A molecular phylogeny of the Drosophila willistoni group: conflicts between species concepts?
The group comprises 25 Neotropical species of the subgenus Sophophora (Throckmorton 1975). Within the group, the six willistoni sibling species include D. willistoni, D. equinoxialis, D. tropicalis, D. insularis, D. pavlovskiana, and D. paulistorum. These species are morphologically indistinguishable, yet they exhibit varying degrees of premating isolation and fail to cross-hybridize (reviewed in Ehrman and Powell 1982). This group also exhibits great variability in geographical ranges including both widespread species and narrowly distributed, insular and mainland endemics. In addition to the species designations, subspecies have been described for D. tropicalis, D. equinoxialis, and D. willistoni. Drosophila paulistorum is composed of six semispecies as defined by ability to produce fertile hybrids. There is also an anomalous category termed the "Carmody strains" (Carmody 1965). These strains originated from two localities, Girardot, COIombia, and Belem, Brazil; the latter is included in this study. These two strains are completely interfertile with one another and do not display any premating isolation. The Belem strain was interfertile with six strains of D. equinoxialis (of 16 tested) and one semispecies (of six) of D. paulistorum. Both Carmody strains display high premating isolation with both D. paulistorum and D. equinoxialis. Thus the willistoni group is a complex of various taxonomic levels.
To date, only two studies have been published on the phylogenetic relationships of the willistoni siblings. Spassky et al. (1971) developed a schematic diagram of the evolutionary relationships based on available biogeographical, genetic, cytological, and biochemical evidence [ILLUSTRATION FOR FIGURE 1A OMITTED]. In the second study, Ayala et al. (1974a) based a phylogeny on the genetic differentiation of the species at 36 allozyme loci. Nei's D (Nei 1972) and Wagner's distance method (Farris 1972) were used to construct a dendrogram of the relationships [ILLUSTRATION FOR FIGURE 1B OMITTED].
There are several reasons why knowledge of the evolutionary relationships of the D. willistoni group is important. A phylogeny provides opportunities to better understand the processes of speciation in the group, for example, it will allow more detailed examination of the evolution of premating and postmating isolation of the species. Courtship song differences have recently been investigated and have been found to be widely divergent (Ritchie and Gleason 1995), but the patterns of evolution of song cannot be assessed without a phylogeny.
In addition, a phylogeny of the group provides an opportunity to compare and contrast two species concepts. All of the species of the D. willistoni group were originally defined by the biological species concept; that is, species are groups of reproductively compatible populations that are reproductively incompatible with other species (Dobzhansky 1937; Mayr 1963). In contrast, a phylogenetic species is a diagnosably distinct cluster with a common ancestry and descent (Cracraft 1989). By definition, a phylogenetic species is monophyletic. The biological species concept does not imply any history of a species and thus a biological species is not necessarily monophyletic. A phylogenetic species may be reproductively isolated but this is not necessary because reproductive isolation is only a subset of effects produced during differentiation. Thus these two species concepts are not always in agreement in defining species.
Here we present a phylogeny of the willistoni sibling species based on the DNA sequences of three genes: the nuclear genes period (per) and Alcohol dehydrogenase (Adh) and the mitochondrial gene Cytochrome oxidase subunit I (COI). Sequencing only one gene to retrieve a phylogeny can be problematic because gene trees do not always reflect species trees. In a species tree, the time of divergence between two species is the time when the two species became reproductively isolated from each other (following the biological species concept). When there is allelic polymorphism within species, a tree constructed from DNA sequences for a given gene can be quite different from the species tree, especially when the time of divergence between species is short (Nei 1987). To help determine if gene trees coincide with species trees, it is helpful to have DNA sequence data for different loci that have evolved independently (Pamilo and Nei 1988).
For multiple datasets, there are competing opinions as to whether the datasets should be analyzed separately and studied for congruence or if they should be analyzed together in a "total evidence" approach (for reviews see Kluge and Wolf 1993; de Queiroz et al. 1995; Miyomoto and Fitch 1995; Hueselbeck et al. 1996; Nixon and Carpenter 1996). A third alternative, conditional combination, depends upon the degree of incongruence in the datasets (Bull et al. 1993; de Queiroz 1993; Larson 1994; Huelsenbeck et al. 1996). Using this approach we examined the degree of incongruence among our datasets before combining and found that even with incongruence among datasets, support for a phylogenetic hypothesis based on one of the genes increased by using all available data. Furthermore, these data produce trees that do not always reflect the taxonomic designations determined primarily by criteria related to the biological species concept. Thus there are conflicts between the biological species concept and the phylogenetic species concept.
MATERIALS AND METHODS
Strains and Sequencing
Strains for all of the D. willistoni sibling species were used (Table 1). As outgroups, the nonsibling species D. nebulosa, D. capricorni, and D. sucinea were also included. Drosophila nebulosa (Sturtevant 1916) is the nearest nonsibling outgroup and was also used by Ayala et al. (1974a). In some analyses, the Carmody strain (Carmody 1965) was also included. This strain is of unknown taxonomic status because individuals of the strain can interbreed with some strains of D. equinoxialis and some strains of D. paulistorum, but not all. Because the original purposes of these studies were different, not all strains or species were sequenced for each of the three genes (Table 1). Approximately 1.2 kb of period (per) coding sequence, 1.3 kb of Alcohol dehydrogenase (Adh) including both coding and noncoding sequence, and 495 bp of mitochondrial Cytochrome oxidase I(COI) coding sequence were analyzed. Genetically, these three loci are independent. COI is mitochondrial. In D. melanogaster, per is located on the X-chromosome, which is homologous to the X-chromosome in D. willistoni (Lakovaara and Saura 1972). The D. willistoni Adh gene is located on the 2R autosome (Rohde et al. 1995).
Sequencing and alignment of most of exon 5 for the period locus is described in Gleason and Powell (1997). The region sequenced corresponds to positions 4843 to 6156 inclusive in the D. melanogaster sequence (GenBank accession number M11969).
Sequencing of the 1.3-kb region of Adh, including 5[prime] and 3[prime] noncoding sequences and an intron, is described in Griffith and Powell (1997). The sequences were aligned using the Jotun Hein method (Hein 1989) in the DNA Star program Megalign with the gap penalty set to 11 and the gap length penalty set to 3. The alignment was adjusted by eye to avoid disruption of codons.
Cytochrome Oxidase I
The COI dataset is composed of 495 bases of the mitochondrial gene Cytochrome oxidase subunit I. Genomic DNA was extracted from 50-200 flies using standard phenol-chloroform extraction followed by proteinase K digestion (Werman et al. 1990). The primers used for PCR were COIe (5[prime]-3[prime]: CCAGTAAATAATGGGTATCAGTG) and COIf (5[prime]-3[prime]: CCAGCTGGAGGAGGAGATCC). The first base of each [TABULAR DATA FOR TABLE 1 OMITTED] corresponds to positions 2672 and 2131, respectively, in the D. yakuba mtDNA sequence (Clary and Wolstenholme 1985).
PCR reactions used approximately 1 [[micro]gram] of total genomic DNA, 200 ng of each primer, 1 unit of Amplitaq DNA polymerase (Perkin Elmer), and final concentrations of 1 x PCR buffer (as formulated by Perkin Elmer) and 0.4 mM dNTP (equimolar mixture of dATP, dCTP, dGTP, dTTP), in a total volume of 100 [[micro]liter]. Amplification was carried out in a Hybaid Omnigene Thermocycler using 30 cycles of 30 s denaturing at 94 [degrees] C, 30 s annealing at 50 [degrees] C, and 30 s extension at 72 [degrees] C.
COI sequencing proceeded by two methods. For the D. paulistorum strains, the PCR products were cloned and sequenced by the procedure described in Gleason and Powell (1997). The rest of the strains were sequenced by direct sequencing. One microliter of the first PCR product was used in a single-stranded PCR reaction. PCR conditions were the same as above except that the primers were unbalanced so that one primer was at the normal concentration and only 20 ng of the other was used. In addition, the annealing temperature was 55 [degrees] C. The resulting PCR product was concentrated on a Centricon 30 COIumn (Amicon). Sequencing followed a modification of the Sequenase 2.0 kit protoCOI (USB) and was performed on both strands. Sequences were aligned by eye to the D. yakuba sequence (Clary and Wolstenholme 1985), as there were no indels.
Maximum-parsimony analyses were done using PAUP (vers. 3.0, Swofford 1990). In all analyses, D. nebulosa was designated as the outgroup. For the per analysis with all strains and species, D. capricorni and D. sucinea were also used as outgroups. Gaps were treated as missing data. For the smallest datasets, exhaustive searches were done. For datasets of more than 11 and less than or equal to 15 taxa, branch-and-bound searches were done. The options "compute via stepwise," MULPARS and "simple" addition sequence were used. For the larger datasets, heuristic search settings were: random addition sequence, 100 replicates, treebisection-reconnection (TBR) branch swapping, the MULPARS option. The MAXTREES setting ranged from 200 to 1000.
One hundred bootstrap replicates were done for each branch-and-bound search with the same settings. For bootstraps of a heuristic search, 100 replicates were done with the heuristic search set to the same values except that only one heuristic replicate was done per bootstrap replicate.
The PHYLIP (vers. 3.57c, Felsenstein 1995) programs were used for phylogenetic analyses employing distance data. Using the DNADIST program, distances were computed by both the maximum-likelihood distance method (Felsenstein 1978) and by the Kimura two-parameter method (Kimura 1980) using a transition/transversion ratio of one. Because the species studied here are so closely related, these distances were equivalent. Only the results of the Kimura two-parameter model are presented here. Phylogenies were constructed using the neighbor-joining algorithm (Saitou and Nei 1987) in the NEIGHBOR program of PHYLIP. For bootstrap analysis, 100 bootstrap replicates were done by SEQBOOT to create 100 distance datasets. From the resulting neighbor-joining trees, a consensus tree was derived using CONSENSE.
The smaller datasets were also examined using maximum likelihood with the DNAML program of PHYLIP (vers. 3.57c, Felsenstein 1995). Ten jumbles and a transition/transversion ratio of one were used for each analysis. Bootstraps analysis on 100 datasets were also performed by the same method, but without jumbling.
Incongruence length difference tests (ILD; Mickevich and Farris 1981; Farris et al. 1995) were performed using PAUP* (vers. 4.0.0.d63, Swofford, in press). Invariant characters were removed, and 10,000 replicates were performed for each test.
Phylogenetic Analyses of Separate Datasets: All Strains
Because the strains sequenced for each gene do not completely overlap, the datasets were first examined with all strains and species available for each gene. Subsequently, we performed analyses with only the strains that overlap among datasets.
The entire per dataset has a total of 1231 characters. Analysis of the ratio of transitions and transversions in per indicates that the transitions are not saturated (Gleason 1996); therefore, transitions and transversion were weighted equally in all analyses. One region of the fragment of per has a variable number of glycines and thus many indels (Gleason and Powell 1997). Because of problems in the alignment of this repeat region, maximum-parsimony analysis was performed with and without this region. Heuristic analysis without the repeat region (nucleotides 616-666) produced 356 trees of length 476 (results not shown). The multiple mostparsimonious trees are the result of rearrangements of strains within D. willistoni, D. equinoxialis, and D. paulistorum and not of rearrangements among species.
Another heuristic search was done with the entire per data sequence fragment. In this case, because of computer time limitations not all most parsimonious trees were found, although there are at least 660 trees of length 504. Bootstrap analysis was completed [ILLUSTRATION FOR FIGURE 2 OMITTED]. The tree is slightly more resolved than the analysis without the repeat region. Resolution is also slightly greater for the strains within species when the repeat is included. The bootstrap values are generally higher for the results of the entire dataset. There is only one bootstrap value less than 100 for a node joining all strains within a species (the D. equinoxialis strains). Note that there is strong support for the Carmody strain belonging to the D. equinoxialis clade [ILLUSTRATION FOR FIGURE 2 OMITTED]. By including the repeat, the bootstrap value for this clade increases from 94% to 97%. The only other bootstrap value that changes among the species is an increase from 71% to 89% for the node linking D. willistoni to D. equinoxialis and D. paulistorum. The repeat was included in all further analyses. An identical topology was obtained by the neighbor-joining method using the Kimura two-parameter model (results not shown).
Because over the entire Adh sequence transitions and transversion are approximately equal (Gleason 1996), they were not weighted differently in maximum-parsimony analysis. There were 1285 characters in this dataset, of which 85 were informative in a parsimony analysis. A heuristic maximum parsimony search produced 26 trees of length 322. The only differences among the trees lie in the arrangement of the D. willistoni strains. A 50%-majority-rule consensus tree shows the same topology for the species [ILLUSTRATION FOR FIGURE 3 OMITTED], as the results of analysis of per. However, bootstrap values for this tree are mixed. Nodes joining all members of a single species are 99% or 100%. As with the per phylogenies, the node with the lowest support (53%) is the one joining the D. paulistorum/D. equinoxialis clade to D. willistoni. Similarly, the bootstrap value for the node joining D. tropicalis to the others is only 82%, as compared to 100% in the per tree. In addition, the node joining D. equinoxialis to D. paulistorum is only 52%, much lower than the 100% found for this node in the per maximum-parsimony tree. A different species topology was obtained using neighbor-joining (see below). In this case, D. tropicalis is the sister taxon to D. equinoxialis.
An unweighted maximum parsimony search was initially done using the branch-and-bound option for the COI dataset, which includes a total of 495 bases. Six trees were found with length of 74. A 50%-majority-rule consensus tree (results not shown) indicates that there are several sources of ambiguity. In all trees, the D. willistoni strains are united, D. tropicalis and D. insularis are sister taxa, the Carmody strain is joined to the D. equinoxialis strains, and all of the D. paulistorum semispecies and D. pavlovskiana group together with the exception of the Amazonian semispecies (paulistorumA), which does not cluster with the other D. paulistorum semispecies. Branch-and-bound bootstrap analysis of this dataset indicates that the phylogeny is not well resolved [ILLUSTRATION FOR FIGURE 4 OMITTED]. The major contributing source of the low bootstrap values is probably the low number of informative sites (n = 28).
A similar topology was obtained by the neighbor-joining method, although some of the relationships among the strains were different (results not shown). Again, the D. insularis/D. tropicalis clade is joined to D. willistoni, although the bootstrap value for this node (65%) is not very high. Carmody again clusters with D. equinoxialis and as before, the Amazonian semispecies does not cluster with the other D. paulistorum strains.
Phylogenetic Analyses of Separate Datasets: Overlapping Strains
Six strains were sequenced for all three genes; these six were used in all subsequent analyses. A different D. nebulosa strain was used in the COI dataset as compared to the per and Adh datasets but this difference has been ignored. For per, the resulting dataset has 1216 characters, of which 79 are informative in a parsimony analysis. An exhaustive search produced one most-parsimonious tree of length 336; the next shortest tree is 339 steps. The topology of this tree was the same as that obtained for both the neighbor-joining and maximum-likelihood methods [ILLUSTRATION FOR FIGURE 5A OMITTED]. For the maximum-likelihood tree, all branches are significant at the P [less than] 0.01 level and the log-likelihood value is -3351.1. Bootstrap analysis using all three methods resulted in values equal to or greater than 79% for all nodes (Table 2).
[TABULAR DATA FOR TABLE 2 OMITTED]
There are 1284 characters in the Adh dataset, of which 51 are informative in a parsimony analysis. An exhaustive search produced one most parsimonious tree of length 271; the next shortest tree is 272 steps. The most parsimonious tree is identical to that for per and the same species topology was obtained by neighbor-joining and maximum-likelihood methods. All branch lengths for maximum likelihood are significant at the P [less than] 0.01 level and the log-likelihood score is - 3143.5. However, a bootstrap analysis for neighbor joining results in a different topology [ILLUSTRATION FOR FIGURE 5B OMITTED]. For all trees, bootstrap values are low when joining species except for node 4 (Table 2, [ILLUSTRATION FOR FIGURE 5B OMITTED]).
The COI dataset is the smallest with only 495 bases, of which 13 are informative in a parsimony analysis. COI is not saturated for transitions among the ingroups (Gleason 1996). An unweighted exhaustive maximum parsimony search for the COI dataset yielded one most-parsimonious tree of 50 steps; there are seven trees at 52 steps. This most-parsimonious tree differs from that for per and Adh [ILLUSTRATION FOR FIGURE 5C OMITTED]. This same tree is obtained for neighbor joining and maximum likelihood. Once again, under maximum likelihood all branch lengths, with the exception of those leading to the D. willistoni strains, are significant at the P [less than] 0.01 level and the loglikelihood value is - 928.9. Bootstrap values for the nodes joining different species range from 62% to 94% (Table 2, [ILLUSTRATION FOR FIGURE 5C OMITTED]).
Incongruence Length Difference Tests
One concern when analyzing multiple datasets in phylogenetic analyses is whether incongruence among datasets precludes combining data. Strong incongruence indicates that different datasets have had different histories, a scenario that violates the assumption of phylogenetic reconstruction. In this case, neither the Adh nor the COI datasets strongly support alternative phylogenetic hypotheses to that resulting from the per analyses. To further test this, ILD tests were performed on alternative partitions of these datasets to examine incongruence in parsimony trees.
There is not incongruence between the per and Adh datasets (Table 3) or between either of these genes and the other combined with COI. In contrast, there is significant incongruence between COI and per and Adh both individually and combined at the P = 0.05 level (Table 3). Because the genes all include coding sequences, first and second codon position were also compared with third codon position; none of the comparisons were significant (Table 3).
By these ILD tests, incongruence indicates that the two nuclear gene datasets (per and Adh) can be easily combined, whereas the COI dataset cannot. However, it has been demonstrated that even with an incongruence P-value [greater than] 0.01, combining datasets can improve, or at least not reduce, phylogenetic accuracy (Cunningham 1997). By this observation, COI can be combined with Adh easily. Furthermore, only with P-values [less than] 0.001 has it been demonstrated that combined data suffer relative to individual partitions (Cunningham 1997). Because none of our P-values are [less than] 0.001, combined approaches were used.
TABLE 3. Results of incongruence length difference test partitions. Partitions P-value All three genes per, Adh, and COI 0.0077 per, Adh coding sequences, COI 0.0048 Single genes vs. two genes per vs. Adh and COI 1 Adh vs. per and COI 1 COI vs. Adh and per 0.005 Combinations of single genes per vs. Adh 1 per vs. COI 0.0011 Adh vs. COI 0.0302 1st and 2nd codon positions vs. 3rd positions per, Adh and COI 1 per 1 Adh 1 COI 0.2121
Analyses were performed on all three genes as well as all combinations of two datasets. All resulting trees from all analyses result in the same tree as that for per alone [ILLUSTRATION FOR FIGURE 5A OMITTED]. In all cases the addition of one of the other datasets to the per dataset increases the bootstrap value of each node, with the exception of the neighbor-joining and maximumlikelihood methods at node 3 when combining per and Adh (Table 2). The addition of the COI dataset to the Adh dataset similarly increases the values for nodes 1 and 3 but not for node 4 over the value for either dataset alone (Table 2). Neither the Adh nor COI dataset by itself supported node 3 when using the neighbor-joining method; however, when combined, the node is supported with a bootstrap value of 71%.
Using the computer program MacClade (Maddison and Maddison 1992) various topologies were compared using the entire dataset. The shortest tree is 661 steps (Table 4) and corresponds to that obtained using per alone [ILLUSTRATION FOR FIGURE 5A OMITTED]. The Adh neighbor-joining tree and the allozyme topologies are the next two shortest though requiring seven and 12 extra steps, respectively (Table 4). The COI tree, which is the only one that does not have D. insularis as the most basal of the sibling species, is the worst fit for the dataset among all of the proposed topologies.
Individual and Combined Datasets
Of the three datasets, only per had strong phylogenetic signal as judged by bootstrap analysis [ILLUSTRATION FOR FIGURE 2 OMITTED]; the other two sequences (Adh and COI) each by themselves produced poorly supported trees that sometimes had topologies conflicting with the per tree [ILLUSTRATION FOR FIGURES 3, 4, 5B, C OMITTED]. It has been suggested that when individual datasets do not produce strongly supported phylogenies independently, a combined approach should be used (de Queiroz 1993). It has also been suggested that the [TABULAR DATA FOR TABLE 4 OMITTED] strength of incongruence should be examined (Bull et al. 1993; de Queiroz 1993; Larson 1994; Huelsenbeck et al. 1996). To assess this, various ILD tests were performed that indicated an absence of incongruence between per and Adh (which produce the same tree under maximum-parsimony criteria). COI differs significantly from both per and Adh. The incongruence for COI and Adh is within the range suggested to not adversely affect phylogenetic accuracy (Cunningham 1997). By combining these datasets, support for the resulting tree increases (as measured by bootstrap support) over that for Adh alone for all methods of analysis and at all nodes, except for the most basal, which decreases (Table 2).
The incongruence between per and COI is greater than that between Adh and COI creating an asymmetrical relationship. Asymmetrical patterns of ILD incongruence make it difficult to determine what datasets to include or exclude in a combined analysis; yet by combining, underlying phylogenetic signal can be amplified (Baker and DeSalle 1997). Although per and COI are the most incongruent pair, adding COI to per increases the bootstrap value of all nodes (Table 2). This increase has been seen in other combinations of datasets (e.g., Doyle et al. 1994; Olmstead and Sweere 1994).
Because conflicts in topology arise for weakly supported nodes from a subset of the data, we have confidence in the final topology, which is strongly supported by a combined analysis [ILLUSTRATION FOR FIGURE 5A OMITTED]. In addition, all three methods of phylogenetic analysis (maximum parsimony, neighbor joining, and maximum likelihood) generate the same topology with nearly equally strong bootstrap support (Table 2), a finding that may also may add confidence that the "true" tree has been recovered (Kim 1993).
The relationships of the species determined from these DNA sequences are different from those of previous studies. It is not clear what criteria were used by Spassky et al. (1971) in producing their diagram of relationships [ILLUSTRATION FOR FIGURE 1A OMITTED], so it is not possible to analyze the conflict. For the allozyme tree [ILLUSTRATION FOR FIGURE 1B OMITTED] part of the reason may lie in the construction method used: Wagner's method (Farris 1972) using Nei's D (Nei 1972). Because the sophistication of phylogenetic analysis has greatly improved since then, a reanalysis of the allozyme data today might produce a different result.
Biological versus Phylogenetic Species Concepts
The biological species concept implies no evolutionary mechanism and thus it is not surprising that it does not agree with the phylogeny of the most problematic stains here, the Carmody strain and D. pavlovskiana. In fact, the biological species concept, which has been used to define all the species of this group, does not provide a framework for problematic strains that can hybridize with two well-separated species. Therefore, the taxonomic rank of the Carmody strain has never been defined because it is interfertile with some strains of D. equinoxialis and one semispecies of D. paulistorum. All phylogenetic analyses, including the original Spassky et al. (1971; [ILLUSTRATION FOR FIGURE 1A OMITTED]) proposal, allozymes (Ayala et al. 1974a; [ILLUSTRATION FOR FIGURE 1B OMITTED]), and all three genes analyzed here, have consistently supported the notion that D. equinoxialis and D. paulistorum are sister taxa. From the perspective of the biological species concept, the Carmody strains would appear to be intermediate between these species. However, from a phylogenetic species perspective (Cracraft 1989), in those DNA sequence studies that include the Carmody strain, the intermediate would be classified as D. equinoxialis [ILLUSTRATION FOR FIGURES 2, 4 OMITTED].
As for D. pavlovskiana, the biological species concept has placed it in a separate species because of its failure to hybridize with D. paulistorum, although it was originally included as another semispecies (Guianan) of D. paulistorum. Kastritsis and Dobzhansky (1967) argued that it is significantly more genetically distinct than all other semispecies and thus deserved species status. This was based primarily on chromosomal studies: unlike in hybrids between semispecies, hybrids between the Guianan strains and all other semispecies have unpaired chromosomes. Also, D. paulistorum displays a different chromosomal banding pattern. Furthermore, Kessler (1962) had found the mating behavior of the Guianan semispecies to be both quantitatively and qualitatively different from all other semispecies. In general, the D. paulistorum complex itself has always been problematic in that semispecies do not neatly fit into reproductively isolated populations. Given this, it is perhaps not surprising that the phylogeny should be so messy and a phylogenetic perspective would place all D. paulistorum and D. pavlovskiana together as a single taxon.
It is certainly conceivable that these apparent conflicts are due to the relatively small amount of sequence data available for these strains and species and that further information would make the conflict disappear. However, the placements of these taxa are supported by high bootstrap values and in the strongest dataset [ILLUSTRATION FOR FIGURE 2 OMITTED] these two anomalous taxa cluster deep within the strains representing a single species. Another possibile explanation is allele sharing due to ancestral polymorphism. Unfortunately, only a single strain of each of these taxa are in existence today, so this possibility cannot be tested. That an autosomal [ILLUSTRATION FOR FIGURE 2 OMITTED] and a mitochondrial gene [ILLUSTRATION FOR FIGURE 4 OMITTED] both place D. pavlovskiana within the D. paulistorum clade is evidence against this explanation.
While there is good support for the broader-scale relationships of the sibling species of the willistoni group, our data are less informative at the subspecific or semispecific level of relationships. Although the per data give strong support for the sister status of the Andean-Brazilian (AB) and Orinocan (O) semispecies and for the close relationship of D. pavlovskiana to the Central American (CA) semispecies [ILLUSTRATION FOR FIGURE 2 OMITTED], we do not place much confidence in these results because only a single strain of each semispecies was sequenced for this single gene. The semispecies designation of the D. paulistorum strain TM is not known, so the Adh data do not provide information about semispecies relationships. At the subspecific level, the Lima (L) strain of D. willistoni, described as subspecies D. w. quechua (Ayala 1973), clusters inside the other strains of this species [ILLUSTRATION FOR FIGURE 2 OMITTED]. In D. equinoxialis, strains from Caribbean Islands are considered a separate subspecies (D. e. caribbensis) relative to the continental strains of this species (Ayala et al. 1974b). Thus, in Figure 2 strains of D. equinoxialis designated C (Cuba) and P (Puerto Rico) should be distinct from the other continental strains, which they are not.
Clearly, the willistoni group presents a challenge to systematists and evolutionists alike. In this paper we present considerable progress in reconstructing phylogenetic relationships on a broad scale. While this information is crucial, to examine microevolutionary processes below the species level (e.g., subspecies and semispecies), clearly finer-scale phylogenetic resolution is vital. Although not totally definitive, our results question some of the previous conclusions concerning the systematics of this group. A reexamination of this group, in light of modern concepts and employing modern methodologies, is likely to produce a rich source of information on the evolutionary process.
Fly strains were kindly provided by P. Chabora, V. Valente, L. Ehrman, L. Strausbaugh, and F. Ayala. We would like to thank M. Ritchie, R. DeSalle, and two anonymous reviewers for comments on the manuscript and E. Moriyama for help with some analyses. D. Swofford kindly allowed us to use PAUP* in advance of formal release. Financial support came from a Sigma Xi Grant-in-Aid of Research, National Science Foundation Dissertation Improvement Grants to JMG (DEB 924749) and to ECG (DEB 9122899) and National Science Foundation Grant DEB 9318836 to JRP. JMG and ECG were both supported by PHS Training Grants.
AYALA, F. J. 1973. Two new subspecies of the Drosophila willistoni group. Pan-Pacific Entomol. 49:273-279.
AYALA, F. J., C. A. MOURAO, S. PEREZ-SALAS, R. RICHMOND, AND T. DOBZHANSKY. 1970. Enzyme variability in the Drosophila willistoni group. I. Genetic differentiation among sibling species. Proc. Nat. Acad. Sci. USA 67:225-232.
AYALA, F. J., M. L. TRACEY, D. HEDGECOCK, AND R. C. RICHMOND. 1974a. Genetic differentiation during the speciation process in Drosophila. Evolution 28:576-592.
AYALA, F. J., M. L. TRACEY, L. G. BARR, AND J. G. EHRENFELD. 1974b. Genetic and reproductive differentiation of Drosophila equinoxialis caribbensis. Evolution 28:24-41.
BAKER, R. H., AND R. DESALLE. 1997. Multiple sources of character information and the phylogeny of Hawaiian Drosophilids. Syst. Biol. 46:654-673.
BULL, J. J., J. P. HUELSENBECK, C. W. CUNNINGHAM, D. L. SWOFFORD, AND P. J. WADDELL. 1993. Partitioning and combining data in phylogenetic analysis. Syst. Biol. 42:384-397.
CARMODY, G. 1965. Two unusual strains of the D. willistoni sibling species group. Drosophila Inf. Serv. 40:53.
CLARK, J. B., W. P. MADDISON, AND M. G. KIDWELL. 1994. Phylogenetic analysis supports horizontal transfer of P transposable elements. Mol. Biol. Evol. 11:40-50.
CLARY, D. O., AND D. R. WOLSTENHOLME. 1985. The mitochondrial DNA molecule of Drosophila yakuba: nucleotide sequence, gene organization and genetic code. J. Mol. Evol. 22:252-271.
CRACRAFT, J. 1989. Speciation and its ontology: the empirical consequences of alternative species concepts for understanding patterns and processes of differentiation. Pp. 28-59 in D. Otte and J. A. Endler, eds. Speciation and its consequences. Sinauer, Sunderland, MA.
CUNNINGHAM, C. W. 1997. Can three incongruence tests predict when data should be combined? Mol. Biol. Evol. 14:733-740.
DA CUNHA, A. B., TH. DOBZHANSKY, O. PAVLOVSKY, AND B. SPASSKY. 1959. Genetics of natural populations. XXVIlI. Supplemental data on the chromosomal polymorphism in Drosophila willistoni in relation to the environment. Evolution 13:389-404.
DANIELS, S. B., K. R. PETERSON, L. D. STRAUSBAUGH, M. G. KIDWELL, AND A. CHOVNICK. 1990. Evidence for horizontal transmission of the P transposable element between Drosophila species. Genetics 124:339-355.
DE QUEIROZ, A. 1993. For consensus (sometimes). Syst. Biol. 42: 368-372.
DE QUEIROZ, A., M. J. DONOGHUE, AND J. KIM. 1995. Separate versus combined analysis of phylogenetic evidence. Annu. Rev. ECOI. Syst. 26:657-681.
DOBZHANSKY, TH. 1937. Genetics and the origin of species. Columbia Univ. Press, New York.
-----. 1950. The chromosomes of Drosophila willistoni. J. Heredity 41:156-158.
DOYLE, J. A., M. J. DONOGHUE, AND E. A. ZIMMER. 1994. Integration of morphological and rRNA data on the origin of angiosperms. Ann. Mo. Bot. Gard. 81:419-450.
EHRMAN, L. 1961. The genetics of sexual isolation in Drosophila paulistorum. Genetics 46:1025-1038.
-----. 1965. Direct observation of sexual isolation between allopatric and between sympatric strains of the different Drosophila paulistorum races. Evolution 19:459-464.
EHRMAN, L., AND J. R. POWELL. 1982. The Drosophila willistoni species group. Pp. 193-225 in M. Ashburner, H. L. Carson, and J. N. Thompson Jr., eds. The genetics and biology of Drosophila. Academic Press, New York.
FARRIS, J. S. 1972. Estimating phylogenetic trees from distances matrices. Am. Nat. 106:645-668.
FARRIS, J. S., M. KALLERSJO, A. G. KLUGE, AND C. BULT. 1995. Testing significance of incongruence. Cladistics 10:315-319.
FELSENSTEIN, J. 1978. Evolutionary trees from DNA sequences: a maximum likelihood approach. J. Mol. Evol. 17:368-376.
-----. 1995. PHYLIP: phylogeny inference package. Vers. 3.57c. University of Washington, Seattle, WA.
GLEASON, J. M. 1996. Molecular evolution of the period locus and evolution of courtship song in the Drosophila willistoni sibling species. Ph.D. diss., Yale University, New Haven, CT.
GLEASON, J. M., AND J. R. POWELL. 1997. Interspecific and intraspecific variation in the period locus in the Drosophila willistoni group. Mol. Biol. Evol. 14:741-753.
GRIFFITH, E. C., AND J. R. POWELL. 1997. Adh nucleotide variation in Drosophila willistoni: high replacement polymorphism in an electrophoretically monomorphic protein. J. Mol. Evol. 45:232-237.
HEIN, J. 1989. A new method that simultaneously aligns and reconstructs ancestral sequences for any number of homologous sequences, when phylogeny is given. Mol. Biol. Evol. 6:669-684.
HUELSENBECK, J. P., J. J. BULL, AND C. W. CUNNINGHAM. 1996. Combining data in phylogenetic analysis. Trends ECOI. Evol. 11: 152-158.
KASTRITSIS, C. D., AND TH. DOBZHANSKY. 1967. Drosophila pavlovskiana, a race or a species? Am. Midl. Nat. 78:244-247.
KESSLER, S. 1962. Courtship rituals and reproductive isolation between the races or incipient species of Drosophila paulistorum. Am. Nat. 96:117-121.
KIM, J. 1993. Improving the accuracy of phylogenetic estimation by combining different methods. Syst. Biol. 42:331-340.
KIMURA, M. 1980. A simple method of estimating evolutionary rate of base substitutions through comparative studies of nucleotide sequences. J. Mol. Evol. 16:111-120.
KLUGE, A. O., AND A. J. WOLF 1993. Cladistics: what's in a word? Cladistics 9:183-199.
LAKOVAARA, S., AND A. SAURA. 1972. Location of enzyme loci in chromosomes of Drosophila willistoni. Experientia 28:355-356.
LARSON, A. 1994. The comparison of morphological and molecular data in phylogenetic systematics. Pp. 371-390 in B. Schierwater, B. Streit, G. P. Wagner, and R. DeSalle, eds. Molecular ecology and evolution: approaches and applications. Birkhauser Verglag, Basel, Switzerland.
MADDISON, W. P., AND D. R. MADDISON. 1992. MacClade: Analysis of phylogeny and character evolution. Sinauer, Sunderland, MA.
MAYR, E. 1963. Animal species and evolution. Harvard Univ. Press, Cambridge, MA.
MICKEVICH, M. F., AND J. S. FARRIS. 1981. The implication of congruence in Medidia. Syst. Zool. 30:351-370.
MIYAMOTO, M. M., AND W. M. FITCH. 1995. Testing species phylogenies and phylogenetic methods with congruence. Syst. Biol. 44:64-76.
NEI, M. 1972. Genetic distance between populations. Am. Nat. 106:283-292.
-----. 1987. Molecular evolutionary genetics. Colombia Univ. Press, New York.
NIXON, K. C., AND J. M. CARPENTER. 1996. On simultaneous analysis. Cladistics 12:221-241.
OLMSTEAD, R. G., AND J. A. SWEERE. 1994. Combining data in phylogenetic systematics: an empirical approach using three molecular data sets in the Solanaceae. Syst. Biol. 43:467-481.
PAMILO, P., AND M. NEI. 1988. Relationships between gene trees and species trees. Mol. Biol. Evol. 5:568-583.
POWELL, J. R., AND J. M. GLEASON. 1996. Codon usage and the origin of P-elements. Mol. Biol. Evol. 13:278-279.
RITCHIE, M. G., AND J. M. GLEASON. 1995. Rapid evolution of courtship song pattern in Drosophila willistoni sibling species. J. Evol. Biol. 8:463-479.
ROHDE, C., E. ABDELHAY, H. PINTO JR., A. SCHRANK, AND V. L. S. VALENTE. 1995. Analysis and in situ mapping of the Adh locus in species of the willistoni group of Drosophila. Cytobios 81: 37-47.
SAITOU, N., AND M. NEI. 1987. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4:406-425.
SPASSKY, B., R. C. RICHMOND, S. PEREZ-SALAS, O. PAVLOVSKY, C. A. MOURAO, A. S. HUNTER, n. HOENIGSBERG, TH. DOBZHANSKY, AND F. J. AYALA. 1971. Geography of the sibling species related to Drosophila willistoni and of the semispecies of the Drosophila paulistorum complex. Evolution 25:129-143.
STURTEVANT, A. H. 1916. Notes on North American Drosophilidae with descriptions of twenty-three new species. Ann. Ent. Soc. Am. 9:323-343.
SWOFFORD, D. 1990. PAUP*: phylogenetic analysis using parsimony. Vers. 3.0. Illinois Natural History Survey, Champaign.
-----. In press. PAUP*: phylogenetic analysis using parsimony. Vers. 4.0.0.d63. Sinauer, Sunderland, MA.
THROCKMORTON, L. H. 1975. The phylogeny, ecology and geography of Drosophila. Pp. 421-469 in R. C. King, ed. Handbook of genetics. Plenum Press, New York.
WERMAN, S. D., E. H. DAVIDSON, AND R. J. BRITTEN. 1990. Rapid evolution in a fraction of the Drosophila nuclear genome. J. Mol. Evol. 30:281-289.
|Printer friendly Cite/link Email Feedback|
|Author:||Gleason, Jennifer M.; Griffith, Elizabeth C.; Powell, Jeffrey R.|
|Date:||Aug 1, 1998|
|Previous Article:||Sex and speciation: genetic architecture and evolutionary potential of sexual versus nonsexual traits in the sibling species of the Drosophila...|
|Next Article:||Genetic variation in cricket calling song across a hybrid zone between two sibling species.|