses of single-stranded conformationa. polymorphisms. Mol. Ecol. (1997) 6:1047–1058.[CrossRef][Medline]
Gladstein D., Wheeler W. POY: Phylogeny reconstruction via direct optimization of DNA data (2000) New York, New York: American Museum of Natural History. version 2.0.
Green A. J., Hughes B. Action plan for the white-headed duck Oxyura leucocephala. In: Globally threatened birds in Europe—Heredia B., Rose L., Painter M, eds. (1996) Strasbourg, France: Council of Europe Publishing. 119–146.
Harshman J. Phylogeny, evolutionary rates, and ducks. Ph.D. dissertation (1996) Chicago, Illinois: University of Chicago.
Hasegawa M. H., Kishino Yano T. Dating of the human-ape splitting by a molecular clock of mitochondrial DNA. J. Mol. Evol. (1985) 22:160–174.[CrossRef][Web of Science][Medline]
Hoelzer G. A., Melnick D. J. Patterns of speciation and limits to phylogenetic resolution. TREE (1994) 9:104–107.
Huelsenbeck J. P., Ronquist F. MrBayes: Bayesian inference of phylogenetic trees. Bioinformatics (2001) 17:754–755.
Hughes B., Criado J., Delany S., Gallo-Orsi U., Green A. J., Grussu M., Perennou C., Torres J. A. The status of the North American ruddy duck Oxyura jamaicensis in the Western Palearctic: Towards an action plan for eradication (1999) Strasbourg, France: Council of Europe Publishing. Council of Europe Publication T-PVS/Birds (99) 9.
Johnsgard P. A. Handbook of waterfowl behavior (1965) Ithaca, New York: Cornell University Press.
Johnsgard P. A. Ducks, geese, and swans of the world (1978) Lincoln, Nebraska: University of Nebraska Press.
Johnsgard P. A., Carbonell M. Ruddy ducks and other stifftails: Their behavior and biology (1996) Norman, Oklahoma: University of Oklahoma Press.
Livezey B. C. A phylogenetic analysis of recent anseriform genera using morphological characters. Auk (1986) 103:737–754.[Web of Science]
Livezey B. C. Phylogeny and comparative ecology of stifftailed ducks (Anatidae: Oxyurini). Wilso. Bull. (1995) 107:214–234.
Livezey B. C. A phylogenetic classification of waterfowl (Aves: Anseriformes), including selected fossil species. Ann. Carnegie Mus. (1997) 66:457–496.
Maddison W. Reconstructing character evolution on polytomous cladograms. Cladistics (1989) 5:365–377.[CrossRef][Web of Science]
Marshall H. D., Baker A. J. Structural conservation and variation in the mitochondrial control region of fringilline finches (Fringilla spp.) and in the greenfinch (Carduelis chloris). Mol. Biol. Evol. (1997) 14:173–184.[Abstract]
McCracken K. G., Harshman J., McClellan D. A., Afton A. D. Date set incongruence and correlated character evolution: An example of functional convergence in the hind-limbs of stifftail divin. ducks. Syst. Biol. (1999) 48:683–714.
Moore W. S. Inferring phylogenies from mtDNA variation: Mitochondrial gene trees versus nuclear-gene trees. Evolution (1995) 49:718–726.[CrossRef][Web of Science]
Neigel J. E., Avise J. C. Phylogenetic relationships of mitochondrial DNA under various demographic models of speciation. In: Evolutionary processes and theory—Nevo E., Karlin S., eds. (1986) New York: Academic Press. 515–534.
O'hUigin C., Satta Y., Takahata N., Klein J. Contribution of homoplasy and of ancestral polymorphism to the evolution of genes in Anthropoid primates. Mol. Biol. Evol. (2002) 19:1501–1513.
Pacheco N. M., Congdon B. C., Friesen V. L. The utility of nuclear introns for investigating hybridization and genetic introgression: A case study involving Brachyramphus murrelets. Conserv. Genet. (2002) 3:175–182.[CrossRef]
Posada D., Crandall K. A. Modeltest: Testing the model of DNA substitution. Bioinformatics (1998) 14:817–818.
Primmer C. R., Borge T., Lindell J., Sætre G. P. Single-nucleotide polymorphism characterization in species with limited available sequence information: High nucleotide diversity revealed in the avian genome. Mol. Ecol. (2002) 11:603–612.[CrossRef][Medline]
Quinn T. W., Wilson A. C. Sequence evolution in and around the mitochondrial control region in birds. J. Mol. Evol. (1993) 37:417–425.[CrossRef][Web of Science][Medline]
Raikow R. J. Evolution of diving adaptations in the stiff-tailed ducks. Univ. Calif. Publ. Zool. (1970) 94:1–52.
Rambaut A., Grassly N. C. Seq-Gen: An application for the Monte Carlo simulation of DNA sequence evolution along phylogenetic trees. Comput. Appl. Biosci. (1997) 13:235–238.
Saitou N., Nei M. The number of nucleotides required to determine the branching order of three species, with special reference to the human-chimpanzee-gorilla divergence. J. Mol. Evol. (1986) 24:189–204.[CrossRef][Web of Science][Medline]
Shimodaira H. An approximately unbiased test of phylogenetic tre. selection. Syst. Biol. (2002) 51:492–508.
Sorenson M. D. TreeRot (1999) Boston, Massachusetts: Boston University. version 2.
Sorenson M. D., Ast J. C., Dimcheff D. E., Yuri T., Mindell D. P. Primers for a PCR-based approach to mitochondrial genome sequencing in birds and other vertebrates. Mol. Phylogenet. Evol. (1999) 12:105–114.[CrossRef][Web of Science][Medline]
Sorenson M. D., Fleischer R. C. Multiple independent transpositions of mitochondrial DNA control region sequences to the nucleus. Proc. Nat. Acad. Sci. USA (1996) 93:15239–15243.
Sorenson M. D., Payne R. B. A single, ancient origin of obligate brood parasitism in African finches: Implications for host-parasite coevolution. Evolution (2001) 55:2550–2567.[CrossRef][Web of Science][Medline]
Sorenson M. D., Quinn T. T. W. W. Numts: A challenge for avian systematics and population biology. Auk (1998) 115:214–221.[Web of Science]
Sraml M., Christidis L., Easteal S., Horn P., Collet C. Molecular relationships within Australian waterfowl (Anseriformes). Aust. J. Zool. (1996) 44:47–58.[CrossRef][Web of Science]
Swofford D. L. PAUP*: Phylogenetic analysis using parsimony (*and other methods) (1998) Sunderland, Massachusetts: Sinauer Associates. version 4.
Tajima F. Evolutionary relationships of DNA sequences in finite populations. Genetics (1983) 105:437–460.
Takahata N. Mhc diversity and selection. Immunol. Rev. (1995) 143:225–247.[CrossRef][Web of Science][Medline]
Thompson J. D., Higgins D. G., Gibson T. J. Clustal W: Improving the sensitivity of progressive multiple sequence alignment through sequence weighting, positions-specific gap penalties, and weight matrix choice. Nucleic Acid. Res. (1994) 22:4673–4680.
Todd F. S. Waterfowl: Ducks, geese, and swans of the world (1979) Sand Diego, California: Sea World Press.
Walsh H. E., Kidd M. G., Moum T., Friesen V. L. Polytomies and the power of phylogenetic inference. Evolution (1999) 53:932–937.[CrossRef][Web of Science]
Wheeler W. C. Optimization alignment: The end of multiple sequence alignment in phylogenetics? Cladistics (1996) 12:1–9.[CrossRef][Web of Science]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  K. G. McCracken, M. Bulgarella, K. P. Johnson, M. K. Kuhner, J. Trucco, T. H. Valqui, R. E. Wilson, and J. L. Peters Gene Flow in the Face of Countervailing Selection: Adaptation to High-Altitude Hypoxia in the {beta}A Hemoglobin Subunit of Yellow-Billed Pintails in the Andes Mol. Biol. Evol., April 1, 2009; 26(4): 815 - 827. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Rubinoff, S. Cameron, and K. Will A Genomic Perspective on the Shortcomings of Mitochondrial DNA for "Barcoding" Identification J. Hered., November 1, 2006; 97(6): 581 - 594. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W. Hordijk and O. Gascuel Improving the efficiency of SPR moves in phylogenetic tree search methods based on maximum likelihood Bioinformatics, December 15, 2005; 21(24): 4338 - 4347. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||