© 2004 Society of Systematic Biologists
Phylogenetic Artifacts Can be Caused by Leucine, Serine, and Arginine Codon Usage Heterogeneity: Dinoflagellate Plastid Origins as a Case Study
Edited by Gavin Naylor: Associate Editor
1 Department of Bioscience, Nagahama Institute of Bioscience and Technology 1266 Tamura Nagahama Shiga 526–0829 Japan; E-mail: yinagai{at}dal.ca (Y.I.)
2 Canadian Institute for Advanced Research, Program in Evolutionary Biology, Department of Biology, Dalhousie University Halifax Nova Scotia B3H 4J1 Canada
 |
Abstract |
|---|
 TopNotes Abstract MethodsResults and DiscussionConclusionsAppendixAcknowledgmentsReferences |
|---|
Phylogenetic analyses of first and second codon positions (DNA1 + 2 analysis) and amino acid sequences (protein analysis) are often thought to provide similar estimates of deep-level phylogeny. However, here we report a novel artifact influencing DNA level phylogenetic inference of protein-coding genes introduced by codon usage heterogeneity that causes significant incongruities between DNA1 + 2 and protein analyses. DNA1 + 2 analyses of plastid-encoded psbA genes (encoding of photosystem II D1 proteins) strongly suggest a relationship between haptophyte plastids and typical (peridinin-containing) dinoflagellate plastids. The psbA genes from haptophytes and a subset of the peridinin-type plastids display similar codon usage patterns for Leu, Ser, and Arg, which are each encoded by two separated codon sets that differ at first or first plus second codon positions. Our detailed analyses clearly indicate that these unusual preferences shared by haptophyte and some peridinin-type plastid genes are largely responsible for their strong affinity in DNA analyses. In particular, almost all of the support from DNA level analyses for the monophyly of haptophyte and peridinin-type plastids is lost when the codons corresponding to constant Leu, Ser, and Arg amino acids are excluded, suggesting that this signal comes from rapidly evolving synonymous substitutions, rather than from substitutions that result in amino acid changes. Indeed, protein maximum-likelihood analyses of concatenated PsaA and PsbA amino acid sequences indicate that, although 19' hexanoyloxyfucoxanthin-type (19' HNOF-type) plastids in dinoflagellates group with haptophyte plastids, peridinin-type plastids group weakly with those of stramenopiles. Consequently our results cast doubt on the single origin of peridinin-type and 19' HNOF-type plastids in dinoflagellates previously suggested on the basis of psaA and psbA concatenated gene phylogenetic analyses. We suggest that codon usage heterogeneity could be a more general problem for DNA level analyses of protein-coding genes, even when third codon positions are excluded.
Keywords: Chromalveolate hypothesis; codon usage bias; concatenated gene phylogeny; dinoflagellates; long-branch attraction; model misspecification; phylogenetic artifact; plastid evolution
Received July 2, 2003; Revised November 3, 2003; Accepted March 5, 2004
DNA sequences of protein-coding genes are often used for phylogenetic analyses for several reasons. First, when a protein data set is suspected not to encode enough phylogenetic signal, the corresponding DNA sequences may provide extra information in the form of synonymous codon changes (silent substitutions). Second, on a more practical level, if one wishes to consider both protein-coding genes and RNA genes in a single concatenated analysis using one kind of substitution model, the former must be analyzed at the DNA sequence level (e.g., Yoon et al., 2002b). Finally, phylogenetic analyses on DNA sequences (DNA analysis) may be less computationally expensive than analyses of protein sequences because the substitution matrices are much simpler, entailing far fewer calculations, especially if maximum-likelihood (ML) methods are employed (e.g., Baldauf et al., 2000).
In protein-coding genes, the substitution patterns at first, second, and third codon positions are expected to differ from one another due to the degeneracy of the genetic code. Most substitutions at first or second codon positions change the amino acid encoded (i.e., are nonsynonymous), whereas most substitutions at third codon positions do not result in changes at the amino acid level (i.e., are synonymous). Sites where synonymous substitutions are possible have higher evolutionary rates than those constrained by selection at the amino acid level (Graur and Li, 2000). Thus, although such positions can potentially yield additional phylogenetic information absent in amino acid sequences, they rapidly saturate with multiple substitutions. Consequently they also evolve rapidly to reflect base composition and/or codon usage preferences characteristic of their host genome (Graur and Li, 2000). The rapidly evolving nature of third codon positions, combined with drastic changes in base composition and codon usage preferences over the tree of life, potentially creates strongly misleading signals in the data, if the nucleotide or codon models used assume stationarity of the substitution process (Yang and Roberts, 1995; Galtier and Gouy, 1998; Foster and Hickey, 1999). Because of these difficulties, third codon positions are often omitted from DNA-level analyses, especially for comparisons where substitution saturation is expected (e.g., analyses assessing ancient divergences, see Baldauf et al., 2000). Since changes in first and second codon positions are mostly nonsynonymous, analyses of these positions (DNA1 + 2 analysis) are thought to be less susceptible to such problems and are expected to give results that are consistent with corresponding protein analyses.
However, not all substitutions at first or second codon positions are nonsynonymous. Leucine (Leu), serine (Ser), and arginine (Arg) are each encoded by six different codons that fall into two different codon sets—TTR and CTN for Leu (R for A or G; N for A, C, G, or T), TCN and AGY for Ser (Y for T or C), and CGN and AGR for Arg. Thus, T
C and CA substitutions are synonymous at the first positions of Leu and Arg codons, respectively. In Ser codons, TA substitutions at first positions and C G substitutions at second positions can be synonymous in combination. In principle, therefore, it is possible that for the three amino acids with six codons the base compositional and codon usage heterogeneity evident at third codon positions could also affect first and second positions. In the worst-case scenario, distantly related sequences that exclusively share idiosyncratic codon usage patterns for Leu, Ser, and/or Arg could be grouped artifactually in DNA1 + 2 analyses. Here we present evidence for precisely this kind of artifact in phylogenetic analyses of the psbA genes from photosynthetic eukaryotes.
Extant plastids (or chloroplasts) are the direct or indirect descendants of a single endosymbiotic cyanobacterium (see review in Palmer, 2003). A common ancestor of red algae, green algae (plus land plants), and glaucophytes probably acquired plastids via one cyanobacterium-eukaryote endosymbiosis (‘primary endosymbiosis’), although several photosynthetic eukaryotic lineages most probably captured their plastids through multiple ‘secondary’ endosymbioses of photosynthetic eukaryotes (reviewed in Palmer, 2003). The plastids in haptophytes, cryptophytes, stramenopiles, dinoflagellates, and apicomplexans (this protist group is nonphotosynthetic but retains remnant plastids; see McFadden and Waller, 1997) are now widely thought to have originated from single event of secondary endosymbiosis involving a red alga (Fig. 1A; see Cavalier-Smith, 1999; Fast et al., 2001; Harper and Keeling, 2003).
|
Most photosynthetic dinoflagellates have a ‘peridinin-type plastid’ that is surrounded by three membranes and contains a unique accessory pigment called peridinin. Molecular phylogenetic studies indicate that the peridinin-type plastid is of red algal origin (Takishita and Uchida, 1999; Zhang et al., 1999, 2000; Fast et al., 2001). However, several lineages of dinoflagellates have plastids with four bounding membranes and different complements of chlorophylls and accessory pigments (Delwiche, 1999). These rarer nonstandard plastids were thought to derive from ‘plastid replacement’ events where the ancestral peridinin-type plastid is supplanted by plastids acquired from different types of eukaryotic algae (Fig. 1A; Delwiche, 1999). Most notably, Gymnodinium mikimotoi and its relatives contain plastids with chlorophylls a + c and 19' hexanoyloxyfucoxanthin (‘19' HNOF-type plastids’), a combination otherwise characteristic of haptophytes (Fig. 1A; Delwiche, 1999). In the case of the 19' HNOF-type plastids, the donor lineage was most probably an engulfed haptophyte (Takishita et al., 1999, 2000; Tengs et al., 2000). As haptophytes themselves obtained their plastid through secondary endosymbiosis, the consortium formed from a dinoflagellate host and an haptophyte plastid would be a ‘tertiary endosymbiosis’ (Fig. 1A; Delwiche, 1999).
However, recent concatenated phylogenetic analyses of two plastid-encoded genes, psbA and psaA (encoding photosystem II D1 protein and photosystem I P700 chlorophyll a apoprotein A1, respectively) suggest a radically different scenario (Yoon et al., 2002a). In phylogenetic trees using both DNA and protein sequences, 19' HNOF-type plastids formed a robust clade with peridinin-type plastids, rather than being a direct sister to a haptophyte clade as would be expected under the standard plastid replacement scenario schematically shown in Figure 1A. As a result, the dinoflagellate clade as a whole clusters with haptophyte plastids with high bootstrap support (BP) values. Based on this phylogenetic pattern, Yoon et al. (2002a) proposed that the 19' HNOF-type plastid was actually ancestral to extant dinoflagellates, with the peridinin-type plastid evolving directly from a 19' HNOF-type plastid by descent-with-modification (Fig. 1B), rather than by a separate endosymbiotic event shown in Figure 1A.
Here we analyze a psbA gene data set including both 19' HNOF- and peridinin-type plastids with DNA and protein ML methods. The monophyletic clade of haptophyte and dinoflagellate plastids (‘H + D plastid clade’) receives high bootstrap support in DNA analyses, whereas the same topology is weakly supported by protein analyses. The strong affinity between haptophyte and peridinin-type plastids detected in DNA analyses was assessed using ML bootstrap analyses and comparisons between tree topologies based on log-likelihood differences and ‘approximately unbiased’ tests. Our results clearly suggest that the robust affinity between haptophyte and peridinin-type plastids is largely attributable to the shared pattern of Leu, Ser, and Arg codon usage heterogeneity in haptophyte plastids and a subset of the peridinin-type plastids (Amphidinium spp. and Heterocapsa spp.). We further reanalyzed the concatenated psaA plus psbA data sets used by Yoon et al. (2002a) to examine whether the phylogenetic analyses in the original study were affected by artifacts stemming from the marked codon usage heterogeneity and the extremely high evolutionary rates of dinoflagellate sequences.
|
Methods |
|---|
|
TopNotesAbstractMethodsResults and DiscussionConclusionsAppendixAcknowledgmentsReferences |
|---|
Data Sets
Forty-nine psbA genes were selected from Cyanophora paradoxa (a glaucophyte) and red alga-derived plastids (i.e., red algae, chromists, and dinoflagellates; see Fig. 1A and B). Their putative protein sequences were aligned manually, and then the corresponding DNA alignment was generated. All codons containing gaps and missing data were omitted, except where the missing data were due to the partial dinoflagellate sequences from Karenia brevis, Scrippsiella trochoidea, and Thoracosphaera heimii (447, 135, and 129 nucleotide positions missing from the N-terminus of the trimmed alignment, respectively). In all, 885 nucleotide positions (295 codons) were retained for analysis. The codon usage patterns for Leu, Ser, and Arg were examined for all sequences in this alignment (see below). Based on the results of Yoon et al. (2002a), we removed closely related sequences to create alignments more tractable for phylogenetic analysis containing 23 sequences with 885 nucleotide positions (DNA-23 data set) or 295 amino acid positions (protein-23 data set). Our preliminary bootstrap analyses (ML for DNA data and ML distance for protein data) suggest that the different taxonomic sampling in the data sets (49 versus 23) has no significant impact on our phylogenetic inferences (data not shown).
Phylogenetic Analyses
Simulation studies indicate that ML methods are more robust to ‘long branch attraction (LBA)’ (whereby highly diverged sequences are artifactually attracted to one another) than maximum parsimony or distance methods (e.g., Swofford et al. 2001). Because dinoflagellate psbA sequences are extremely divergent, ML methods are the most appropriate to reduce or avoid potential LBA in the psbA data sets. Another potential difficulty in the DNA analyses is base composition bias—the psbA sequences considered here bear strong G+C composition bias, particularly at third codon positions (values ranging from 16% to 46%). Thus, the third codon positions of the psbA data sets were excluded from DNA ML analyses. However, ML methods are also sensitive to LBA if the evolutionary model is misspecified (e.g., Sullivan and Swofford, 2001). Therefore, to better account for complexities in the substitution process in the problematic psbA data sets, one of the most complex models available for nucleotide evolution, the general-time-reversible (GTR) model incorporating among-site rate variation (ASRV) was used throughout this study.
Using Paup* v.4.0b10 (Swofford, 1998), the ML tree was reconstructed from the first and second codon positions in the DNA-23 data set under the GTR model with ASRV modeled using a discrete gamma (
) distribution with four equally probable rate categories plus invariable sites (GTR++I model). For simplicity, Jukes-Cantor–corrected neighbor-joining trees were used for the estimation of model parameters, as it is known that slight errors in initial trees seldom adversely affect the accuracy of parameter estimation, especially for the ASRV parameters (e.g., Sullivan et al., 1996). Ten random taxon additions and tree bisection reconnection (TBR) topological rearrangements were used to heuristically search for the ML tree. A bootstrap analysis (100 replicates) was performed using the same model, but with one random sequence addition per replicate and TBR. The same analysis was repeated considering only second codon positions to exclude the potential artifact from the first positions of Arg, Leu, and Ser codons.
For protein analyses, ML methods incorporating ASRV were selected to reduce or avoid LBA. The protein-23 data set was subjected to the ML analyses using ProML implemented in Phylip v.3.6a (Felsenstein, 1993). The ML tree was estimated with the JTT amino acid substitution matrix, with ASRV modeled using a discrete approximation to a distribution (JTT+ model), with five random taxon additions and global rearrangements for tree searching. ML bootstrap analyses using the JTT+ model (100 replicates) were performed with global rearrangements after a single stepwise addition sequence per replicate. Conditional mode site rates (eight and four equally probable categories for tree search and bootstrap analysis, respectively) were estimated from the data using Tree-Puzzle v. 5.0 (Schmidt et al., 2002).
Separate Evaluations of Three Groups of Dinoflagellate Plastids
Peridinin-type plastids exhibit an unusually wide range in their usage of different Leu, Ser, and Arg codons (see Fig. 3). Based on a discontinuous distribution in the usage of Ser codons, we divided peridinin-type plastids into two groups: ‘group-1’ including Amphidinium spp. and Heterocapsa spp., and ‘group-2’ including Alexandrium, Gonyaulax, Prorocentrum, Scrippsiella and Thoracosphaera. The two groups also had nonoverlapping distributions of Leu and Arg codon usage, and, significantly, the Leu, Ser, and Arg codon usage of group-1 peridinin-type plastids is strikingly similar to that of haptophytes (see below).
|
If peridinin-type plastids arose from 19' HNOF-type plastids during dinoflagellate evolution (Fig. 1B), haptophyte and peridinin-type plastids would be expected to form a clade in DNA analyses in the absence of 19' HNOF-type plastids. Likewise, if distinct codon usage patterns for Leu, Ser, and Arg detected among peridinin-type plastids do not cause phylogenetic artifacts in DNA analyses, the clade consisting of haptophyte and dinoflagellate plastids should be recovered regardless of the selection of peridinin-type plastids. We tested these conjectures using (I) bootstrap analyses based on the first and second codon positions, and on the second codon positions only (DNA1 + 2 and DNA2 bootstrap analyses), (II) ‘tree log-likelihood (lnL) comparisons,’ and (III) ‘approximately unbiased (AU)’ tests.
Bootstrap analyses
New data sets with four different samples of dinoflagellate plastid sequences were generated from the DNA-23 data sets. These sub-data sets include 16 nondinoflagellate taxa plus either (I) five peridinin-type dinoflagellates; (II) Amphidinium operculatum and Heterocapsa triquetra (peridinin-type plastids with ‘group-1’ codon usage); (III) Alexandrium tamarense, Gonyaulax polyedra, and Prorocentrum micans (peridinin-type plastids with ‘group-2’ codon usage); or (IV) two 19' HNOF-type dinoflagellates. DNA1 + 2 and DNA2 bootstrap analyses were conducted on these data sets as described above, except that 500 replicates were used in the analysis on the data set including five peridinin-type dinoflagellates.
Tree lnL comparisons
Another smaller data set was generated from the DNA-23 data set by removing all dinoflagellates and two redundant haptophytes Pavlova lutheri and Emiliania huxleyi (‘DNA-14 data set’). Again, preliminary bootstrap analyses confirmed that the removal of the two haptophytes has no significant impact on overall phylogenetic inferences (data not shown). This DNA-14 data set was used to create a backbone topology (Fig. 4) by subjecting the first and second codon positions to ML analysis with a GTR+ +I model as described above. Each of the three subgroups of dinoflagellate plastids described above (group-1 peridinin-type, group-2 peridinin-type [Gonyaulax is excluded from this analysis], and 19' HNOF-type; Fig. 4) was grafted as a clade to all edges of this backbone in turn, creating three sets of 25 tree topologies (Fig. 4). The lnLs of the 25 tree topologies were calculated from the first and second codon positions in the DNA data sets (DNA1 + 2 data sets) under the GTR++I model by Paup* (Swofford, 1998). The lnL of topology 1, which corresponds to the ML tree from the DNA1 + 2 data set, was subtracted from those of the alternatives to examine the significance of the H + D plastid clade—lnL differences reflect the degree of dominance of topology 1 over the alternatives. Similar analyses were performed on the corresponding protein data sets using the JTT+ model in Tree-Puzzle (Schmidt et al., 2002). The parameters for the ML analyses and the branch lengths of each tree topology were estimated from the data.
|
AU tests
For each of 25 trees described above, lnLs at sites (site lnLs) were calculated under the same models using Paup* for DNA data, and Codeml implemented in Paml v.3.1 (Yang, 1997) for protein data. The Paup* output files containing site lnLs were reformatted using covMONKEY (C. Blouin, Dalhousie University, Halifax, Canada). Using these site lnL data, AU tests were conducted using Consel v.0.1f with default parameter settings (Shimodaira and Hasegawa, 2001; Shimodaira, 2002).
Analysis of a psbA Data Set With Corrected Arg, Leu, and Ser Codon Usage
The impact of codon usage heterogeneity in Arg, Leu, and/or Ser codons on a relationship between the haptophyte and peridinin-type plastids was also evaluated by a DNA data set including group-1 and -2 peridinin-type dinoflagellates modifying Arg, Leu, and Ser codon usage patterns. Among 49 psbA sequences, CGN codons appear to be preferred over synonymous AGR codons for Arg. Likewise, TTR are preferred over CTN for Leu codons, and TCN are preferred over AGY for Ser codons (data not shown). All Arg, Leu, and Ser codons in the data set were recoded into major CGN, TTR, and TCN, respectively, in order to prevent potential phylogenetic artifacts from the codon usage heterogeneity. This modified data set was subjected to DNA1 + 2 ML bootstrap analysis (500 replicates). The details of the bootstrap analysis are same as described above.
Analyses Excluding ‘Constant L/S/R Codons’
The protein data sets including 14 nondinoflagellate taxa (see the previous section) and either group-1 or group-2 peridinin-type dinoflagellates contain 34 to 35 sites that are fixed to be Leu, Ser, or Arg. We excluded the codons for these positions (‘constant L/S/R codons’) from the corresponding DNA alignments to generate ‘DNA* data sets.’ First and second codon positions from the DNA* data sets were subjected to ML bootstrap analyses (500 replicates), tree lnL comparisons and AU tests as described above. To evaluate the significance of the exclusion of the constant L/S/R codons, we also generated 100 randomly jackknifed data sets using Jacksite (H. Philippe, Université de Montreal, Canada), removing the same number of codons as there are constant L/S/R codons. Then, first and second positions of the jackknifed data sets (‘DNA*1 + 2 data sets’) were used to calculate the lnLs under the 25 tree topologies as described above. A single jackknifed data set was used to estimate the model parameters for the entire analysis.
Analyses of Concatenated Gene Data Sets
Confirmation of the Results Presented in Yoon et al. (2002a)
D. Bhattacharya (University of Iowa, Iowa City, IA) generously provided us with the DNA and protein data sets for concatenated psaA and psbA genes used in their original work (these data sets are described in detail in Yoon et al., 2002a). To confirm the results presented in Yoon et al. (2002a), we first conducted bootstrap analyses (500 replicates) on DNA sequences using the minimum-evolution (ME) method with LogDet/Paralinear distances (‘LogDet distance analysis’) in Paup* (Swofford, 1998). All codon positions were used for this DNA analysis (‘DNAALL analysis’), as per Yoon et al. (2002a). It should be noted that these authors did not account for invariable sites in their LogDet analyses and this failure to account for ASRV may seriously bias phylogenetic estimation (Sullivan and Swofford 2001). Therefore we also performed LogDet analyses allowing for invariable sites (LogDet+I), a measure that should address the problem of ASRV to some degree (Sullivan and Swofford, 2001).
Bootstrap analysis (500 replicates) on the protein concatenated data set was conducted using the ME method with Poisson-corrected distances in Mega2 (Kumar et al., 2001). The settings of these analyses were the same as those described in Yoon et al. (2002a).
Reanalyses of the concatenated DNA and protein data sets using ML methods
The same concatenated gene data sets were then subjected to ML analyses using the GTR++I model (for the DNA sequences) and the JTT+ model (for the protein sequences). Parameters for these models were estimated from the concatenated data, and applied to the entire alignment. The DNA ML analyses were the same as described above, except all codon positions were considered, as per Yoon et al. (2002a). For the protein ML analyses, four equally probable rate categories were used and for bootstrap analysis, heuristic tree searching employed only local rearrangements. Other details of the protein ML analyses are the same as described above.
Protein sequence-based AU tests on the concatenated protein data set
We generated a PsaA protein data set including Amphidinium operculatum and Heterocapsa triquetra (group-1 peridinin-types), which is equivalent to the PsbA data set used for tree lnL comparisons (see Fig. 4 for the taxa considered). Twenty-five tree topologies (see Fig. 4) were assessed by the AU test based on a data set combining PsaA and PsbA sequences. Site lnLs of the PsaA and PsbA portions of the concatenated data were calculated under independent parameters for the JTT+ model and branch lengths estimated from each gene data set. All other aspects of the AU tests are the same as described above.
|
Results and Discussion |
|---|
|
TopNotesAbstractMethodsResults and DiscussionConclusionsAppendixAcknowledgmentsReferences |
|---|
Incongruities Between DNA and Protein Analyses on PsbA Sequences
The ML trees reconstructed from the DNA-23 and protein-23 data sets are broadly similar to each other, recovering haptophyte and dinoflagellate plastids as a clade (‘H + D plastid clade’) to the exclusion of all other plastids (Fig. 2A and B). However, the BP values for the H + D plastid clade differ markedly between analyses. The DNA1 + 2 bootstrap analysis suggests that the H + D plastid clade is robust (BP = 82%; Fig. 2A), consistent with the results presented by Yoon, et al. (2002a). Importantly, similar support values are recovered even if the 19' HNOF-type plastids are excluded (Table 1), indicating that the affinity between haptophyte plastids and peridinin-type plastids is not dependent on attraction to 19' HNOF-type plastid sequences. On the other hand, the clade is very weakly supported by the protein bootstrap analysis (BP = 29%; Fig. 2B) and by DNA2 bootstrap analyses (BP < 5%; Table 1).
|
|
The one important topological incongruity between the analyses concerns the internal relationships amongst the peridinin-type plastids. A strong subgrouping of Alexandrium, Prorocentrum, and Gonyaulax is recovered by the DNA1 + 2 analyses (BP = 94%; Fig. 2A). This subgroup was also recovered with high BP values in the phylogenies of Yoon et al. (2002a). However this clade received weaker BP supports in the protein and DNA2 analyses (BP = 29% and 68%, respectively).
Shared Leu, Ser, and Arg Codon Usage Patterns Across the Haptophyte and Peridinin-Type Plastids
None of the protein sequences in protein-23 data set were rejected by a 
2 test of amino acid composition in Tree-Puzzle (Schmidt et al., 2002). Thus we suspected that the significant discrepancies between the protein and DNA2 analyses on one hand, and the DNA1 + 2 analyses on the other, might be caused by codon usage heterogeneity at Leu, Ser, and/or Arg positions. The psbA genes of Amphidinium spp. and Heterocapsa spp. (hereafter ‘group-1 peridinin-type’) exhibit a strong preference for CTN codons over TTR codons for Leu, TCN over AGY for Ser, and CGN over AGR for Arg (Fig. 3A to C). On the other hand, the rest of peridinin-type dinoflagellates included in this analyses, Alexandrium spp., Gonyaulax, Prorocentrum, Scrippsiella, and Thoracosphaera (hereafter ‘group-2 peridinin-type’) have no strong preference for any of these codon boxes (Fig. 3A to C). In particular, a clear separation of the two groups is observed in Ser codon usage (Fig. 3B). Significantly, the codon usage patterns of group-1 peridinin-type plastids and of haptophyte plastids are similar to one another (Fig. 3A to C), and in the cases of Leu and Ser, are markedly different from those of red algae, cryptophytes, and stramenopiles (Figs. 3A and B). By contrast, the patterns for Leu and Ser in psbA from group-2 peridinin-type plastids are similar to those from red algae, cryptophytes and stramenopiles (Fig. 3A and B).
Based on the results presented above, the robust clade of group-2 peridinin-type dinoflagellates (Alexandrium, Prorocentrum, and Gonyaulax) in the DNA1 + 2 analyses (Fig. 2A) may reflect the differences in the codon usage patterns for Leu, Ser, and Arg amongst the peridinin-type plastids, rather than a true historical relationship. However, it remains unclear whether these patterns contribute to the support for the H + D plastid clade in the DNA1 + 2 analyses. A potential complicating factor is the fact that all dinoflagellate plastid sequences have been evolving more rapidly than homologous sequences from other algal lineages (Fig. 2A and B). Furthermore, the Leu, Ser, and Arg codon usage patterns (Figs. 3A to C) suggest that the process of nucleotide substitutions at the first and second codon positions is nonstationary, violating the assumption of the GTR++I model. Therefore, the highly diverged peridinin- and 19' HNOF-type plastids, which may or may not have a direct sister relationship excluding other taxa considered, could artifactually attract one another in our ML analyses. Consequently, in the following analyses, the two types of dinoflagellate plastids were examined separately to avoid potential LBA between these two lineages. The two groups of peridinin-type plastids with different codon usage patterns for Leu, Ser, and Arg were further examined separately. If the robust H + D plastid clades observed (Table 1) were indeed due to the artifact from the shared codon usage patterns across the haptophyte and group-1 peridinin-type sequences (Fig. 3A to C), the significance may be abolished when group-1 peridinin-type sequences are excluded from DNA analyses.
19' HNOF-Type Plastids
In the tree lnL comparisons including 19' HNOF-type plastids, topology 1, in which the 19' HNOF-type plastids fall within haptophyte plastids as the sister of Isochrysis sp. (Fig. 4), dominates over the alternatives (filled squares in Figs. 5A and B). In particular, the likelihoods of topologies 4 to 25, which do not include the H + D plastid clade (Fig. 4), are at least 20 lnL units in the protein analysis and 50 lnL units in the DNA1 + 2 analysis lower than topology 1 (Fig. 5A and B). The AU tests and bootstrap analyses complement the results of the tree lnL comparisons. The monophyly of haptophyte and 19' HNOF-type plastids is supported by bootstrap analyses of both DNA1 + 2 and DNA2 data sets (BP = 100% and 74%; Table 1). Topologies 4 to 25 are all rejected by the DNA sequence-based AU test (the sixth column in Table 2). The protein sequence-based test rejects all topologies except 1 (the third column in Table 2). The results clearly support the evolutionary affinity between the 19' HNOF-type plastids from dinoflagellates and those from haptophytes, consistent with (I) the fact that these plastids share an unique accessory pigment, 19' HNOF, and (II) the results from other phylogenetic studies (Takishita et al., 1999, 2000; Tengs et al., 2000). As no hint of phylogenetic artifact for the connection between haptophyte and 19' HNOF-type plastids was detected, this matter was not pursued further.
|
|
Peridinin-Type Plastids
More equivocal results are obtained from the protein data sets using either group of peridinin-type plastids from dinoflagellates (open and filled circles in Fig. 5A). In these comparisons, topology 1 was still the most likely, but all topologies tested are within 20 lnL units of each other (Fig. 5A). In each case, protein sequence-based AU tests fail to reject several topologies that have no H + D plastid clade (the first and second columns in Table 2). In particular, topologies 4 to 10, 15, and 23 are never rejected (see Fig. 4 for the details of these topologies).
Despite the weak support for the monophyly of haptophyte and group-1 peridinin-type plastids in the protein analyses, the equivalent DNA analyses provide strong support for the H + D plastid clade. The differences in lnLs (
lnLs) between the topologies with and without the H + D plastid clade appear to be >20 lnL units (filled circles in Fig. 5B). A DNA analysis with a codon model using Codeml implemented in Paml (Yang, 1997) showed a similar result (data not shown). The AU test rejects topologies 4 to 25, as well as topology 3 (the fourth column in Table 2), in which the group-1 peridinin-type plastid sequences are the sister group to a monophyletic haptophyte plastid clade (Fig. 4). Furthermore the DNA1 + 2 bootstrap analysis strongly supports the H + D plastid clade (Table 1). By contrast, the DNA2 bootstrap analysis, like the protein analysis, provides weak support for this clade (BP = 40%; Table 1).
The results from the DNA analyses considering group-2 peridinin-type plastids are more or less consistent with the corresponding protein analyses. In the tree lnL comparison, the lnL between topologies 1 to 3 and the alternatives, are generally <20 lnL units (open circles in Fig. 5B) and the DNA sequence-based AU test fails to reject topologies 4–9, in addition to topologies 2 and 3 (the fifth column in Table 2). The monophyly of the haptophyte and group-2 peridinin-type plastids is poorly supported either by DNA1 + 2 or DNA2 bootstrap analysis (BP = 55% or 23%; Table 1). We obtained a similar result from an analysis employing a codon model (data not shown).
The examinations of group-1 peridinin-type plastids circumstantially, but strongly, suggest that the robustness of the H + D plastid clade in our DNA1 + 2 analyses is attributable to codon usage preferences for Leu, Ser, and Arg shared uniquely by the two lineages. On the other hand, the DNA analyses examining group-2 peridinin-type plastids are not affected by these codons, because none of codon usage patterns for the three amino acids in this plastid lineage matches those in haptophyte plastids (Fig. 3A to C).
The Impact of Constant L/S/R Codons on psbA Phylogeny
Given the relatively conserved nature of PsbA proteins, much of the supposed artifactual signal may come from constant amino acids fixed to Leu, Ser, or Arg (codons corresponding to constant Leu, Ser, or Arg amino acids will henceforth be referred to as "constant L/S/R codons"). Thus, to evaluate directly the impact of these codon usage patterns to the psbA DNA phylogeny, we reanalyzed the DNA1 + 2 data sets in the absence of constant L/S/R codons (‘DNA*1 + 2 data sets’).
As expected, exclusion of constant L/S/R codons drastically reduces the support for the H + D plastid clade in the DNA1 + 2 analyses including group-1 peridinin-type plastids. In the tree lnL comparison based on the DNA*1 + 2 data set, <20 lnL units separate topology 1 and topologies 4 to 25 (filled diamonds in Fig. 5C). This result is similar to that from the comparison based on the corresponding protein data set (filled circles in Fig. 5A). The observed lnLs for the topologies 4 to 25 are significantly smaller than the null distribution from 100 randomly jackknifed data sets (P < 0.01; Fig. 5C), indicating that the lower likelihood ratio cannot be attributed solely to the reduction in data. After the exclusion of constant L/S/R codons, the AU test fails to reject topologies 1 to 11 (the seventh column in Table 2), contrasting sharply with the test on the original data that rejects all topologies except 1 and 2. The bootstrap support for the H + D plastid clade from the DNA*1 + 2 analysis was also reduced to 57% (Table 1).
Also as anticipated, the removal of constant L/S/R codons has little impact on analyses considering the group-2 peridinin-type plastids. In comparisons based on the DNA*1 + 2 data set, the observed lnLs are similar to those calculated from the data set including constant L/S/R codons (filled diamonds in Fig. 5D), and fall within the null distributions (Fig. 5D). The results of the AU tests are little changed by the exclusion of the constant L/S/R codons, with topologies 4 to 9 not rejected (the eighth column in Table 2). The DNA*1 + 2 bootstrap analyses provide very weak support for the H + D plastid clade (BP = 27%; Table 1).
To further test the effect of the codon usage heterogeneity, we recoded all Arg, Leu, and Ser codons into CGN, TTR, and TCN codons, respectively (for these analyses third codon positions were excluded, so the identity of this position is irrelevant) in a data set containing the group-1 and -2 peridinin-type dinoflagellate plastids and this data set was then subjected to ML bootstrap analysis. In this analysis, the bootstrap support for the H + D plastid clade was diminished to 38% (Table 1), in contrast to the corresponding value from the analysis without the recoding (BP = 89%; Table 1). These results indicate that the high support for the H + D plastid clade is largely due to the heterogeneity in Arg, Leu, and Ser codon usage.
We also returned to the original DNA-23 data set (containing group-1 and -2 peridinin-type and 19' HNOF-type dinoflagellate plastids) and removed the constant L/S/R codons. With all dinoflagellate sequences considered, the BP value for the H + D plastid clade remains high (92%; Table 1). However, when the 19' HNOF-type plastids are excluded, the BP value drops markedly (57%; Table 1), in contrast to our initial analysis, where exclusion of the 19' HNOF-type plastids did not weaken support for this clade. Clearly factors other than codon usage patterns also affect the topology recovered here—classical LBA between 19' HNOF-type and peridinin-type plastids in our data set is a likely possibility.
Concatenated Gene and Protein ML Analyses of PsbA and PsaA Do Not Resolve the Origin of Peridinin-Type Plastids
Our results presented above strongly suggest that a major signal uniting haptophyte and peridinin-type plastids in DNA1 + 2 analysis comes from synonymous changes at Leu, Ser, and Arg codon positions, rather than nucleotide positions under selection at the amino acid level. Therefore, although the ML trees from both protein and DNA1 + 2 analyses have a H + D plastid clade (Fig. 2A and B), the robustness of this clade in DNA analyses is highly suspect. Interestingly, although not presented here, our analyses indicate that the phylogenetic artifact caused by constant L/S/R codons also occurs in a psaA DNA data set including group-1 peridinin-type plastids. It is also probable that a similar artifact affected the DNA analyses of Yoon et al. (2002a), because only group-1 peridinin-type plastids (i.e., Amphidinium operculatum and Heterocapsa triquetra) were considered in their concatenated gene phylogenetic analyses. Here we reanalyzed the concatenated gene and protein data sets used in Yoon et al. (2002) using DNA and protein ML methods.
Concatenated DNA analyses
The DNAALL analysis using the ML method under the GTR++I model on the concatenated gene data set reconstructed both a clade consisting of peridinin-type and 19' HNOF-type plastids and the monophyly of dinoflagellate plus haptophyte plastids (Fig. 6A) in accordance with the results of Yoon et al. (2002a). However, these clades receive significantly different BP values in the ML versus LogDet distance analyses (Fig. 6A). Although LogDet analyses show strong support for these groupings, ML analyses with the complex GTR++I model received much lower support in bootstrap analysis. Interestingly, accounting for invariable sites in the LogDet analyses (LogDet+I), lowered the BP values for the groups above in much the same way as the ML analyses (Fig. 6A), with the exception that a monophyletic dinoflagellate plastid clade (peridinin-type plus 19' HNOF-type plastids) is not recovered at all in the optimal tree (BP < 5%). Several aspects of model misspecification seem to be at work here. First the original LogDet analyses are obviously influenced by the failure to model ASRV—the high BP values for groupings identified above are at least in part an artifact stemming from this model misspecification, and decay somewhat when it is taken into account. However, the assumptions of the GTR++I model used in the ML analyses are also violated due to the heterogeneous base composition in the data caused by the codon usage heterogeneity. Nevertheless, Rosenberg and Kumar (2003) recently showed that, when the assumption of homogeneous base composition is violated, ML methods perform better than MP and distance methods, including the LogDet method that was designed to be robust to heterogeneous base composition. In summary, it seems that the ML analyses and LogDet+I analyses on the DNAALL data set recover some of the same groups as the original LogDet analysis, but with much reduced bootstrap support, casting doubt on their validity.
|
Concatenated protein analyses
Poisson-corrected distance analyses of the concatenated PsaA and PsbA sequences strongly supported the monophyly of 19' HNOF-type plus peridinin-type plastids in dinoflagellates, as well as a H + D plastid clade (BP = 100% and 96%, respectively; Fig. 6B), as seen in Yoon et al. (2002a). However, the extremely long branch lengths of dinoflagellate plastids (Fig. 6C) suggest that the high support values obtained from these distance analyses could result from LBA due to the use of the overly simple Poisson model of protein evolution and ignoring ASRV. Indeed, all methods of analysis including ML are known to be susceptible to LBA if ASRV is not taken into account (Sullivan and Swofford, 2001). Consistent with this, bootstrap support for the monophyly of the two types of dinoflagellate plastids and the H + D plastid clade are negligible in distance analyses under the JTT+
model (BP < 5%; not shown). In contrast to all previous analyses, our protein ML analyses weakly group peridinin-type plastids with those of stramenopiles, whereas 19' HNOF-type plastids remained within a haptophyte clade (Fig. 6C). This branching pattern is consistent with the common origin of peridinin-type plastids with other ‘chromalveolate’ plastids as proposed previously on other evidence (Fig. 1A; Cavalier-Smith, 1999; Fast et al., 2001; Harper and Keeling, 2003). Unfortunately most of the relevant branches are poorly supported in ML bootstrap analyses (Fig. 6C). Likewise, protein sequence-based AU tests based on the combined PsaA plus PsbA data set fail to reject many of the alternative hypotheses for the origin of peridinin-type plastids (see the right-most column in Table 2).
The low phylogenetic resolution observed in our ML analyses (Fig. 6A to C and Table 2) suggests that further multigene studies, preferably using analyses where the complications stemming from the codon usage heterogeneity described above are avoided, will be required to settle this intriguing issue in plastid evolution. Until such data are forthcoming, it seems premature to assume that the peridinin-type plastid is directly related to and descended from the 19' HNOF-type plastid now found in a few dinoflagellates.
|
Conclusions |
|---|
|
TopNotesAbstractMethodsResults and DiscussionConclusionsAppendixAcknowledgmentsReferences |
|---|
Unusual codon usage patterns for Leu, Ser, and Arg may be common in data sets considered for phylogenetic analyses. In theory, consideration of all three codon positions at the DNA level should yield more phylogenetic information than the corresponding amino acid sequences of proteins. However, it would be prudent to examine codon usage for marked heterogeneity in a data set before DNA-level analyses. If third codon positions are to be excluded because of their rapid evolutionary rate, it may be considered sensible to also exclude some or all codon positions where Leu, Ser, or Arg codons predominate (e.g., constant L/S/R codons).
|
Appendix |
|---|
|
TopNotesAbstractMethodsResults and DiscussionConclusionsAppendixAcknowledgmentsReferences |
|---|
Accession numbers for the sequences used in this study.
Cyanophora paradoxa (NC_001675
[GenBank]
/NC_001675), Bangia atropurpurea (AY119734
[GenBank]
), Bangia fuscopurpurea(AY119735
[GenBank]
), Bangiopsis subsimplex (AY119736
[GenBank]
), Chondrus crispus (AY119746
[GenBank]
), Compsopogon coeruleus (AY119737
[GenBank]
), (NC_001840
[GenBank]
/NC_001840), Cyanidioschyzon merolae (AY119730
[GenBank]
), Dixonielloa grisea (AY119738
[GenBank]
/AY119702), Erythrotrichia carnea (AY119739
[GenBank]
/AY119703), Galdieria sulphuraria strain DBV063 (AY119733
[GenBank]
), Galdieria sulphuraria strain SAG108.79 (AY119731
[GenBank]
), Porphyra purpurea (NC_000925
[GenBank]
/NC_000925), Porphyridium aerugineum (AY119741
[GenBank]
/AY119705), Rhodosorus marinus (AY119744
[GenBank]
), Rhodella violacea (AY119742
[GenBank]
/AY119706), Stylonema alsidii (AY119745
[GenBank]
), Thorea violacea (AY119747
[GenBank]
), Emiliania huxleyi (AY119752
[GenBank]
), Isochrysis sp. SAG927-2 (AY119753
[GenBank]
/AY119717), Pavlova gyrans (AY119754
[GenBank]
/AY119718), Pavlova lutheri (AY119755
[GenBank]
), Phaeocystis antarctica (AY119756
[GenBank]
), Pleurochrysis carterae (AY119757
[GenBank]
), Prymnesium parvum (AY119758
[GenBank]
) Chroomonas sp. SAG980-1 (AY119749
[GenBank]
), Guillardia theta (NC_000926
[GenBank]
/NC_000926), Pyrenomonas helgolandii (AY119750
[GenBank]
), Rhodomonas abbreviata (AY119751
[GenBank]
/AY119715), endosymbiotic diatom in Peridinium foliaceum (AY119764
[GenBank]
), Heterosigma carterae (HCU18090), Odontella sinensis (NC_001713
[GenBank]
/NC_001713), Pylaiella littoralis (AY119760
[GenBank]
/AY119724), Skeletonema costatum (AY119761
[GenBank]
) Alexandrium catenella (AB025590
[GenBank]
), Alexandrium tamarense (AB025589
[GenBank]
), Amphidinium carterae (AY004259
[GenBank]
), Amphidinium operculatum (AJ250262
[GenBank]
/AJ250264), Gonyaulax polyedra (AB025588
[GenBank]
), Heterocapsa niei (AF206709
[GenBank]
), Heterocapsa pygmaea (AF206707
[GenBank]
), Heterocapsa rotundata (AY004262
[GenBank]
), Heterocapsa triquetra (AB025587
[GenBank]
/AF130031), Prorocentrum micans (AB025585
[GenBank]
), Scrippsiella trochoidea (AF206710
[GenBank]
), Thoracosphaera heimii (AF206712
[GenBank]
), Gymnodinium mikimotoi (AB027234
[GenBank]
), Karlodinium micrum (AY119763
[GenBank]
), Karenia brevis (AY119762
[GenBank]
). Accession numbers for psbA and psaA sequences are listed before and after the divider, respectively. The alignment of these sequences are available at TreeBase (accession number SN1811). |
Acknowledgments |
|---|
|
TopNotesAbstractMethodsResults and DiscussionConclusionsAppendixAcknowledgmentsReferences |
|---|
We would like to thank D. Bhattacharya for original DNA and protein data sets for psaA and psbA. This work is supported by an operating grant 227085-00 from NSERC and a Sloan Research Fellowship awarded to AJR. AJR is supported as a Scholar by the CIAR Program in Evolutionary Biology and by a New Investigator Fellowship from the Canadian Institutes for Health Research (CIHR). YI is supported by the "Prokaryote Genome Evolution and Diversity Project" of Genome Atlantic. AGBS thanks the CIHR (for fellowship support), the CIAR, and Genome Atlantic. JBD is supported by a doctoral research award from the CIHR, and an operating grant MT4467 from the CIHR (awarded to W. F. Doolittle, Dalhousie University, Halifax, Canada).
Associate Editor: Gavin Naylor
|
Notes |
|---|
|
TopNotesAbstractMethodsResults and DiscussionConclusionsAppendixAcknowledgmentsReferences |
|---|
3 Current address: Department of Zoology, National History Museum Cromwell Road London SW7 5BD United Kingdom
|
References |
|---|
|
TopNotesAbstractMethodsResults and DiscussionConclusionsAppendixAcknowledgmentsReferences |
|---|
-
Baldauf S. L., Roger A. J., Wenk-Siefert I., Doolittle W. F. A kingdom-level phylogeny of eukaryotes based on combined protein data. Science (2000) 290:972–977.
Cavalier-Smith T. Principles of protein and lipid targeting in secondary symbiogenesis: Euglenoid, dinoflagellate, and sporozoan plastid origins and the eukaryotic family tree. J. Eukaryot. Microbiol. (1999) 46:347–366.[CrossRef][Web of Science][Medline]
Delwiche C. F. Tracing the thread of plastid diversity through the tapestry of life. Am. Naturalist (1999) 154:S164–S177.[CrossRef][Medline]
Fast N. M., Kissinger J. C., Roos D. S., Keeling P. J. Nuclear-encoded, plastid-targeted genes suggest a single common origin for apicomplexan and dinoflagellate plastids. Mol. Biol. Evol. (2001) 18:418–426.
Felsenstein J. PHYLIP (Phylogeny Inference Package) (1993) Seattle: University of Washington.
Foster P. G., Hickey D. A. Compositional bias may affect both DNA-based and protein-based phylogenetic reconstructions. J. Mol. Evol. (1999) 48:284–290.[CrossRef][Web of Science][Medline]
Galtier N., Gouy M. Inferring pattern and process: Maximum-likelihood implementation of a nonhomogeneous model of DNA sequence evolution for phylogenetic analysis. Mol. Biol. Evol. (1998) 15:871–879.[Abstract]
Graur D., Li W. Rates and patterns of nucleotide substitution. In: Fundamentals of molecular evolution (2000) Sunderland, Massachusetts: Sinauer. 99–164. Pages.
Harper J. T., Keeling P. J. Nucleus-encoded, plastid-targeted glyceraldehyde-3-phosphate dehydrogenase (GAPDH) indicates a single origin for chromalveolate plastids. Mol. Biol. Evol. (2003) 20:1730–1735.
Kumar S., Tamura K., Jakobsen I. B., Nei M. MEGA2: Molecular Evolutionary Genetics Analysis Software. Bioinformatics (2001) 17:1244–1245.
McFadden G. I., Waller R. F. Plastids in parasites of human. Bioessays (1997) 19:1033–1040.[CrossRef][Web of Science][Medline]
Palmer J. D. The symbiotic birth and speard of plastids: How many times and whodunit? J. Phycol. (2003) 39:4–11.[CrossRef][Web of Science]
Rosenberg M. S., Kumar S. Heterogeneity of nucleotide frequency among evolutionary lineages and phylogenetic inference. Mol. Biol. Evol. (2003) 20:610–621.
Schmidt H. A., Strimmer K., Vingron M., Von Haeseler A. TREE-PUZZLE: Maximum likelihood phylogenetic analysis using quartets and parallel computing. Bioinformatics (2002) 18:502–504.
Shimodaira H. An approximately unbiased test of phylogenetic tree selection. Syst. Biol. (2002) 51:492–508.
Shimodaira H., Hasegawa M. CONSEL: For assessing the confidence of phylogenetic tree selection. Bioinformatics (2001) 17:1246–1247.
Sullivan J., Holsinger K. E., Simon C. The effect of topology on estimates of among-site rate variation. J. Mol. Evol. (1996) 42:308–312.[CrossRef][Web of Science][Medline]
Sullivan J., Swofford D. L. Should we use model-based methods for phylogenetic inference when we know that assumptions about among-site rate variation and nucleotide substitution pattern are violated? Syst. Biol. (2001) 50:723–729.
Swofford D. L. PAUP* version 4.0b3 (1998) Sunderland, Massachusetts: Sinauer.
Swofford D. L., Waddell P. J., Huelsenbeck J. P., Foster P. G., Lewis P. O., Rogers J. S. Bias in phylogenetic estimation and its relevance to the choice between parsimony and likelihood methods. Syst. Biol. (2001) 50:525–539.
Takishita K., Nakano K., Uchida A. Preliminary phylogenetic analysis of plastid-encoded genes from an anomalously pigmented dinoflagellate Gymnodinium mikimotoi (Gymnodiniales, Dinophyta). Phycol. Res. (1999) 47:257–262.[CrossRef]
Takishita K., Nakano K., Uchida A. Origin of the plastid in the anomalously pigmented dinoflagellate Gymnomodium mikimotoi (Gymnodiniales, Dinophyta) as inferred from phylogenetic analysis based on the gene encoding the large subunit of form I-type RuBisCO. Phycol. Res. (2000) 48:85–89.[CrossRef]
Takishita K., Uchida A. Molecular cloning and nucleotide sequence analysis of psbA from the dinoflagellates: Origin of the dinoflagellate plastid. Phycol. Res. (1999) 47:207–216.[CrossRef]
Tengs T., Dahlberg O. J., Shalchian-Tabrizi K., Klaveness D., Rudi K., Delwiche C. F., Jakobsen K. S. Phylogenetic analyses indicate that the 19' Hexanoyloxy-fucoxanthin-containing dinoflagellates have tertiary plastids of haptophyte origin. Mol. Biol. Evol. (2000) 17:718–729.
Yang Z., Roberts D. On the use of nucleic acid sequences to infer early branchings in the tree of life. Mol. Biol. Evol. (1995) 12:451–458.[Abstract]
Yang Z. PAML: A program package for phylogenetic analysis by maximum likelihood. Comput. Appl. Biosci. (1997) 13:555–556.
Yoon H. S., Hackett J. D., Bhattacharya D. A single origin of the peridinin- and fucoxanthin-containing plastids in dinoflagellates through tertiary endosymbiosis. Proc. Natl. Acad. Sci. U.S.A. (2002a) 99:11724–11729.
Yoon H. S., Hackett J. D., Pinto G., Bhattacharya D. The single, ancient origin of chromist plastids. Proc. Natl. Acad. Sci. U.S.A. (2002b) 99:15507–15512.
Zhang Z., Green B. R., Cavalier-Smith T. Single gene circles in dinoflagellate chloroplast genomes. Nature (1999) 400:155–159.[CrossRef][Medline]
Zhang Z., Green B. R., Cavalier-Smith T. Phylogeny of ultra-rapidly evolving dinoflagellate chloroplast genes: A possible common origin for sporozoan and dinoflagellate plastids. J. Mol. Evol. (2000) 51:26–40.[Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  K. Shalchian-Tabrizi, M. Skanseng, F. Ronquist, D. Klaveness, T. R. Bachvaroff, C. F. Delwiche, A. Botnen, T. Tengs, and K. S. Jakobsen Heterotachy Processes in Rhodophyte-Derived Secondhand Plastid Genes: Implications for Addressing the Origin and Evolution of Dinoflagellate Plastids Mol. Biol. Evol., August 1, 2006; 23(8): 1504 - 1515. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. R. Bennett and S. Mathews Phylogeny of the parasitic plant family Orobanchaceae inferred from phytochrome A Am. J. Botany, July 1, 2006; 93(7): 1039 - 1051. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. R. Bachvaroff, M. V. Sanchez Puerta, and C. F. Delwiche Chlorophyll c-Containing Plastid Relationships Based on Analyses of a Multigene Data Set with All Four Chromalveolate Lineages Mol. Biol. Evol., September 1, 2005; 22(9): 1772 - 1782. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. S. Yoon, J. D. Hackett, F. M. Van Dolah, T. Nosenko, K. L. Lidie, and D. Bhattacharya Tertiary Endosymbiosis Driven Genome Evolution in Dinoflagellate Algae Mol. Biol. Evol., May 1, 2005; 22(5): 1299 - 1308. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||