Introduction

It is a general tenet of recent evolutionary biology that the red algae, green plants, and glaucophytes constitute the “primary photosynthetic eukaryotes,” whose plastids have two bounding membranes and likely originated directly from a cyanobacterium-like prokaryote via primary endosymbiosis (e.g., Bhattacharya and Medlin 1995; Delwiche 1999; McFadden 2001; Cavalier-Smith 2002). By contrast, the plastids of other lineages of eukaryotic phototrophs appear to be the result of secondary or tertiary endosymbiotic events (involving a phototrophic eukaryote and a host cell) because they are surrounded by three or four bounding membranes, and the endosymbiotic remnant of the photosynthetic eukaryotic nucleus, called the nucleomorph, may be recognized between the bounding membranes (e.g., Delwiche 1999; McFadden 2001; Cavalier-Smith 2002a; Yoon et al. 2002a). Based on phylogenetic analyses of plastid-coding genes and the similarity of the plastid genome organization (e.g., Morden et al. 1992; Nelissen et al. 1995; Bhattacharya and Medlin 1995), the plastids from all photosynthetic eukaryotes can be considered the product of a single primary endosymbiosis. Recently, complete sequences from various plastids have been determined (e.g., Ohyama et al. 1986; Ohta et al. 2003) and phylogenetic analyses using multiple plastid genes from a wide-range of eukaryotic lineages have been carried out in order to resolve robust phylogenetic relationships of plastids (e.g., Martin et al. 2002; Yoon et al. 2002a, b; Maul et al. 2002; Ohta et al. 2003). However, some of the phylogenetic relationships of plastids remain ambiguous or conflict between different phylogenetic methods used for nucleotide and amino acid sequences, especially the phylogenetic positions of the secondary plastids (e.g., Martin et al. 2002; Yoon et al. 2002a, b; Maul et al. 2002; Ohta et al. 2003). Therefore, an alternative methodology to infer the plastid phylogeny is needed.

The most dramatic feature distinguishing plastids from free-living cyanobacteria is the tremendous reduction in the size and gene content of plastid genomes (Martin and Herrmann 1998; Palmer and Delwiche 1998; Delwiche 1999; Martin et al. 2002). In addition, the size of the photosynthetic plastid genomes and number of genes are variable, ranging from 118 to 201 kbp with 62–209 genes (see Ohta et al. 2003). This suggests that plastids have gradually lost their genes during evolution, even after the primary endosymbiosis, since the reverse evolution of plastid gene loss is generally impossible (Martin et al. 2002). Therefore, the loss of plastid genes seems to be a good indicator of plastid evolution and it should be an alternative to using nucleotide/amino acid substitutions to construct the plastid phylogeny.

This study inferred plastid phylogeny and evolution based on the “loss of plastid genes” deduced from complete plastid genome sequences from a wide range of eukaryotic phototrophs. We used the Camin-Sokal or irreversible model (Camin and Sokal 1965) to conduct a cladistic analysis of the loss of plastid genes. The cladistic analysis of 274 genes from 20 OTUs (operational taxonomic units) produced robust support for the phylogenetic positions of the glaucophyte plastid and the secondary plastids of Euglena and the red lineage.

Materials and Methods

The data matrix we used to examine the maintenance or loss of plastid protein-coding genes was that used by Martin et al. (2002, supplementary data). It included the five additional OTUs listed in Table 1 and the cysW gene (Ohta et al. 2003), but excluded two nonphotosynthetic organisms (Plasmodium falciparum and Epifapus virsiniana) whose plastids are highly reduced. Seven pseudogenes of Pinus (ndhB, ndhC, ndhD, ndhE, ndhH, ndhI, and ndhK, Wakasugi et al. 1994), which Marin et al. (2002) designated as loss (-), were considered to represent the maintenance of the genes in the present analyses. The annotations of the genes from the additional plastid sequences were based on their descriptions in GenBank, and on their similarity using the search program Blastx at NCBI (http://www.ncbi.nlm.nih.gov/BLAST/ ). The data matrix consisted of 274 characters from 20 eukaryotic OTUs (Fig. 1), and is available from HN on request.

Table 1 List of additional OTUs (Martin et al. 2002) included in the cladistic analyses (Fig. 1) and origin of the plastid genome sequences
Full size table
Figure 1
figure 1

Single most parsimonious (MP) tree (with tree length of 557 and consistency index of 0.4147) based on the present cladistic analyses of 274 plastid genes from 20 OTUs representing a wide-range of eukaryotic phototrophs. The tree was constructed based on the branch-and-bound search by PAUP 4.0b10 (Swofford 2002). Irreversible Camin-Sokal model (Camin and Sokal 1965; Wiley et al. 1991) was used for setting character types in loss or maintenance of 269 plastid genes by PAUP 4.0b10, whereas five other genes harbored in the mobile group I and II introns were designated unordered. Branch lengths are proportional to numbers of character changes, which are indicated by numbers below branches, based on ACCTRAN by PAUP 4.0b10. Numbers above branches are the bootstrap values (50% or more) based on 1000 replications of the branch-and-bound search by PAUP 4.0b10. Single and double asterisks indicate the primary and secondary plastids, respectively.

Full size image

We carried out a cladistic analysis of the loss of plastid genes after primary endosymbiosis using this data matrix with PAUP 4.0b10 (Swofford 2002). Since the regaining of plastid genes during plastid evolution cannot generally be deduced, we used an irreversible Camin-Sokal model (Camin and Sokal 1965; Wiley et al. 1991) to set character types for the loss or maintenance of 269 plastid genes in PAUP4.0b10. Five other genes (cuvI, matK, ycf13, ycf68, and ycf74) harbored in the mobile group I and II introns (see Lambowitz and Belfort 1993) were designated as unordered characters. The loss of plastid genes is a derived character. Therefore, the loss and maintenance of a gene are polarized a priori in the transformation series. Consequently, the root or ingroup node of the phylogenetic tree was determined without outgroup comparison after the cladistic analyses using PAUP4.0b10 (see Wiley et al. 1991). A cladistic or maximum parsimony analysis of the data matrix was carried out using a branch-and-branch search with PAUP 4.0b10 and the robustness of the resulting lineages was tested with bootstrap analyses (Felsenstein 1985) involving 1000 replications of the branch-and-branch search with PAUP 4.0b10.

Results

Based on our cladistic analysis of 274 genes from 20 OTUs, a single MP tree was constructed and is presented in Fig. 1. The tree was 557 steps long, with a consistency index of 0.4147 and a retention index of 0.8939.

Two major monophyletic groups were resolved with relatively high bootstrap values (70–84%), with the root of the plastids positioned between these two monophyletic groups. One group consisted of the green lineage (green plants plus Euglena) and the glaucophyte Cyanophora, and the other (the red lineage) contained three red algal plastids (Porphyra, Cyanidium, and Cyanidioschyzon) and secondary plastids from a diatom (Odontiella) and cryptophyte (Guillardia).

Within the former group, Cyanophora is basal to the green lineage, which was subdivided into two clades supported with relatively high bootstrap support (84–86%). One of the two clades was composed of two prasinophycean algae (Mesostigma and Nephroselmis) in basal positions and a robust monophyletic group (98% bootstrap support) consisting of Chlorella, Chlamydomonas, and Euglena, with high bootstrap support (99%) for the sister relationship between Chlamydomonas and Euglena. The other clade consisted of five angiosperms (Zea, Oryza, Spinacea, Oenothera, and Nicotiana), the gymnosperm Pinus, the fern Psilotum, the liverwort Marchantia, and the charophycean alga Chaetosphaeridium, although the basal phylogenetic relationships within the clade were almost completely ambiguous (Fig. 1).

Two robust monophyletic groups were resolved within the red lineage with 95–100% bootstrap values: one composed of the two secondary plastids (Guillardia and Odontella) and the other consisting of two cyanidiophycean algae (Cyanidium and Cyanidioschyzon) (Fig. 1). However, the phylogenetic position of the red alga Porphyra was ambiguous within the red lineage.

Discussion

Our cladistic analysis of 274 genes from 20 OTUs, including three secondary plastids, resolved two major lineages with relatively high bootstrap support: the red lineage and a large monophyletic group composed of the green lineage (green plants and Euglena) and the basal glaucophyte plastid. These phylogenetic relationships suggest that the in-group node of the plastid lineage is positioned between these two major lineages and a sister relationship of the plastids in the green lineage and the Glaucophyta (Fig. 1). In contrast, recent plastid multigene phylogenies involving more than 8,000 amino acid sequences demonstrated that the glaucophyte Cyanophora occupies the basal position within the plastids (Adachi et al. 2000; Martin et al. 2002, Ohta et al. 2003). This discrepancy may result from the small taxon sampling (see Zwickl and Hillis 2002) from the Glaucophyta (only Cyanophora paradoxa) for both phylogenetic analyses. Additional OTUs from the Glaucophyta may resolve this problem. However, the sister relationship between the green lineage and the glaucophyte plastid resolved using our cladistic method is consistent with the rbcL gene phylogeny (Delwiche and Palmer 1997) and one recent phylogenetic study using multiple nuclear genes (Nozaki et al. 2003). In contrast to the monophyly of most of the plastid genes within the cyanobacterial lineage (e.g., Morden et al. 1992; Nelissen et al. 1995; Bhattacharya and Medlin 1995), the rbcL gene tree contains two separate lineages: the red lineage and the green/glaucophyte lineage (Delwiche and Palmer 1997). This polyphyletic status of the plastids is generally attributed to horizontal gene transfer (Morden et al. 1992; Delwiche and Palmer 1997). Very recently, Nozaki et al. (2003) carried out phylogenetic analyses of various lineages of only mitochondria-containing eukaryotic organisms using nuclear multigene sequences, including the complete sequence from the primitive red alga Cyanidioschyzon merolae. They resolved the position of the red algae as basal to a large monophyletic group including green plants and the Glaucophyta. If one assumes that primary endosymbiosis occurred only once and the phylogenetic relationships resolved by Nozaki et al. (2003) are reliable, our phylogenetic position of the Glaucophyta (Fig. 1) should be natural.

Since the plastids of the Euglenophyta or Euglenozoa (Discicristata) are surrounded by three bounding membranes and contain chlorophylls a and b, they are considered to have originated from the secondary endosymbiosis of green plants (e.g., Delwiche 1999; McFadden 2001). However, the phylogenetic position of the Euglena plastid conflicts between different methods used to deduce multigene phylogenies, even based on the same amino acid sequence data (8,308 a.a. from 16 OTUs) by Martin et al. (2002). The basal position of the Euglena plastid within the green lineage was resolved with 86–100% bootstrap support using distance methods, whereas ML analyses resolved the sister relationship between Euglena and Chlorella with 64% bootstrap support (Martin et al. 2002). Turmel et al. (1999), on the basis of ML analyses of 7,499 a.a. from 37 plastid genes, showed that the Euglena plastid is sister to the clade composed of Chlorella (Trebouxiophyceae) and Chlamydomonas (Chlorophyceae). Our cladistic analysis demonstrated that Euglena and Chlamydomonas formed a very robust clade (99% bootstrap support) that is sister to Chlorella (Fig. 1). Since recent molecular phylogenetic analyses based on 18S rRNA genes and comparative ultrastructural data suggest the sister relationship between the Chlorophyceae and Trebouxiophyceae (e.g., Friedl 1997; Graham and Wilcox 2000), the ancestor of the euglenophyte plastid should be a chlorophycean alga.

Very recently, several cyanobacterial or plant-like genes were found in plastid-lacking organisms (Heterolobosea and Kinetoplastida) belonging to the Discicristata (Andersson and Roger 2002; Hannaert et al. 2003). Since the Euglenozoa have secondary plastids and are closely related to the Kinetoplastida, Hannaert et al. (2003) discussed that a common ancestor of both Euglenozoa and Kinetoplastida already acquired the secondary plastids that gave rise to the plant-like genes in the Kinetoplastida nuclear genome. However, phylogenetic analyses of various euglenoid taxa suggested that the evolution of phototrophy occurred in the common ancestor of the monophyletic plastid-containing group (Euglenales) distally positioned within the Euglenozoa (Montegut-Felkner and Triemer 1997; Preisfeld et al. 2000; Müllner et al. 2001), and the Heterolobosea are positioned outside the clade composed of Euglenozoa and Kinetoplastida (Baldauf et al. 2000; Nozaki et al. 2003). In addition, the present cladistic analysis suggested the relatively recent acquisition of the secondary plastid of Euglena (after the divergence of Chlorophyceae and Trebouxiophyceae) (Fig. 1). Therefore, the secondary plastid endosymbiosis can hardly be considered before the divergence of Euglenozoa and Kinetoplastida or Heterolobosea, and the cyanobacterial or plant-like genes found in Heterolobosea and Kinetoplastida (Andersson and Roger 2002; Hannaert et al. 2003) are possibly derived from the primary plastid endosymbiosis. Nozaki et al. (2003) suggested that the primary plastid endosymbiosis likely occurred once in the common ancestor of three primary plastid-containing lineages as well as in Discicristata, Heterokontophyta, and Alveolata (apicomplexans and Ciliophora), from which the primary plastids were subsequently lost, but some of their genes have remained in the nuclear genomes (Andersson and Roger 2002; Nozaki et al. 2003).

The phylogenetic positions of the secondary plastids within the red lineage conflict in recent phylogenetic analyses of multigene sequences (Martin et al. 2002; Maul et al. 2002; Yoon et al. 2002a, b; Ohta et al. 2003). Based on combined psaA and psbA sequences, the cryptophytes are robustly separated from the clade composed of the Cyanidiophyceae (Cyanidium, Cyanidioschyzon, etc.) and other secondary plastid-containing algae (Heterokontophyta and Haptophyta) (Yoon et al. 2002a). Nonmonophyly of the secondary plastids in the red lineage was also resolved in the phylogenetic studies using more than 8,000 amino acid sequences from limited numbers of OTUs (Martin et al. 2002; Maul et al. 2002; Ohta et al. 2003). In contrast, Yoon et al. (2002b) resolved a large monophyletic group composed of the secondary plastids from Heterokontophyta, Cryptophyta, and Haptophyta in the red lineage, in an analysis of the combined 16S rRNA, psaA, psbA, rbcL, and tufA genes DNA sequences (5827 nt from 29 OTUs of the red lineage plus 6 OTUs). They argue that the separate phylogenetic position of the Cryptophyta is derived from the small number of OTUs analyzed by Martin et al. (2002). Our results based on gene loss (Fig. 1) are essentially consistent with Yoon et al. (2002b) in that the secondary plastids of the red lineage have a single ancestor.

The phylogenetic position of Mesostigma is conflicting in recent phylogenetic analyses using multiple plastid gene sequences (Lemieux et al. 2000; Martin et al. 2002; Maul et al. 2002; Ohta et al. 2003). This primitive green alga is positioned most basally within the green plants (Lemieux et al. 2000; Maul et al. 2002) or represents the most basal lineage within the Streptophyta (embryophytes and Charophyceae) (Martin et al. 2002; Ohta et al. 2003). In contrast, the present cladistic analysis of loss of plastid genes suggested the basal position of Mesostigma within the Chlorophyta (Nephroselmis, Chlamydomonas, etc.) (Fig. 1). This situation may result from limited taxon sampling from Chlorophyta in the present study (especially in lacking ulvophycean algae). However, pigment composition (Fawley and Lee 1990) and presence of stigma (Graham and Wilcox 2000) in Mesostigma suggest its phylogenetic position within the Chlorophyta.

As discussed above, the phylogenetic positions of the secondary plastids in both the red and green lineages appeared ambiguous, even when a large nucleotide or amino acid sequences from multiple plastid genes were analyzed. This is probably based on the unusual bias for gene substitutions related to secondary endosymbiosis, which also caused highly divergent gene sequences of the vestigial nucleus or nucleomorph in the Cryptophyta and Chlorarachniophyta (see Van de Peer et al. 1996). According to Adachi et al. (2000), Euglena deviates most strongly from other plastid genomes at the level of amino acid composition. Recently, Itoh et al. (2002) compared the complete genome sequences of two different endosymbionts, and demonstrated that the rate of amino acid substitution is two times higher in symbionts than in their relatives, suggesting that the elevated evolutionary rate is mainly due to an enhanced mutation rate in endosymbiosis. Therefore, traditional molecular phylogenetic methods using nucleotide or amino acid substitutions should have limited efficiency for resolving the phylogeny of secondary and tertiary plastids especially when the numbers of OTUs are limited. Since the loss of the plastid genes are not directly related to such unusual gene substitutions, the present phylogenetic results regarding the secondary plastids (Fig. 1) can be reliable alternative hypotheses. However, as in the traditional molecular phylogenetic studies using nucleotide or amino acid substitutions, the present cladistic method is potentially associated with the possible clustering of some lineages due to convergent and/or accelerated evolution especially when the taxon sampling is poor. Further information of complete plastid genome sequences should resolve more natural phylogenetic relationships of plastids.