Molecular Biology and Evolution 18:1117-1131 (2001)
© 2001 Society for Molecular Biology and Evolution
ARTICLE |
The Phylogenetic Analysis of Variable-Length Sequence Data: Elongation Factor–1
Introns in European Populations of the Parasitoid Wasp Genus Pauesia (Hymenoptera: Braconidae: Aphidiinae)
Institut Cavanilles de Biodiversitat i Biología Evolutiva, Universitat de València, Valencia, Spain
Department of Biology, Imperial College at Silwood Park, Ascot, Berks, England
Departament of Conservation Biology, Uppsala, Sweden
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionAcknowledgementsliterature cited |
|---|
Elongation factor–1
(EF-1) is a highly conserved nuclear coding gene that can be used to investigate recent divergences due to the presence of rapidly evolving introns. However, a universal feature of intron sequences is that even closely related species exhibit insertion and deletion events, which cause variation in the lengths of the sequences. Indels are frequently rich in evolutionary information, but most investigators ignore sites that fall within these variable regions, largely because the analytical tools and theory are not well developed. We examined this problem in the taxonomically problematic parasitoid wasp genus Pauesia (Hymenoptera: Braconidae: Aphidiinae) using congruence as a criterion for assessing a range of methods for aligning such variable-length EF-1 intron sequences. These methods included distance- and parsimony-based multiple-alignment programs (CLUSTAL W and MALIGN), direct optimization (POY), and two "by eye" alignment strategies. Furthermore, with one method (CLUSTAL W) we explored in detail the robustness of results to changes in the gap cost parameters. Phenetic-based alignments ("by eye" and CLUSTAL W) appeared, under our criterion, to perform as well as more readily defensible, but computationally more demanding, methods. In general, all of our alignment and tree-building strategies recovered the same basic topological structure, which means that an underlying phylogenetic signal remained regardless of the strategy chosen. However, several relationships between clades were sensitive both to alignment and to tree-building protocol. Further alignments, considering only sequences belonging to the same group, allowed us to infer a range of phylogenetic relationships that were highly robust to tree-building protocol. By comparing these topologies with those obtained by varying the CLUSTAL parameters, we generated the distribution area of congruence and taxonomic compatibility. Finally, we present the first robust estimate of the European Pauesia phylogeny by using two EF-1 introns and 38 taxa (plus 3 outgroups). This estimate conflicts markedly with the traditional subgeneric classification. We recommend that this classification be abandoned, and we propose a series of monophyletic species groups. |
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsliterature cited |
|---|
A major problem in molecular phylogenetics is the establishment of base pair homology among variable-length sequences, for example, in noncoding regions such as ribosomal genes and introns. Typically, base positions are aligned by the insertion of gaps, representing insertion or deletion events (indels), prior to tree-building. Several computer programs employing different algorithms are available for generating such multiple alignments, and Morrison and Ellis (1997)
demonstrated that the choice of program could have a greater effect on the estimated phylogeny than the choice of tree-building method. As either a complementary or an alternative strategy, many authors simply exclude regions they consider difficult to align, usually on the grounds that they are unlikely to be phylogenetically informative and will merely introduce noise (Olsen and Woese 1993
). This inevitably introduces an unwelcome element of subjectivity and may reflect an undue concern over the danger of noise. Wenzel and Siddall (1999)
demonstrated that levels of noise, such as that caused by completely saturated third codon positions, are unlikely to have a significantly detrimental effect on the phylogenetic signal from elsewhere in the gene. Although secondary structure can be used to improve multiple alignments by identifying homologous regions (Kjer 1995
), its estimation is problematic, and many types of sequences do not produce such structure. A further problem with multiple alignments is how to treat gaps, which are informative evolutionary events, but most distance and maximum likelihood analyses would treat them as "missing data." Despite this, most discussions of molecular phylogenetic methods (e.g., Swofford et al. 1996
) do not address these questions, and few studies explore the sensitivity of their results to alignment methods as thoroughly as their sensitivity to tree-building algorithms (but see Giribet and Wheeler 1999
; Maddison, Baker, and Ober 1999
). Recently, a direct optimization approach (Wheeler 1996
) to such data has become available. This avoids the conceptually problematic separation of alignment and tree-building and uses parsimony to infer phylogenetic relationships directly from variable-length sequences.
In this study, we compare methods of analyzing variable-length sequence data, in the present case, introns in the elongation factor-1 gene from closely related species of the parasitoid wasp genus Pauesia (Hymenoptera: Braconidae: Aphidiinae). Elongation factor–1 (EF-1) is a nuclear coding gene involved in the GTP-dependent binding of charged tRNAs to the acceptor site of the ribosome during translation. This gene has been characterized in several animal species (Brands et al. 1986
; Lenstra et al. 1986
; Rao and Slobin 1986
; Roth et al. 1987
; Hovemann et al. 1988
; Walldorf and Hovemann 1990
; Danforth and Ji 1998
). It has been used to infer evolutionary relationships among early eukaryotes (Hasegawa et al. 1993
; Kamaishi et al. 1996
) and also within insect taxa at different levels (Friedlander, Regier, and Mitter 1992
; Brower and DeSalle 1994
; Friedlander 1994
; Belshaw and Quicke 1997
; Mitchell et al. 1997
; Regier and Shultz 1997
). However, highly conserved protein-coding nuclear genes may also be phylogenetically informative among more recent divergences if they contain more rapidly evolving introns (Clark, Leicht, and Muse 1996
; Kelchner and Clark 1997
; Liss et al. 1997
; Belshaw et al. 1999
; Fabry, Köhler, and Coleman 1999
; Oakley and Phillips 1999
). Sequencing across introns from conserved primer sites in the adjacent exon (exon primed intron crossing [EPIC]) allows these variable regions to be targeted and provides an efficient method of looking at closely related taxa.
Assessing different methods of analyzing variable-length sequences is not easy. These methods optimize different criteria, and all are dependent on both the parameters chosen (especially the cost of gaps as opposed to substitutions) and the computational effort applied to optimization. Morrison and Ellis (1997)
compared alignments with a preferred alignment based on secondary structure, Wheeler (1995)
used congruence with existing taxonomy, and Giribet and Wheeler (1999)
used congruence with morphological characters (in incongruence length difference [ILD] tests). In the present study, it was not possible to use any of these methods. First, we suspected that the reliability of many morphological characters and our preliminary results were clearly incongruent with the traditional classification of the genus Pauesia. We were unable to use measures of character congruence such as the ILD test because the program POY does not work via production of a multiple alignment. Besides, these noncoding sequences do not give rise to a secondary structure. Therefore, we compared different alignment methods using two other measures of congruence: (1) congruence between the exon characters (which have no alignment ambiguity) and trees produced using the different alignment methods on the introns alone, and (2) congruence between pairs of trees produced using the different alignment methods on the two introns.
We are assuming that the exon and the two introns have had the same phylogenetic history and hence should be fully congruent with each other. Incongruence will result from several sources of homoplasy, one of which will be the introduction of errors during alignment. Hence, the more congruent the results given with one alignment method, the more accurate they should be.
Additionally, we thoroughly explored the phylogenetic consequences of changes in the parameters of one commonly used alignment method, the program CLUSTAL W, by creating and analyzing 72 alignments of the whole data set (exon plus introns). For this, we employed different measures for assessing the different sets of parameters (congruence between different tree-building protocols and robustness of each estimate by performing bootstrap tests).
Taxonomic Background
Recent advances in Aphidiinae (Hymenoptera: Braconidae) systematics pose questions about the taxonomic and evolutionary status of many populations. It has been proposed that sympatric population splitting as a consequence of host specialization has led to the formation of rapidly evolving groups of closely related species (Tremblay and Pennacchio 1988
). The genus Pauesia Quilis, 1931, is a good group for testing this phenomenon. The genus has 55 species recognized at present (Medvedev 1995
) and is highly specialized in its host range, with all the species restricted to Lachninae aphids (Homoptera: Aphididae) mostly on Pinatacea (Stary 1960, 1966, 1970, 1976, 1979
). The genus is taxonomically problematic: the phylogenetic relationships between species are unknown and, in practice, it is often impossible to determine species boundaries using morphology (many specimens cannot be ascribed confidently to described species). Although three subgenera are currently recognized (Pauesiella, Pauesia s.s., and Paraphidius), their monophyly has not been convincingly demonstrated.
Within the subgenus Paraphidius, we can recognize one species group, here called Pauesia species group jezoensis, that in western Europe includes such described species as Pauesia jezoensis (=Paraphidius piceaecollis Stary, 1960
), Pauesia pinicollis, and Pauesia cupressobii. The North American species Pauesia ahtanumensis apparently also belongs to this group. Another species group within Paraphidius, here called the Pauesia species group pini, includes published species like Pauesia pini, Pauesia silvestris, Pauesia silana, Pauesia juniperorum, Pauesia similis, and Pauesia alpina, but also the unpublished Pauesia sp. A (on Larix). The morphologically distinctive species Pauesia infulata is also included in the subgenus Paraphidius, but its phylogenetic affinities have not been established.
Within the subgenus Pauesia s.s., the species Pauesia unilachni is morphologically distinctive and has a unique host range (Schizolachnus spp.). In Europe, another two species, Pauesia laricis and Pauesia picta, are assigned to Pauesia s.s. but, although the extreme morphological characters between them are clearly distinctive, intermediate forms appear to occur in nature. Thus, we assign these two species to Pauesia species group laricis. Morphologically, there seems to be a third unpublished taxon belonging to this group, here referred to as Pauesia sp. B (on Pinus nigra nigra).
To elucidate the phylogenetic relationship between the three current subgenera, we also included material of one unpublished species from Syria (Pauesia sp. C), which we placed in the subgenus Pauesiella because of its spatulate ovipositor sheaths (Sedlag and Stary 1980
).
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsliterature cited |
|---|
Taxon Sampling
Parasitized aphids were collected from 20 localities, primarily in Spain (by A.S. and J.M.M.) and Sweden (by U.G.). Individual wasps were reared and placed alive into 100% ethanol. Representatives of two other closely related genera, Aphidius colemani Viereck, 1912, and Pseudopauesia prunicola Halme, 1986, were used as outgroups.
The 38 specimens of Pauesia included in this study provided a good representation of the genus, including 13 of the 22 described European species. We analyzed single individuals, and when multiple individuals of the same putative species were studied, each came from a different locality (details in table 1 ).
|
Specimens were identified using the keys for the group (Stary 1960, 1966, 1970, 1976, 1979
; Medvedev 1995
). Two EF-1 introns, separated by approximately 300 bp, were then sequenced in individual wasps. Intron 1 varied in length between 172 and 232 bp, and intron 2 was between 103 and 244 bp. In addition to the introns, we sequenced adjacent exon regions totaling 227 bp in length (intron lengths and GenBank accession numbers are given in table 1 ).
Laboratory Protocols
Total genomic DNA was isolated from individual wasps as described elsewhere (Sanchis et al. 2000
) and resuspended in 10 µl of TE buffer (Tris HCl 10 mM, EDTA 1 mM [pH 8.0]). The EF-1 was amplified following a technique of nested PCR. Primers used in the first reaction were 5'-AGA TGG CYA ARG GTT CCT TCA A-3' (forward) (Belshaw and Quicke 1997
) and 5'-ATG TGA GCA GTG TGG CAA TCC AA-3' (reverse) ("INDNA35"; John Hobbs, Nucleic Acid Protein Service, University of British Columbia). PCR reactions were carried out in a Perkin-Elmer 2400 thermal cycler in a 50-µl volume containing 0.5 µl DNA extract, 1.25 U Taq polymerase (Amersham Pharmacia Biotech), 20 pmol of each primer, 10 nmol dNTPs (Amersham Pharmacia Biotech), and 5 µl buffer. PCR conditions consisted of (1) an initial denaturation for 2 min at 94°C, (2) 35 cycles of denaturation for 15 s at 96°C, annealing for 1 min at 45°C, and extension for 2 min at 72°C, and (3) a final extension for 7 min at 72°C. Second PCR reactions were carried out with internal primers 5'-GAT GGC ACG GAG ACA ACA TG-3' (forward) and 5'-CCA TTG CTG ATT TGT CCA GGG TGG-3' (reverse), corresponding to positions 2741–2760 and 3735–3758, respectively, in the F2 copy of Drosophila melanogaster (Hovemann et al. 1988
). These amplifications were carried out in a 100-µl volume containing 1 µl of the previous PCR reaction as a template, 2.50 U Taq polymerase, 40 pmol of each primer, 20 nmol dNTPs, and 10 µl buffer. PCR conditions consisted of (1) an initial denaturation for 2 min at 94°C, (2) 35 cycles of denaturation for 15 s at 96°C, annealing for 30 s at 55°C, and extension for 2 min at 72°C, and (3) a final extension for 7 min at 72°C. The products were cut from the agarose gel, purified with the Sephaglas Band Prep kit (Pharmacia), and then sequenced directly on a PE/ABI 373 automated sequencer using ABI PRISM Dye Terminators (Perkin-Elmer). Two other internal primers were used for sequencing, 5'-ACA CCA GTT TCA ACA CGA CC-3' (reverse) and 5'-ACG AAG CTC TCA CTG AAG CCG TTC C-3' (forward).
Sequence Alignment Methods
We used five methods for aligning each intron separately (exons and outgroups excluded):
- A multiple alignment was made using CLUSTAL W, version 1.5. (Thompson, Higgins, and Gibson 1994
). This method produces an alignment from a series of progressive pairwise alignments between sequences and clusters of sequences. The pairwise alignments use the method of Needleman and Wunsch (1970)
, while the clustering order of progressive sequence alignment (Feng and Doolittle 1987
) is determined from a neighbor-joining guide tree. For comparison with other methods, we tested values of 8 for the gap opening penalty and 0.5 for the gap extension penalty. These values represent a compromise between the more restrictive and the more permissive values proposed by some authors (Wheeler 1995
; Morrison and Ellis 1997
).
- A typical "by eye" alignment was made by modification of a default CLUSTAL alignment under a compression length criterion. A minimum number of gaps were inserted manually in order to produce what looked like a reasonable alignment.
- A second "by eye" alignment was made by modification of a default CLUSTAL alignment using a hierarchical approach to overcome the problem caused by variable regions, which can be unambiguously aligned within groups of taxa but not across all taxa. Such regions were aligned independently and also represented as binary characters in a separate data matrix. This was done in a NEXUS file by shifting blocks of bases and inserting gaps in the corresponding positions of the remaining taxa. No homoplasy was found in this second matrix, and a constraint tree was created from it. Under this constraint, the most parsimonious tree was then found as described below for all the sequences, with gaps treated as missing data. Where gaps occurred within the unambiguously aligned blocks, they were coded as additional single binary characters and included in the analysis (note that the constraint tree was written manually and taxa sharing regions of unambiguous alignment were treated as monophyletic, which cannot be demonstrated in an unrooted analysis). Regions that could not be unambiguously aligned even within groups were excluded from the analysis.
- A multiple alignment was made using MALIGN, version 2.7, parallel version 1.5 (Wheeler and Gladstein 1994
). In contrast to CLUSTAL, this method utilizes parsimony as the optimality criterion. The gap cost was set to twice the substitution cost, and 100 random additions were used with limited alignment swapping (commands "quick" and "alignswap"; smaller numbers of random additions, with more exhaustive swapping, failed to find shorter alignments).
- A direct optimization procedure (see Wheeler 1996
) implemented in POY native code, version 2.0, by David Gladstein and Ward Wheeler (obtained from anonymous ftp.amnh.org/pub/wheeler/programs/poy) was used. We performed three analyses with gap costs set at one, two, and three times that of a substitution. For each gap cost, we performed 100 random additions holding a maximum of 100 trees (branch swapping with both SPR and TBR).
All of the methods except POY produced a multiple alignment from which we built MP trees using PAUP* (Swofford 1998
) (equal weighting and with 500 random additions, branch swapping using TBR on a maximum of 10 trees up to a limit of 1,000 trees). Additionally, for all methods except POY and the hierarchic "by eye" method, we treated gaps both as missing data and as fifth bases. No regions were excised except in method 3. All alignments are available on http://www.bio.ic.ac.uk/research/data/pauesia. The 27 alignment positions that appeared heterozygous were coded as missing data. No heterozygosity for length was present except for one unreadable region (around 150 bp) in some individuals (P. picta 1, 2, 3 and P. picta/P. laricis), which might be a sequencing artifact caused by homopolymers.
Comparison of Alignment Methods
Two measures of congruence were used to compare the five alignment methods:
- Congruence between the 227 exon characters and trees produced using the different alignment methods on the introns alone. For each of the two introns, we built trees using the five alignments described above. Also using PAUP*, we then measured the lengths of these trees according to the exon characters only, which were treated as unordered characters (i.e., we counted the minimum number of changes required by mapping these characters onto the trees). The most congruent alignment methods are those that give the lowest values (i.e., require the fewest changes).
- Comparison of the topological congruence (a measure of similarity between different topologies) between trees produced separately from the two introns using the different alignment methods. For this, we chose to use the partition metric of Penny and Hendy (1985)
, implemented as the symmetric distance in PAUP*, and give the modal value for ease of calculation.
Phylogenetic Analysis of Complete Data Set
After measuring the congruence of the different alignment methods, we combined in a series of data matrices the exon data plus the two introns aligned using the five methods. On the basis of the complete phylogenetic information, we reconstructed new maximum-parsimony (MP) trees in order to examine their node sensitivity.
An additional method of examining the sequence alignment process is to choose a single alignment algorithm and vary its parameter values (e.g., Wheeler 1995
; Morrison and Ellis 1997
). In this way, it is possible to cover most of the phylogenetic variation concerning alignment (Wheeler 1995
). Following Morrison and Ellis (1997)
, we chose the alignment program CLUSTAL W and examined in detail the sensitivity of the phylogenetic estimates to changes in some of the multiple-alignment parameters. We focused on the gap cost ratios, because they are recognized as one of the most important alignment parameters (Tyson 1992
; Vingron and Waterman 1994
; Wheeler 1995
; Morrison and Ellis 1997
). There is no way of determining these values a priori (Rinsma-Melchert 1993
), and hence we followed Morrison and Ellis' (1997)
strategy of logarithmically varying the gap opening penalty (GOP) and the gap extension penalty (GEP). We varied the GOP (the cost of inserting a new gap into a sequence) from 0.5 to 64 times the cost of a substitution (log2 GOP = -1, 0, 1, 2, 3, 4, 5, 6). The GEP (the cost of extending an existing gap) was varied from 0.031 to 8 times the cost of a substitution (log2 GEP = -5, -4, -3, -2, -1, 0, 1, 2, 3). We tested all possible combinations of these values.
In these analyses, we included the complete data set (introns plus exons) and the three outgroups. We also tested the sensitivity of the 72 generated alignments to tree-building protocol. For this purpose, three common methods were used: neighbor joining (NJ), MP, and maximum likelihood (ML).
NJ analyses were performed using the MEGA program, version 1.01 (Kumar, Tamura, and Nei 1993
); distances were estimated using the Jukes-Cantor correction (Jukes and Cantor 1969
). MP analyses were performed using PAUP*. We employed unweighted parsimony and three weighting schemes: (1) intron versus exon positions weighted 2 to 1, (2) first plus second codon positions versus third plus intron positions weighted 2 to 1, and (3) first plus second codon positions versus third plus intron positions weighted 5 to 1. ML analyses were performed using DNAML (PHYLIP, version 3.53c; Felsenstein 1993
). Bootstrapping (Efron 1982
) was used to establish the relative support of each node in the trees. Gaps were treated as missing data in this set of analyses.
Constraint Analysis
To measure the support for groupings recovered in our analyses, we used the Kishino-Hasegawa (Kishino and Hasegawa 1989
) and Templeton (1983)
tests as implemented in the program package PHYLIP, version 3.53c. We also constrained one topology representing the monophyly of the traditional subgenus Paraphidius; we employed the previous tests in order to compare this topology with the optimal topologies obtained in our analyses.
|
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsliterature cited |
|---|
Comparison of Alignment Methods
The EF-1
introns had a high nucleotide bias with several runs of A's and/or T's (A+T = 72%). This made aligning the sequences by eye very difficult but—perhaps surprisingly—these alignments scored well in both measures of congruence (table 2 ). In fact, there was no striking difference among the different methods, and we found considerable variation in the performance of exon characters on different equally parsimonious trees produced by each method. Two of the specimens (P. juniperorum 1 and 2) had particularly aberrant second-intron sequences (table 1
), so we repeated the analyses pruning these from the trees: the results obtained were similar (not shown). Thus, our analyses did not show any method as being markedly more congruent than the others.
|
Phylogenetic Analysis of Complete Data Set
Figure 1 shows the strict consensus of all MP trees built from the five multiple-alignment strategies. Regardless of alignment method, we found two main clades (1 and 2), which, in turn, contained six clades plus three branches represented by single specimens. The relationships among these nine branches were very sensitive to alignment protocol (see below).
|
We focused our attention on phenetic-based criteria (less computationally demanding) for examining the sensitivity of our data to tree-building protocol. Method 2 obtained some of the best scores and produced a reasonable number of MP trees (table 2 ). However, in order to avoid the subjectivity of dealing with "by eye" alignments, we based the NJ, MP, and ML reconstructions on method 1 as well as method 2. Hence, we worked with 796 (after removing gap columns in the 918 initially aligned positions) and 756 aligned positions, respectively. In spite of their similar lengths, these two alignments differed in their phylogenetic estimates. Figure 2 shows the NJ tree obtained from the CLUSTAL alignment (method 1). The figure legend gives details of how this topology varied when other tree-building protocols were used on this alignment and when other alignment methods were employed. It is clear that the branches that were sensitive to tree-building protocol were the same as those that were sensitive to alignment method. Within clade 1, the relationship between P. unilachni, Pauesia sp. B, and Pauesia sp. C could not be established, nor could it be determined whether these species belonged to the clade ahtanumensis-jezoensis-pinicollis-cupressobii (AJPC) or to the clade laricis-picta (LP). Similarly, within clade 2, the relationships among the clade infulata-juniperorum (IJ), the clade sp. A, the clade silvestris-silana-pini (SSP), and the clade similis-alpina (SA) could not be inferred reliably.
|
Sequence Gap Penalties
The combination of the GOP/GEP values in CLUSTAL produced results similar to those produced by variation in both the alignment method and the tree-building protocol. The 216 trees showed the same basic structure (figs. 1 and 3 ) with the same rearrangements in the tips (fig. 2 ). Operating in this way, however, we were able to recover a greater variation in the relationships among sensitive clades than was recovered by trying several alignment methods (figs. 4 and 5 ).
|
Realignment and Analyses Within Supported Clades
After finding two main clades (robust both to alignment and to tree-building protocol), we realigned separately the sequences belonging to each clade in order to find more accuracy. The new alignments were performed using the default parameters of CLUSTAL W. The tree-building—by NJ, ML, and unweighted MP—was performed as explained above.
For clade 2, the resulting alignment contained 686 positions (with 113 variable and 71 informative). The three tree-building methods gave the same topology 2 (fig. 5A ), but only the branch splitting the clade infulata-juniperorum (IJ) was bootstrap-supported (91% by NJ and 87% by MP); unweighted MP resulted in two trees of length 167 and a consistency index (CI) of 0.910.
|
Regarding clade 1, the new alignment did not resolve the position of the three problematic branches. Excluding the more variable regions from the analyses did not lead to the complete resolution of these branches. Nevertheless, most of these trees reproduced the same topology 1 (fig. 4A ) (data not shown), which would be quite in agreement with the current classification for the genus Pauesia (both P. unilachni and Pauesia sp. B are ascribed to the subgenus Pauesia s.s.).
Hereinafter, we use the robust relationships established in topology 2 (fig. 5A ) and the higher affinity of P. unilachni and Pauesia sp. B to Pauesia sg. laricis as criteria for testing the taxonomic congruence of the estimates obtained by varying the CLUSTAL parameters.
Comparison of Sequence Gap Penalties and Tree-Building Protocols
As expected, increasing the GOP and the GEP relative to the cost of a substitution decreased the resulting aligned sequence length, and there was a convergence for this data set to 750 nucleotide positions. We also observed a periodicity in the parameters defining both trees and alignments. This periodicity was dependent on the GOP values (fig. 6
). Thus, the higher the GOP value, the higher the number of variable and informative positions and the higher the likelihood of the tree, but also the shorter the alignment length and the lower the CI. However, this correlation between parameters did not correspond with topology similarity, indicating that the GEP also influences the phylogenetic inference. Consequently, there are no apparent means of predicting cladogram similarity from the gap penalty values (figs. 4 and 5
).
|
Therefore, we decided to check which GOP/GEP combinations led to the most congruent and best bootstrap-supported trees. We compared the resulting topologies under the taxonomic criteria defined above.
Congruence between tree-building protocols was measured as the total number of different topologies obtained by NJ, MP, and ML with each parameter combination (counting topologies for clade 1 plus those for clade 2; figs. 4 and 5 , respectively). These numbers ranged from a minimum of three topologies to a maximum of six. The highest congruence (only three different topologies) was reached with the log2 GOP/log2 GEP combinations of -1/-5, 1/-5, -1/-4, 1/-4, -1/-3, 1/-3, 0/-2, and 1/-2 (corresponding to the alignments numbered 1, 3, 9, 11, 17, 19, 26, and 27 in fig. 6 ). All these combinations reconstructed the same topologies: (1) topology 3 for clade 1 (fig. 4A ), and (2) topology 2 (by ML and MP) and topology 5 (by NJ) for clade 2 (fig. 5A ). These topologies matched most of the majority-rule consensus trees shown in figure 3 and exhibited some of the highest CIs. However, the parameters employed to compare phylogenetic trees (e.g., CI and likelihood (-ln L)) are usually correlated with the number of aligned positions, and hence it is not possible to use them as a criterion for choosing the best alignments. Also, in clade 1, these topologies were incongruent with the taxonomic criteria discussed above.
|
Regarding the robustness of the phylogenetic estimates, we performed bootstrap tests in the 72 NJ analyses (fig. 3 ). We found a bias in the distribution of topologies with bootstrap-supported nodes (figs. 4A, 4B, 5A, and 5B ). Hence, most restrictive gap penalties supported topology 1 within clade 1 (fig. 4A and B ). The bias within clade 2 was not apparent (fig. 5A and B ). Nevertheless, the GOP/GEP combinations leading to supported relationships within clade 1 were not coincident with those combinations supporting nodes within clade 2 (except for the log2 GOP/log2 GEP combination of 4/0). In general, the bootstrap-supported nodes were congruent with our taxonomic criteria. In only four GOP/GEP combinations were some incongruencies detected within clade 2. Such incongruencies dealt with topologies 4 and 5 showing their internal nodes to be bootstrap-supported (fig. 5A and B ). Comparing the different reconstruction methods, the log2 GOP/log2 GEP combination (4/0) leading to the most robust NJ topology (i.e., the topology with the greatest number of bootstrap-supported internal nodes) recovered topologies that were incongruent with each other and with our taxonomic criteria (e.g., topology 3 for clade 2; fig. 5A ). This shows that bootstrap values of single alignments do not necessary reflect their robustness to variation in alignment parameters.
Out of eight different topologies obtained for clade 1 (fig. 4A ) and seven topologies for clade 2 (fig. 5A ), some of them were totally incongruent with those produced with the within-clade alignments and with our taxonomic criteria. Regarding clade 2, these topologies were 1, 3, 6, and 7 (fig. 5A ). Due to the low resolution gained for clade 1, we lacked an explicit criterion for discarding topologies. We tentatively discard topologies 3, 6, 7, and 8 as the most unrealistic. Taking these things into consideration, the comparison of tree-building protocols showed the ML estimation to be the most sensitive to alignment-parameter variation and to lead to many incongruent topologies. NJ recovered a large number of congruent topologies but never succeeded in recovering the preferred topology (topology 2) for clade 2, and MP showed a contradictory result, as it recovered the highest number of different topologies for clade 1 (most of them incongruent with the preferred topology) but the lowest for clade 2 (largely congruent with topology 2).
Kishino-Hasegawa and Templeton Tests
Due to the paraphyletic status of subgenus Paraphidius, found in all our topologies, we questioned the current subgeneric classification within the genus Pauesia. Hence, we performed the tests of Kishino and Hasegawa (1989)
and Templeton (1983)
to check whether or not this apparent paraphyly could be due to chance. In order to discard background noise, we constrained eight topologies including the main representatives of each clade: two outgroups (A. colemani 1 and 2), three specimens of Pauesia sg. jezoensis (P. ahtanumemsis, P. pinicollis 1, and P. cupresobii 2), three specimens of Pauesia sg. laricis (P. laricis and P. picta 2 and 3), two specimens of the IJ clade (P. infulata and P. juniperorum 1), one specimen of the SA clade (P. similis 1), one specimen of the SSP clade (P. silvestris), and one specimen of the clade sp. A (Pauesia sp. A 1). Pauesia unilachni, Pauesia sp. B, and Pauesia sp. C were not included in these topologies because their positions were not resolved in our analyses, and these tests do not accept polytomies.
Except for the eighth topology, all the topologies showed P. sg. jezoensis and P. sg. laricis as sister groups forming clade 1 (a relationship obtained in all our analyses), and for clade 2, each numbered topology in table 3 reproduced the corresponding topology in figure 5A . The eighth topology forced the subgenus Paraphidius (formed by P. sg. jezoensis plus of all the clusters forming our clade 2) and the subgenus Pauesia s.s. (formed by P. sg. laricis) to be monophyletic.
|
Basing on the CLUSTAL alignment (method 1), we compared all eight of these topologies by performing the previous tests. The eighth topology was clearly rejected by both tests, so we cannot accept the monophyly of the subgenus Paraphidius (this eighth topology was also rejected when the tests were based on the "by eye" alignment [method 2]; data not shown). The rest of topologies were considered statistically similar by the tests (except for topology 7, which was rejected by Templeton's test; table 3 ).
|
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsliterature cited |
|---|
Variable-Length Sequence Alignment
Under the different alignment algorithms tested, all of the topologies showed considerable similarity (figs. 1 and 3 ). This shows that a strong phylogenetic signal is present in the EF-1
introns. Both the basic structure (fig. 1
) and the minor rearrangements in the tips (fig. 2
) were always recovered and were usually supported by high bootstrap values (>70%; Berry and Gascuel 1996
). Testing the internal consistency of the five alignment methods revealed no striking difference between them. All of them showed considerable variation both in the performance of exon characters (on different equally parsimonious trees produced by each method) and between variants of the same method. The most striking result was that alignments made by eye and by computationally less demanding programs such as CLUSTAL appeared to perform as well as those produced by the more theoretically justifiable algorithms. With regard to "by eye" alignments, we recognize that they have the disadvantage of being unrepeatable and dependent on the investigator's criterion. However, their phylogenetic estimations deserve consideration if they are done blind (i.e., aligned without knowledge of the sequence names) and their results are contrasted with others coming from a different alignment algorithm. In addition, alignment algorithms sometimes produce misleading solutions (e.g., insertion of gap columns) that need to be manually refined.
Most of the prominent differences between alignment strategies, dealing with the three problematic taxa of clade 1 and with the relationships among the four monophyletic clusters of clade 2, were also found when different tree-building protocols were used on an alignment. Hence, most of the ML and MP analyses recovered P. unilachni, Pauesia sp. C, and Pauesia sp. B in the same cluster as P. sg. jezoensis (87% and 73% in the consensus tree, respectively; fig. 3 ), but the NJ analyses tended to recover them with P. sg. laricis (57%; fig. 3 ). This latter relationship was supported most of the time by high bootstrap values (figs. 3 and 4B ). On the other hand, the MP analyses revealed a bias favoring topology 2 within clade 2 (fig. 5A and D ).
The subsequent within-group alignments reduced the number of ambiguous positions (at least for clade 2), and the topologies estimated with these new alignments gained robustness and resolution. Based on these alignments, we propose topology 2 (fig. 5A ) as the most probable taxonomic hypothesis for the relationships within clade 2. If topology 2 is the best estimate, then in our study the most restrictive CLUSTAL parameters led to the most incongruent alignments. On the same basis, ML appears to be the most alignment-sensitive reconstruction method, and MP appears to be the most robust.
We also find, like Morrison and Ellis (1997)
, that testing alignment strategy is as important as tree-building protocol. Both trying different alignment methods and varying parameter values in the same alignment algorithm seem to be similarly effective in revealing the sensitive branches. Regarding CLUSTAL, there is apparently no such thing as a topological distribution pattern depending on the parameter values (also found by Morrison and Ellis 1997
). Therefore, exhaustive searches of the parameter space seem to be necessary. However, we could have detected the strength of the phylogenetic signal contained in our data just by testing the three different tree-building protocols with six of the GOP/GEP combinations. These six combinations would include three GOP values (representing strong, medium, and weak gap penalties) versus two GEP values (paying attention to the periodicity observed in fig. 6
). In the present case, we would discard the more restrictive parameters (both log2 GOP = 6 and log2 GEP = 3), as they led mostly to incongruent topologies (see figs. 4 and 5
).
Phylogeny of European Pauesia Species
There are several conclusions regarding the phylogeny of Pauesia that are not sensitive to either alignment or tree-building method. First, the genus Pauesia seems to be monophyletic with respect to Pseudopauesia. Only under a few extreme alignment parameters was Pseudopauesia recovered within Pauesia (e.g., with the most restrictive GOP value but never combined with the more restrictive GEP; see fig. 5B and D
).
Second, we found two well-supported sister clades (fig. 2 ). Clade 1 includes two species groups, (1) P. sg. jezoensis and (2) P. sg. laricis, plus three other species (P. unilachni, Pauesia sp. C, and Pauesia sp. B) whose relationships could not be clarified. Both our sequences and the morphological characters (not shown) indicate that Pauesia sp. B and Pauesia sp. C represent valid species. However, the relationship of Pauesia sp. B within the group P. sg. laricis and Pauesia sp. C within the subgenus Pauesiella are at present uncertain.
Clade 2 includes the species of Pauesia sg. pini as well as P. infulata (table 1 ). This clade contains four subclades: (1) similis-alpina, (2) sp. A, (3) silvestris-silana-pini, and (4) infulata-juniperorum. Although relationships among these subclades were sensitive both to alignment and to tree-building method, further analyses (based on alignments within clade 2) indicated that the subclade formed by the species P. infulata plus P. juniperorum forms a sister group to the clade P. sg. pini (SSP + SA + SP. A). Interestingly, P. juniperorum appears to be a well-supported sister species to P. infulata.
All of the analyses showed that the current classification with three subgenera (Paraphidius, Pauesia s.s., and Pauesiella) is not reliable, at least with reference to the Paraphidius monophyly (table 3 and fig. 2 ). In the absence of a clear morphological distinction between clades 1 and 2, we can only recommend that the current subgeneric classification be abandoned.
Species Boundaries
Following alignment, we examined groups of similar sequences to estimate where the species boundaries lay. Although we cannot confidently draw species boundaries from only a few individuals sequenced at a single locus, we can make some inferences. Belshaw et al. (1999)
sequenced EF-1 intron 2 in 12 individuals of another aphidiine species of the genus Lysiphlebus, each from a different locality across western Europe, and found no variation, indicating that the introns may be relatively conserved at the species level.
Within the P. sg. jezoensis species complex, the five P. pinicollis/P. jezoensis individuals are distinguished from the three P. cupressobii individuals by a fixed 2-bp indel, which may mark a species boundary. The status of P. pinicollis and P. jezoensis remains uncertain: individuals are distinguished by only a single substitution. The North American P. ahtanumensis is a clearly distinctive species, always appearing basal to the other members of this group. Hence, we are probably dealing with three or four species within this clade, depending on the status of P. pinicollis.
The P. sg. laricis species complex constitutes the most problematic area in our study. Although these sequences cluster together in most phylogenetic analyses (often with high bootstrap support), we found extensive variation between all of the sequences. The most similar individuals (P. picta 2 and 3) are distinguished in the second intron by a short region that contains perhaps 10 fixed substitutions or indels (it is difficult to align). Also, P. picta/P. laricis and P. picta 1, 2, and 3 are characterized by the presence of one homopolymer in intron 1. Therefore, the traditional division between P. laricis and P. picta, the two recognized species, is not supported, and we cannot ignore the possibility that each individual in this complex represents a different species (from a minimum of three to a maximum of six).
Sequences of the five individuals identified as P. similis and P. alpina (the similis-alpina clade) are identical except at one exon position in P. similis 3, and this position is clearly homoplastic with the alternate allele shared with P. sg. laricis and P. prunicola (the outgroup). Hence, all five individuals may well belong to a single species or form a species complex that would deserve the name Pauesia species group similis.
The three individuals identified as Pauesia sp. A are identical, and the only difference between them and P. cupresobii 4 is an additional T in a T run (which may be an artifact). Therefore, all of them probably belong to the same species (still undescribed).
Three species can be drawn within the silvestris-silana-pini clade. Nevertheless, from the present data arises the question of whether P. silana could be conspecific with P. silvestris.
In conclusion, we believe that our 38 Pauesia individuals, although morphologically referable to 13 described and 3 new species, probably represent between 15 and 21 species which conflict with the morphologically defined species boundaries.
|
|
Acknowledgements |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsliterature cited |
|---|
This work was funded by the Conselleria d'Educació i Ciència (GV-3216/95), by grant PB96-0793C04-01 from DGES, and by NERC in the U.K. We thank P. González-Funes, S. Fowler, S. Shaw, P. Stary, and W. Völkl for material; D. Swofford and W. Wheeler for their programs, PAUP and POY, respectively; and F. González-Candelas and three reviewers for their useful comments on an earlier draft. The facilities at SCSIE (Universitat de València) were used for sequencing, and the Servei de Bioinformàtica provided computer support.
|
Footnotes |
|---|
Manolo Gouy, Reviewing Editor
1 Keywords: phylogenetics
variable-length sequences
alignment strategies
tree-building methods
Aphidiinae
Pauesia
EF-1 introns
elongation factor–1 
2 Address for correspondence and reprints: Amparo Latorre, Institut Cavanilles de Biodiversitat i Biologia Evolutiva, Genetica Evolutiva, Universitat de València, Edificio Institutos Investigación, Campos Paterna, Apdo. Correos 22085, 46071 Valencia, Spain. amparo.latorre{at}uv.es
|
literature cited |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsliterature cited |
|---|
Belshaw, R., D. L. J. Quicke. 1997. A molecular phylogeny of the Aphidiinae (Hymenoptera: Braconidae). Mol. Phylogenet. Evol. 7:281–293[Web of Science][Medline]
Belshaw, R., D. L. J. Quicke, W. Völkl, C. J. Godfray. 1999. Molecular markers indicate rare sex in a predominantly asexual parasitoid wasp. Evolution. 53:1189–1199
Berry, V., O. Gascuel. 1996. On the interpretation of bootstrap trees: appropriate threshold of clade selection and induced gain. Mol. Biol. Evol. 13:999–1011
Brands, J. H. G. M., J. A. Maassen, F. J. Van-Hemert, R. Amons, W. Möller. 1986. The primary structure of the alpha subunit of human elongation factor 1: structural aspects of guanine nucleotide binding sites. Eur. J. Biochem. 155:167–172[Web of Science][Medline]
Brower, A. V. Z., R. DeSalle. 1994. Practical and theoretical considerations for choice of a DNA sequence region in insect molecular systematics, with a short review of published studies using nuclear gene regions. Ann. Entomol. Soc. Am. 87:702–716
Clark, A. G., B. G. Leicht, S. V. Muse. 1996. Length variation and secondary structure of introns in the Mlc1 gene in six species of Drosophila.. Mol. Biol. Evol. 13:471–482[Abstract]
Danforth, B. N., S. Ji. 1998. Elongation factor-1 occurs as two copies in bees: implications for phylogenetic analysis of EF-1 sequences in insects. Mol. Biol. Evol. 15:225–235[Abstract]
Efron, B.. 1982. The jackknife, the bootstrap and other resampling plansSIAM, Philadelphia
Fabry, S., A. Köhler, A. W. Coleman. 1999. Intraspecies analysis: comparison of ITS sequence data and gene intron sequence data with breeding data for a worldwide collection of Gonium pectorale.. J. Mol. Evol. 48:94–101[Web of Science][Medline]
Felsenstein, J.. 1993. PHYLIP (phylogeny inference package)Version 3.5c. Distributed by the author, Department of Genetics, University of Washington, Seattle
Feng, D. F., R. F. Doolittle. 1987. Progressive sequence alignment as a prerequisite to correct phylogenetic trees. J. Mol. Evol. 25:351–360[Web of Science][Medline]
Friedlander, T. P.. 1994. Phylogenetic information content of five nuclear gene sequences in animals: initial assessment of character sets from concordance and divergence studies. Syst. Biol. 43:511–525
Friedlander, T. P., J. C. Regier, C. Mitter. 1992. Nuclear gene sequences for higher level phylogenetic analysis: 14 promising candidates. Syst. Biol. 41:483–490
Giribet, G., W. C. Wheeler. 1999. On gaps. Mol. Phylogenet. Evol. 13:132–143[Web of Science][Medline]
Hasegawa, M., T. Hashimoto, J. Adachi, N. Iwabe, T. Miyata. 1993. Early branchings in the evolution of eukaryotes: ancient divergence of entamoeba that lacks mitochondria revealed by protein sequence data. J. Mol. Evol. 36:380–388[Web of Science][Medline]
Hovemann, B., S. Richter, U. Walldorf, C. Cziepluch. 1988. Two genes encode related cytoplasmic elongation factors 1 (EF-1) in Drosophila melanogaster with continuous and stage specific expression. Nucleic Acids Res. 16:3175–3194
Jukes, T. H., C. R. Cantor. 1969. Evolution of protein moleculesPp. 21–132 in H. N. Munro, ed. Mammalian protein metabolism. Academic Press, New York
Kamaishi, T., T. Hashimoto, Y. Nakamura, F. Nakamura, S. Murata, N. Okada, K. I. Okamoto, M. Shimizu, M. Hasegawa. 1996. Protein phylogeny of translation elongation factor EF-1 suggests microsporidians are extremely ancient eukaryotes. J. Mol. Evol. 42:257–263[Web of Science][Medline]
Kelchner, S. A., L. G. Clark. 1997. Molecular evolution and phylogenetic utility of the chloroplast rpl16 intron in Chusquea and the Bambusoideae (Poaceae). Mol. Phylogenet. Evol. 8:385–397[Web of Science][Medline]
Kishino, H., M. Hasegawa. 1989. Evaluation of the maximum likelihood estimate of the evolutionary tree topologies from DNA sequence data, and the branching order in Hominoidea. J. Mol. Evol. 29:170–179[Web of Science][Medline]
Kjer, K. M.. 1995. Use of rRNA secondary structure in phylogenetic studies to identify homologous positions: an example of alignment and data presentation from the frogs. Mol. Phylogenet. Evol. 4:314–330[Web of Science][Medline]
Kumar, S., K. Tamura, M. Nei. 1993. MEGA: molecular evolutionary genetic analysisVersion 1.0. Pennsylvania State University, University Park
Lenstra, J. A., A. Van Vliet, A. C. Arnberg, F. J. Van Hemert, W. Möller. 1986. Genes coding for the elongation factor EF-1 in Artemia.. Eur. J. Biochem. 155:475–484[Web of Science][Medline]
Liss, M., D. L. Kirk, K. Beyser, S. Fabry. 1997. Intron sequences provide a tool for high-resolution phylogenetic analysis of volvocine algae. Curr. Genet. 31:214–227[Web of Science][Medline]
Maddison, D. R., M. D. Baker, K. A. Ober. 1999. Phylogeny of carabid beetles as inferred from 18S ribosomal DNA (Coleoptera: Carabidae). Syst. Entomol. 24:103–138
Medvedev, G. S.. 1995. Keys to the fauna of the URSSVol. III. Hymenoptera. Part V. Science Publishers, Leningrad
Mitchell, A., S. Cho, J. C. Regier, C. Mitter, R. W. Poole, M. Mathews. 1997. Phylogenetic utility of elongation factor-1 in Noctuoidea (Insecta: Lepidoptera): the limits of synonymous substitution. Mol. Biol. Evol. 14:381–390[Abstract]
Morrison, D. A., J. T. Ellis. 1997. Effects of nucleotide sequence alignment on phylogeny estimation: a case study of 18S rDNAs of Apicomplexa. Mol. Biol. Evol. 14:428–441[Abstract]
Needleman, S. B., C. D. Wunsch. 1970. A general method applicable to the search for similarities in the amino acid sequences of two proteins. J. Mol. Biol. 48:444–453
Oakley, T. H., R. B. Phillips. 1999. Phylogeny of salmonine fishes based on growth hormone introns: Atlantic (Salmo) and Pacific (Oncorhynchus) salmon are not sister taxa. Mol. Phylogenet. Evol. 11:381–393[Web of Science][Medline]
Olsen, G. J., C. R. Woese. 1993. Ribosomal RNA: a key to phylogeny. FASEB J. 7:113–123[Abstract]
Penny, D., M. D. Hendy. 1985. The use of tree comparison metrics. Syst. Zool. 34:75–82
Rao, T. R., L. Y. Slobin. 1986. Structure of the amino-terminal end of mammalian elongation factor Tu. Nucleic Acids Res. 14:2409
Regier, J. C., J. W. Shultz. 1997. Molecular phylogeny of the major arthropod groups indicates polyphyly of the crustaceans and a new hypothesis for the origin of hexapods. Mol. Biol. Evol. 14:902–913[Abstract]
Rinsma-Melchert, Y.. 1993. The expected number of matches in optimal global sequence alignments, N. Z. J. Bot. 31:219–230
Roth, W. W., P. W. Bragg, M. V. Corrias, N. S. Reddy, J. N. Dholakia, J. A. Wahba. 1987. Expression of a gene for mouse eukaryotic elongation factor TU during murine erythroleukemic cell differentiation. Mol. Cell. Biol. 7:3929–3936
Sanchis, A., A. Latorre, F. Gonzalez-Candelas, J. M. Michelena. 2000. An 18S rDNA based molecular phylogeny of Aphidiinae (Hymenoptera: Braconidae). Mol. Phylogenet. Evol. 14:180–194[Web of Science][Medline]
Sedlag, U., P. Stary. 1980. Pauesia (Pauesiella subg.n.) spatulata sp.n., a parasitoid of Cinara aphids from Central Europe (Hymenoptera: Aphidiidae; Homoptera: Lachninae). Acta Entomol. Bohem. 77:383–386
Stary, P.. 1960. A taxonomic revision of the European species of the genus Paraphidius Stary, 1958. Acta Faun. Entomol. Mus. Nat. Pragae. 6:5–38
———.1966. The Aphidiidae of Italy (Hymenoptera: Ichneumonoidea). Boll. Ist. Entomol. Univ. Bologna. 28:65–139
———.1970. Biology of aphid parasites (Hymenoptera: Aphidiidae) with respect to integrated controlW. Junk, B. V. Publishers, the Hague
———.1976. Aphid parasites (Hymenoptera: Aphidiidae) of the Mediterranean areaW. Junk, B. V. Publishers, the Hague
———.1979. Aphid parasites (Hymenoptera: Aphidiidae) of Central Asian areaW. Junk, B. V. Publishers, the Hague
Swofford, D. L.. 1998. PAUP*Phylogenetic analysis using parsimony (*and other methods). Version 4. Sinauer, Sunderland, Mass
Swofford, D. L., G. J. Olsen, P. J. Waddell, D. M. Hillis. 1996. Phylogenetic inferencePp. 407–514 in D. M. Hillis, C. Moritz, and B. K. Mable, eds. Molecular systematics. Sinauer, Sunderland, Mass
Templeton, A. R.. 1983. Phylogenetic inference from restriction endonuclease cleavage site maps with particular reference to the evolution of human and apes. Evolution. 37:221–224[Web of Science]
Thompson, J. D., D. G. Higgins, T. J. Gibson. 1994. CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position specific gap penalties and weight matrix choice. Nucleic Acids Res. 22:4673–4680
Tremblay, E., F. Pennacchio. 1988. Speciation in aphidiine Hymenoptera (Hymenoptera: Aphidiidae)Pp. 139–146 in V. K. Gupta, ed. Advances in parasitic Hymenoptera research. E. J. Brill, Leiden
Tyson, H.. 1992. Relationships between amino acid sequences determined through optimum alignments, clustering, and specific distance patterns: application to a group of scorpion toxins. Genome. 35:360–371[Medline]
Vingron, M., M. S. Waterman. 1994. Sequence alignment and penalty choice: review of concepts, case studies and implications. J. Mol. Biol. 235:1–12[Web of Science][Medline]
Walldorf, U., B. T. Hovemann. 1990. Apis mellifera cytoplasmic elongation factor 1- (EF-1 ) is closely related to Drosophila melanogaster EF-1. FEBS Lett. 267:245–249[Web of Science][Medline]
Wenzel, J. W., M. E. Siddall. 1999. Noise. Cladistics. 15:51–64[Web of Science]
Wheeler, W. C.. 1995. Sequence alignment, parameter sensitivity, and the phylogenetic analysis of molecular data. Syst. Biol. 44:321–331[Web of Science]
———.1996. Optimization alignment: the end of multiple sequence alignment in phylogenetics?. Cladistics. 12:1–9
Wheeler, W. C., D. S. Gladstein. 1994. MALIGN: a multiple sequence alignment program. J. Hered. 85:417–418
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  E. Benavides, R. Baum, D. McClellan, and J. W. Sites Molecular Phylogenetics of the Lizard Genus Microlophus (Squamata:Tropiduridae): Aligning and Retrieving Indel Signal from Nuclear Introns Syst Biol, October 1, 2007; 56(5): 776 - 797. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. B. Smythe, M. J. Sanderson, and S. A. Nadler Nematode Small Subunit Phylogeny Correlates with Alignment Parameters Syst Biol, December 1, 2006; 55(6): 972 - 992. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Creer, C. E. Pook, A. Malhotra, and R. S. Thorpe Optimal Intron Analyses in the Trimeresurus Radiation of Asian Pitvipers Syst Biol, February 1, 2006; 55(1): 57 - 72. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Y. Ogurtsov, S. Sunyaev, and A. S. Kondrashov Indel-Based Evolutionary Distance and Mouse-Human Divergence Genome Res., August 1, 2004; 14(8): 1610 - 1616. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Pons, T. G. Barraclough, K. Theodorides, A. Cardoso, and A. P. Vogler Using Exon and Intron Sequences of the Gene Mp20 to Resolve Basal Relationships in Cicindela (Coleoptera:Cicindelidae) Syst Biol, August 1, 2004; 53(4): 554 - 570. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Kawakita, T. Sota, J. S. Ascher, M. Ito, H. Tanaka, and M. Kato Evolution and Phylogenetic Utility of Alignment Gaps Within Intron Sequences of Three Nuclear Genes in Bumble Bees (Bombus) Mol. Biol. Evol., January 1, 2003; 20(1): 87 - 92. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||