Ribosomal DNA phylogeny of the Bangiophycidae (Rhodophyta) and the origin of secondary plastids^†

Abstract

We sequenced the nuclear small subunit ribosomal DNA coding region from 20 members of the Bangiophycidae and from two members of the Florideophycidae to gain insights into red algal evolution. A combined alignment of nuclear and plastid small subunit rDNA and a data set of Rubisco protein sequences were also studied to complement the understanding of bangiophyte phylogeny and to address red algal secondary symbiosis. Our results are consistent with a monophyletic origin of the Florideophycidae, which form a sister-group to the Bangiales. Bangiales monophyly is strongly supported, although Porphyra is polyphyletic within Bangia. Bangiophycidae orders such as the Porphyridiales are distributed over three independent red algal lineages. The Compsopogonales sensu stricto, consisting of two freshwater families, Compsopogonaceae and Boldiaceae, forms a well-supported monophyletic grouping. The single taxon within the Rhodochaetales, Rhodochaete parvula, is positioned within a cluster containing members of the Erythropeltidales. Analyses of Rubisco sequences show that the plastids of the heterokonts are most closely related to members of the Cyanidiales and are not directly related to cryptophyte and haptophyte plastid genomes. Our results support the independent origins of these secondary algal plastids from different members of the Bangiophycidae.

The Rhodophyta (red algae) form a distinct eukaryotic lineage (Bhattacharya et al., 1990; Van de Peer et al., 1996; Stiller and Hall, 1997) that likely shares a most recent common ancestry with the green algae (Burger et al., 1999; Moreira, Le Guyader, and Philipper, 2000). Members of the red algae share a suite of characters that do not occur together in any other eukaryote, such as a complete lack of flagellated stages and basal bodies, a two-membraned “simple” plastid (Bhattacharya and Medlin, 1995) that lacks chlorophyll b or c, have unstacked thylakoids with phycobilisomes, and photosynthetic reserves stored as floridean starch (Garbary and Gabrielson, 1990). Traditionally, systematists have divided the red algae into two groups, sometimes with the taxonomic status of classes, i.e., Bangiophyceae and Florideophyceae (cf. Van den Hoek, Mann, and Jahns, 1995), or subclasses, Bangiophycidae and the Florideophycidae within a single class Rhodophyceae (Gabrielson, Garbary, and Scagel, 1985), with the latter approach being more generally accepted. The Florideophycidae includes morphologically complex red algae in orders such as the Gigartinales and the Ceramiales and is widely considered to be a derived, monophyletic group (Garbary and Gabrielson, 1990; Freshwater et al., 1994; Ragan et al., 1994; Saunders and Kraft, 1997).

The Bangiophycidae (bangiophytes) retain morphological characters that are considered to have been present in the ancestral pool of red algae and range from unicells to multicellular filaments (uniseriate or multiseriate; branched or unbranched) or sheet-like thalli (monostromatic or distromatic; Garbary, Hansen, and Scagel, 1980). Cells typically have a single axial stellate plastid with a large pyrenoid, though some taxa contain a cup-shaped plastid, or a complex, interconnected plastid with no pyrenoid (Gantt, Scott, and Lipschultz, 1986; Broadwater and Scott, 1994). Pit connections are rare, and sexual reproduction has not been described in many members of this subclass (Garbary, Hansen, and Scagel, 1980; Gabrielson et al., 1990). Studies of biochemical, morphological, and ultrastructural characters (Gabrielson and Garbary, 1985; Gabrielson, Garbary, and Scagel, 1985; Garbary and Gabrielson, 1990) and ribosomal DNA (rDNA) sequence comparisons (Ragan et al., 1994; Oliveira et al., 1995; Oliveira and Bhattacharya, 2000) suggest that the Bangiophycidae is a diverse assemblage of algae that includes the Florideophycidae. For this reason, only one class of red algae, the Rhodophyceae, is recognized by some researchers (e.g., Gabrielson et al., 1990). Nevertheless, the name Bangiophycidae continues to be used to identify the early diverging red algae from which the Florideophycidae have evolved. The Bangiophycidae have been divided into four orders: Bangiales, Compsopogonales, Porphyridiales, and Rhodochaetales (Gabrielson et al., 1990; Garbary and Gabrielson, 1990). Recently, two new orders were suggested, the Cyanidiales (Seckbach and Ott, 1994), which is proposed to include thermophilic genera such as Galdieria and Cyanidium, and the order Erythropeltidales (Garbary, Hansen, and Scagel, 1980; Silva, Basson, and Moe, 1996). The latter grouping is based on the greater morphological complexity in members of the Compsopogonaceae than in those of the Erythrotrichiaceae. Nonetheless, Rintoul, Vis, and Sheath (1999) provided molecular evidence that does not support recognition of the Eryhtropeltidales.

Previous molecular studies indicate that, whereas the Bangiales and Compsopogonales appear to be monophyletic, other orders, such as the Porphyridiales, may be either biphyletic or paraphyletic within the Bangiophycidae (Ragan et al., 1994; Rintoul, Vis, and Sheath, 1999; Oliveira and Bhattacharya, 2000). In their analysis of red algal orders using a variety of morphological, ultrastructural, and biochemical data, Gabrielson and Garbary (1985), Gabrielson, Garbary, and Scagel (1985), and Garbary and Gabrielson (1990) also hypothesized that the Porphyridiales were probably paraphyletic and found considerable differences in plastid ultrastructure and highly variable vegetative and reproductive morphologies. They noted that the remaining three orders, except perhaps for the Compsopogonales (including the Erythrotrichiaceae), were probably monophyletic.

In this study, we compare nuclear small subunit (SSU) rDNA sequences from members of all the orders of the Bangiophycidae to address the following questions: (1) Is there phylogenetic support for the present ordinal classification of the Bangiophycidae? (2) Which Bangiophycidae order is the sister group of the Florideophycidae? and (3) It is now widely accepted that the cryptophytes have gained their plastid through a secondary symbiosis involving a bangiophyte red alga (Douglas et al., 1991; Bhattacharya and Medlin, 1995; Palmer and Delwiche, 1996; Douglas and Penny, 1999), yet the specific donor of this plastid remains unknown (see also Oliveira and Bhattacharya, 2000). We ask, therefore, if there is a specific evolutionary relationship between the nucleomorphs in cryptophyte algae and any of the Bangiophycidae orders. The goals of this research are to present a phylogenetic hypothesis for the Bangiophycidae that can be tested in future studies to ultimately create a natural classification of the Rhodophyta and to gain further insights into the origin of plastids through red algal secondary symbiosis.

MATERIALS AND METHODS

Bangiophycidae were either collected in the field or obtained from the culture collections at the University of Texas (UTEX) or the Sammlung von Algenkulturen at the University of Göttingen (SAG; Schlösser, 1994). These taxa are listed in Table 1. Field-collected samples were cleaned of visible epiphytes, bases were removed to prevent contamination, and the specimens were stored at −20°C. Thalli were ground in liquid nitrogen, and the DNA was extracted according to the protocol outlined by Saunders (1993) with modifications given in Vis and Sheath (1997). Cultured material (100–400 mg, fresh mass) was ground in liquid nitrogen and total genomic DNA was extracted using the DNeasy Plant Mini Kit (Qiagen, Santa Clarita, California, USA).

For DNA isolated from the SAG cultures, the SSU rDNA coding region was amplified with the polymerase chain reaction (PCR) using synthetic oligonucleotide primers that recognize conserved sequences at the termini of the nuclear-encoded SSU rDNA genes (modified from Medlin et al., 1988): 18S5′ (5′-AACCTGGTTGATCCTGCCAGTRGTSAT ATGCTTGT-3′) and 18S3′ (5′-GATCCTTCTGCAGGTTCACCTACGGAA-3′). For the remaining taxa of the Bangiophycidae (excluding the Bangiales), the rDNA coding region was amplified in three subfragments, ranging in size from 500 to 600 base pairs (bp). The following primers were used in these experiments: G01.1 (5′-AACCTGATTGATCCTGCC AG-3′) and G10.1 (5′-GCGCGCCTGCTGCTTCCTTGG-3′); G02.1 (5′-CGATTCCGGAG AGGGAGCCTG-3′) and G08.1 (5′-GAACGGCCATGCACCACCAAC-3′) (modified from Saunders and Kraft [1994]); G06 and G07 (Saunders and Kraft, 1994) or the primers designed by White et al. (1990): NS1 and NS2; NS3 and NS4; NS5 and NS8 or primers designed by Medlin et al. (1988). Bangiales SSU rDNA were amplified in two subfragments using the primers G01.1 and G10.1, and G02.1 and G15.1 (modified from Saunders and Kraft [1994]). Coding regions that contained group I introns in position 516 (E. coli numbering; see Bhattacharya, 1998) of the SSU rDNA, required additional primers (not shown) to allow for sequencing of these insertions.

The purified rDNA fragments were then either sequenced directly or cloned using the TA cloning kit and the plasmid vector pCR2.1 (Invitrogen, Carlsbad, California, USA) prior to sequencing. Plasmid DNA was isolated and purified using the Plasmid Midi Kit (Qiagen). The rDNA sequences were determined over both strands using an ABI-310 Genetic Analyzer (Perkin-Elmer, Norwalk, Connecticut, USA), and the primers listed above or in Elwood, Olsen, and Sogin (1985), with the Dye Terminator Cycle Sequencing Ready Reaction Kit (Perkin-Elmer). The GenBank accession numbers of the SSU rDNA sequences determined in this study are shown in Table 1.

Sequences were aligned according to the secondary structure models of Gutell (1996) and Van de Peer et al. (1997), optimized manually using the SeqApp program (Gilbert, 1992), and a total of 1582 sequence positions (626 parsimony-informative characters) were used in the phylogenetic analyses (alignment available from D. Bhattacharya). A tree was inferred with fastDNAml 1.1 (Olsen et al., 1994) for maximum likelihood (ML) analyses using the global search option, transition/transversion ratio = 2, and a jumbled sequence input. The ML tree was rooted on the branch length leading to green algal (chlorophyte) rDNA sequences. The data set was also analyzed with a distance method in which the matrix was calculated with the HKY-85 model (Hasegawa, Kishino, and Yano, 1985) with equal rates of change across sites and the tree was built with neighbor-joining (Saitou and Nei, 1987). A neighbor-joining tree was also built with LogDet distances (Lockhart, Steel, and Penny, 1994). Bootstrap analyses (Felsenstein, 1985; 2000 replications) were done using both HKY-85 and LogDet distances. Missing data and gaps were excluded from pairwise comparisons in these analyses. All phylogenetic analyses except for fastDNAml were done with PAUP* (V4.0b4a; Swofford, 2000). The GenBank accession numbers of the existing nuclear SSU rDNA sequences are shown in the ML tree.

To test hypotheses generated by the initial tree, we concatenated all available nuclear and plastid SSU rDNA sequences from the same bangiophyte strains (ten sequences) and inferred trees from this data set (2953 total characters, 784 parsimony-informative characters) using PAUP*. The tree was inferred with the ML method in PAUP* using empirically determined nucleotide frequencies, equal rates across sites, with the transition/transversion ratio = 2, and the HKY-85 model of sequence evolution. Initial trees were built stepwise with ten random-sequence additions, and these trees were rearranged with tree-bisection reconnection (TBR) to find the phylogeny of highest likelihood. Two-hundred bootstrap replicates were analyzed with this method. The ML tree was rooted on the branch length leading to the rDNA sequences of the chlorophyte Mesostigma viride. We also did bootstrap analyses (2000 replications) of this data set using neighbor-joining and HKY-85 or LogDet distances as described above. Finally, unweighted maximum parsimony (MP) was used to calculate bootstrap values (2000 replications) for the concatenated rDNA sequences using the same sequence addition and tree rearrangement protocols as described above for the ML method. The GenBank accession numbers of the existing plastid-encoded SSU rDNA sequences are shown in the ML tree.

A third data set of existing plastid-encoded large subunit sequences of ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco, EC 4.1.1.39) from rhodophytes, cryptophytes, haptophytes, and heterokonts was assembled to provide another perspective on bangiophyte phylogeny and secondary symbiosis. A total of 494 amino acids (161 parsimony-informative characters) were used for the phylogenetic analyses using PAUP*. The tree was built with the unweighted MP method as described above, and the stability of monophyletic groups was tested with the bootstrap method (2000 replications). The MP tree was rooted on the branch length leading to Rubisco sequences from Cyanidiales and heterokonts. Bootstrap analysis (2000 replications) was also done using protein distances (mean character difference). The Rubisco sequences were used as input for quartet puzzling (Strimmer and van Haeseler, 1996) with protein distance (minimum evolution) as the optimality criterion (10 000 puzzling steps). Lastly, the first and second positions of Rubisco codons (954 characters, 292 parsimony-informative sites) were analyzed with quartet puzzling with DNA sequence distance (minimum evolution) as the optimality criterion. The GenBank accession numbers of the existing Rubisco sequences are shown in the MP tree. We also changed the topology of the Rubisco tree using the MacClade computer program (version 3.07; Maddison and Maddison, 1997) and compared these rearranged trees to the “best” tree using the Kishino-Hasegawa test (Kishino and Hasegawa, 1989 [within PAUP*]) to test different hypotheses about bangiophyte and secondary plastid evolution. In this analysis, the number of steps required for all trees were compared to test for significant (P < 0.05) differences among them.

RESULTS

The complete SSU rDNA gene was sequenced for all studied Bangiophycidae (except Stylonema alsidii CR for which the 5′-terminal 400 bp could not be obtained, likely due to PCR primer mismatch in this region). All rDNA genes were free of any major insertions or deletions except for Bangia fuscopurpurea and Porphyra leucosticta, which were of lengths 3341 and 2404 bp, respectively. Bangia fuscopurpurea contained two group I introns at positions 516 and 1506, whereas P. leucosticta contained a single group I intron at the 516 position. Bangia atropurpurea (LM01) and Bangia sp. (VIS7) also contain introns at the 516 position (Müller et al., 1998). These introns will not be further analyzed here. Maximum likelihood analysis of the SSU rDNA data set resulted in the tree shown in Fig. 1 (−Ln likelihood = 19963.4577). The distance analyses provided bootstrap support for most nodes in the ML tree. These analyses show the order Bangiales to be a well-supported monophyletic grouping (100% bootstrap with both methods) in which Porphyra is polyphyletic within Bangia. The Bangiales is positioned as a sister group to the Florideophycidae in all the analyses, but without bootstrap support. To test the effect of the highly divergent Florideophycidae rDNAs on the position of the Bangiales, the Florideophycidae sequences were removed from the data set shown in Fig. 1 and the data were analyzed with the distance method (HKY-85 model). This phylogeny supported the monophyletic origin of the Bangiales and their distant relationship to other red algal orders (tree not shown).

The Porphyridiales form three well-supported, yet independent clades. The first clade (Porphyridiales 1) contains the unicells, Rhodella maculata, R. violacea, and Glaucosphaera vacuolata. Another unicell, Dixoniella grisea, also groups with high bootstrap support with Porphyridiales 1 (99% LogDet, 98% HKY-85) as a sister of G. vacuolata (91% LogDet, 89% HKY-85). The length of the branch leading to D. grisea and its divergence position are shown in Fig. 1. This rDNA has been excluded from the analyses because its extreme divergence rate disrupts the phylogenies leading to the likely artifactual placement of Porphyridiales 1 as a sister group of the Florideophycidae (e.g., Zuccarello et al., 2000). This phenomenon is most certainly explained by long-branch attraction between the highly divergent D. grisea and Florideophycidae rDNA sequences (see Fig. 1 and below).

Porphyridiales 2 consists primarily of pseudofilamentous forms, Stylonema alsidii, S. alsidii (CR), S. cornu-cervi, Goniotrichopsis sublittoralis, Bangiopsis subsimplex, Chroodactylon ornatum, and C. ramosum, but also two unicells, Rhodosorus marinus and Rhodosorus sp. This clade is well supported by the bootstrap analyses (100%). Porphyridiales 3 contains the unicells Porphyridium aerugineum, P. sordidium, and Flintiella sanguinaria and is also well supported by the bootstrap analyses (100%). Monophyly of the Erythropeltidales, Erythrotrichia carnea and the closely related Erythrocladia sp. and Smithora naiadum, is strongly supported by the bootstrap analyses (100%). Rhodochaete parvula, which is the only member of the order Rhodochaetales, is a derived taxon that is sister to the Erythropeltidales (100% bootstrap; Fig. 1). Taxa within the Compsopogonales sensu stricto (Compsopogon coeruleus, Compsopogonopsis leptoclados, and Boldia erythrosiphon), all of which inhabit freshwater environments, are positioned on a strongly supported branch (99% bootstrap) with weak support for a sister group relationship with the Erythropeltidales/Rhodochaetales (64% LogDet, 69% HKY-85). The Galdieria sulphuraria sequence has strong bootstrap support as the earliest divergence within the Rhodophyta and the cryptophyte nucleomorphs form a well-supported monophyletic group in Fig. 1 with no obvious sister-group relationship to the bangiophyte orders in the analysis.

Phylogenetic analyses of the concatenated nuclear and plastid SSU rDNA sequences resulted in an identical tree with all methods in which virtually all nodes had strong bootstrap support. This tree is also remarkably similar to that shown in Fig. 1. The ten random additions and heuristic search with the ML method identified a single tree (−Ln likelihood = 17778.9589; see Fig. 2). This tree supports, with moderate to strong bootstrap support, the sister-group relationship of Bangiales and Florideophycidae, an early divergence of G. sulphuraria within the Rhodophyta, and the monophyly of previously identified bangiophyte orders and the cryptophyte nucleomorphs. The lack of plastid SSU rDNA sequences from Dixoniella and members of the Erythropeltidales/Rhodochaetales does not allow us to compare more precisely the nuclear vs. the nuclear + plastid rDNA phylogenetic frameworks.

The Rubisco tree shown in Fig. 3 is also generally consistent with the results of the rDNA analyses. The monophyly of the Bangiales and Florideophycidae is found in both protein- and DNA-level analyses, but only the protein distance method provides moderate bootstrap support (73%) for this node, as does the protein quartet puzzling analysis (puzzle value = 71). There is weak support for the monophyly of the Cyanidiales and heterokont plastids as was found with a recent analysis of plastid SSU rDNA (Oliveira and Bhattacharya, 2000). There is moderate to strong bootstrap support for the monophyly of the orders Bangiales, Erythropeltidales, and the Cyanidiales as there is for the monophyly of the secondary plastids in the haptophytes and of the secondary plastids in the heterokonts. The single sequence from the cryptophyte, Guillardia theta, groups with moderate bootstrap support with the haptophyte plastids. Surprisingly, the Compsopogonales do not have bootstrap support in the different Rubisco analyses, but all phylogenetic methods show the monophyly of this lineage, in agreement with the rDNA analyses (see Fig. 1).

Five different hypotheses about the phylogenetic history of bangiophytes were tested using the Kishino-Hasegawa test (Kishino and Hasegawa, 1989) with rearrangements of the “best” MP Rubisco tree shown in Fig. 3 (Tree 1). The hypotheses tested the strength of the following relationships: Bangiales/Florideophycidae (Tree 2), cryptophyte/haptophyte (Tree 3), and Cyanidiales/heterokont plastid (Tree 4), as well as the monophyly of the Compsopogonales (Tree 5) and the forced monophyly of cryptophyte/haptophyte/heterokont plastids (Tree 6; see Fig. 3). The results of the Kishino-Hasegawa test (Table 2) show that rearranged Trees 4 and 6 are significantly “worse” than the best tree shown in Fig. 3. This result provides strong support for the close evolutionary relationship of the plastids in the Cyanidiales and heterokont and argues against a single origin of the secondary plastids in the cryptophyte, haptophyte, and heterokont algae (Tree 6). Of the remaining trees, only rearranged Tree 2 showed a near significant difference (P = 0.0956) with respect to the best tree. Taken together, these data suggest that the Bangiales/Florideophycidae grouping has some support with the Rubisco data (as evident in the moderate protein distance bootstrap value), but the monophyly of the Compsopogonales sensu stricto and the cryptophyte/haptophytes are not robust results. In contrast, the nuclear rDNA trees clearly support the monophyly of the Compsopogonales, whereas there is little evidence for the monophyly of cryptophyte and haptophyte plastids when a broad diversity of plastid SSU rDNA sequences are analyzed (see Fig. 2 in Oliveira and Bhattacharya, 2000).

DISCUSSION

In spite of several attempts at classification (Skuja, 1939; Fritsch, 1945; Kylin, 1956; Chapman, 1974; Garbary, Hansen, and Scagel, 1980; Magne, 1989), the taxonomy of the subclass Bangiophycidae remains uncertain. Much of the existing taxonomy is based on the presence of unicellular or multicellular form, a likely artificial character (Garguilo, De Masi, and Tripodi, 1987). Gabrielson, Garbary, and Scagel (1985) used cladistic methods to determine a phylogenetic ranking of orders within the Rhodophyta. Based on their analysis, there was little support for separating red algal orders into two distinct taxa of higher rank, the Bangiophycidae and Florideophycidae. In addition, the Porphyridiales were determined to be paraphyletic within the Bangiophycidae, whereas the Bangiales, Rhodochaetales, and possibly the Erythropeltidales were monophyletic (Gabrielson, Garbary, and Scagel, 1985).

A completely different evolutionary scenario was proposed by Magne (1989). Based on previous morphological analyses of thallus structure by the French school of phycology (e.g., Chadefaud, 1969), and his own studies on reproductive cells, called cysts, he divided the Rhodophyceae into three subclasses: (1) Archaeorhodophycidae, lacking naked fertile cell-producing cysts, with a single order Porphyridiales; (2) Metarhodophycidae, defined by the presence of metacystis (only part of the fertile cell produces the reproductive cell) including the orders Erythropeltidales, Rhodochaetales, and Compsopogonales; and (3) Eurhodophycidae, defined by the presence of eucystis (the entire content of the fertile cell is transformed into the reproductive cells) including the order Bangiales and all the orders of the former Florideophycidae.

Ragan et al. (1994) used the nuclear SSU rDNA sequences to infer a phylogeny of the Rhodophyta; however, their analysis contained only six bangiophyte genera belonging to three of the five orders. Despite this limited sampling, Ragan et al. (1994) postulated that the Bangiophycidae were paraphyletic because it included the Florideophycidae and probably encompassed three or more distinct lineages. Oliveira and Bhattacharya (2000) studied plastid SSU rDNA coding regions and showed that Porphyridiales formed three distinct lineages, that Cyanidiales was likely an independent order, and postulated a sister-group relationship between Bangiales and Florideophycidae as predicted by Magne (1989). The present study is, however, the first to include numerous bangiophyte taxa including all five orders in an analysis of the nuclear SSU rDNA gene to resolve the systematics of this pivotal group of red algae.

Bangiales

This order comprises two genera, Porphyra and Bangia within one family, the Bangiaceae, which has been well defined based on life history and developmental characteristics (Garbary, Hansen, and Scagel, 1980; Garbary and Gabrielson, 1990). In the present study, this order is strongly supported as being monophyletic in all analyses, and this finding is in agreement with previous molecular studies (e.g., Ragan et al., 1994; Oliveira and Bhattacharya, 2000). However, the taxonomic relationship between the two genera within this order is not clear. Oliveira et al. (1995), Müller et al. (1998), Broom et al. (1999) and Oliveira and Bhattacharya (2000), found that Porphyra is polyphyletic within Bangia. A great deal more work will have to be done with naturally occurring populations of Bangiales (e.g., Broom et al., 1999) before the evolutionary history of this complex species assemblage can be resolved.

Compsopogonales and Erythropeltidales

In most taxonomic schemes (e.g., Bird and McLachlan, 1992; Womersley, 1994; Stegenga, Bolton, and Anderson, 1997), the order Compsopogonales (= Erythropeltidales sensu Garbary, Hansen, and Scagel, 1980) consists of three families: the Compsopogonaceae, Boldiaceae, and Erythrotrichiaceae (= Erythropeltidaceae; Womersley, 1994). The first two families are typically freshwater in their occurrence (Sheath, 1984), although Compsopogon coerulus (Balbis) Montagne has also been collected from brackish waters (Tomas et al., 1980). In contrast, the Erythrotrichiaceae are composed of strictly marine representatives (Womersley, 1994; Silva, Basson, and Moe, 1996). The Compsopogonales were erected based primarily on the mode of monosporangium formation (Garbary, Hansen, and Scagel, 1980). The sporangia develop from undifferentiated vegetative cells that undergo oblique divisions by a curved cell wall, producing two cells, the smaller of which develops into a monosporangium (Garbary, Hansen, and Scagel, 1980). However, Silva, Basson, and Moe (1996) suggested that the Compsopogonales should circumscribe only members of the Compsopogonaceae because their morphology is considerably more complex than that of the Erythrotrichiaceae. In addition, Silva, Basson, and Moe (1996) noted that the Erythropeltidales are a valid order if they contain the Erythrotichiaceae, exclusive of the Compsopogonaceae. In contrast to this view, Rintoul, Vis, and Sheath (1999) concluded that the Compsopogonales sensu lato, including the three families, is supported as a valid taxonomic entity based on phylogenetic analyses inferred from the SSU rDNA and Rubisco genes. The much larger data set used in the current study shows that the Erythrotrichiaceae are well separated from the Compsopogonaceae and Boldiaceae. Therefore, the recommendation of Silva, Basson, and Moe (1996) to recognize the Erythropeltidales, which contains the single family Erythrotrichiaceae, is substantiated by our analyses. Thus, Erythrotrichia, Erythrocladia, and Smithora should be considered to be members of the family Erythrotrichiaceae within the order Erythropeltidales. Other genera believed also to belong to this family include Porphyropsis, Membranella, Porphyrostromium, and Sahlingia (Garbary, Hansen, and Scagel, 1980).

The Compsopogonales, as resolved in our study, contain the freshwater families Boldiaceae and Compsopogonaceae, each of which constitutes a well-supported monophyletic group. In terms of genera contained within this order, only Boldia and Compsopogon appear to be valid. The identical sequences of Compsopogon coeruleus and specimens formerly classified as Compsopogonopsis leptocladus (Table 1) is in accord with Rintoul, Vis, and Sheath (1999), who concluded that the Compsopogonaceae contains the single species, Compsopogon coeruleus.

Rhodochaetales

This order has only one family, the Rhodochaetacae, and a single species, Rhodochaete parvula (Garbary, Hansen, and Scagel, 1980; Gabrielson et al., 1990). This species is thought to be central to understanding red algal evolution because of its combination of characteristics, including filamentous organization with apical growth, triphasic and slightly heteromorphic life history, and pit plugs with no cap layers or membranes (Magne, 1960; Boillot, 1978; Pueschel and Magne, 1987; Gabrielson et al., 1990; Garbary and Gabrielson, 1990). Van den Hoek, Mann, and Jahns (1995) postulated that the remaining bangiophyte orders may be derived from an ancestral Rhodochaete-like group and have secondarily lost more “advanced” characters (e.g., Porphyridiales) or have evolved additional advanced characters (e.g., Florideophycideae). The present study and that of Zuccarello et al. (2000) do not support this idea. Instead, Rhodochaete parvula groups with members of the Erythropeltidales in a strongly supported clade in our SSU rDNA trees. These data suggest that the Rhodochaetales is not well supported as being a separate order and that R. parvula may best be reclassified in the Erythropeltidales.

In a cladistic analysis of red algal orders, Gabrielson, Garbary, and Scagel (1985) noted that the order Rhodochaetales is distinguished from the Compsopogonales (including members of the Erythropeltidales) by the presence of apical growth and sexual reproduction. The authors suggested that these characters may have been lost in members of the Compsopogonales. Later, Garbary and Gabrielson (1990) presented the argument that, based on similarity of sexual reproduction with Porphyrostromium and pit plugs with Compsopogon, one could argue that R. parvula is most closely related to the Compsopogonales (including the Erythrotrichiaceae). In the present study, R. parvula appears to be basal to members of the Erythrotrichiaceae, substantiating the concept of a reduction in sexuality, apical growth, and pit plugs in this family (for details, see Zuccarello et al., 2000). Given the clear separation of the Erythrotrichiaceae and R. parvula in this clade, the family Rhodochaetaceae should be retained.

Porphyridiales

Resolving the taxonomic status of the order Porphyridiales is the most difficult issue within the Bangiophycidae because this order does not form a single monophyletic group. In fact, there appear to be at least three well-supported clades. The Porphyridiales has been separated into two families, the Porphyridiaceae and Phragmonemataceae. The first family has single stellate plastids with pyrenoids, whereas the taxa within the Phragmonemataceae do not have stellate plastids and pyrenoids are absent (Garbary, Hansen, and Scagel, 1980; Gabrielson et al., 1990). These families are not well supported in our study. In fact, both the Porphyridiaceae and Phragmonemataceae are polyphyletic within the Rhodophyta.

Porphyridiales 1 contain primarily unicells currently classified in the family Porphyridiaceae (Dixoniella and Rhodella). However, Dixoniella grisea is positioned on an extremely long branch within the distance analyses. Dixoniella and Rhodella differ in the thylakoid penetration and nuclear projection into the pyrenoid, Golgi association with other organelles, and presence of a peripheral thylakoid (Scott et al., 1992; Broadwater and Scott, 1994). These two genera are unique among the unicellular red algae that have been examined ultrastructurally because they have stellate-like plastids with deep lobing and excentric, unstalked pyrenoids (Broadwater and Scott, 1994). Hence, this grouping appears to represent a distinct taxonomic entity separated on the basis of plastid type, and the clade is well separated from the other red algal taxa currently classified in the Porphyridiales. Glaucosphaera vacuolata, which was positioned in this clade, is classified within the Glaucocystophyta (Kies and Kremer, 1986), but analyses of both nuclear- and plastid-encoded SSU rDNA sequences demonstrate that this taxon is not directly related to other glaucocystophytes and is a member of the Rhodophyta (Bhattacharya et al., 1995; Helmchen, Bhattacharya, and Melkonian, 1995). The suggestion, based on ultrastructural analyses, that G. vacuolata is a member of the Porphyridiales (McCracken, Nadakavukaren, and Cain, 1980; Broadwater et al., 1995) is substantiated by our analysis.

Porphyridiales 2 comprise unicellular (Rhodosorus) and multicellular (pseudofilamentous forms) taxa currently classified in the family Porphyridiaceae, with the exception of Goniotrichopsis sublittoralis (Phragmonemataceae; Garbary, Hansen, and Scagel, 1980; Gabrielson et al., 1990). The two species of Rhodosorus form a well-supported group in this large clade. They are unique among unicellular red algae in having a cup-shaped plastid without a peripheral thylakoid but with an eccentric pyrenoid (Broadwater and Scott, 1994). The pseudofilamentous genera Bangiopsis, Chroodactylon, Goniotrichopsis, and Stylonema are positioned on a weakly supported separate branch. All of these genera have a stellate plastid with a central pyrenoid, except for Goniotrichopsis, which has numerous discoid plastids without pyrenoids (Taylor, 1960; Abbott and Hollenberg, 1976; Womersley, 1994). Clearly, this diverse clade does not fit the currently conceived orders of the Bangiophycidae. The order Stylonematales and family Stylonemataceae can be reinstated to include Stylonema and Chroodactylon (Drew, 1956), as noted by Silva, Basson, and Moe (1996). According to Abbott and Hollenberg (1976), Goniotrichopsis belongs to the same family as Stylonema (as Goniotrichum) and Chroodactylon (as Asterocytis), based on having branched, filamentous thalli in which cells are separated by gelatinous material. The remaining taxon, Rhodosorus, appears to be well associated with the above grouping and can be considered to be a member of the Stylonematales. Nonetheless, a new family is warranted for this genus based on its unicellular habit, unique plastids, and occurrence on a well-supported and separate branch in the SSU rDNA gene trees. Within the cluster containing the pseudofilamentous forms, Bangiopsis subsimplex (Porphyridiaceae) and Chroodactylon ornatum (Porphyridiaceae) are unresolved with respect to the cluster containing Stylonema (Porphyridiaceae) and Goniotrichopsis (Phragmonemataceae). Using plastid SSU rDNA sequences, Oliveira and Bhattacharya (2000) also found a well-supported group composed of four species of the Porphyridiales, C. richteriana, C. ramosum, S. alsidii, and R. marinus.

The well-supported Porphyridiales 3 lineage consists solely of unicellular forms—two species of Porphyridium and one species of Flintiella. This cluster contains the nominate genus and thus the order Porphyridiales and family Porphyridiaceae must accompany this genus. Flintiella has a branched plastid with no obvious pyrenoid in contrast to Porphyridium, which contains a stellate plastid with a central pyrenoid (Broadwater and Scott, 1994). Flintiella has been classified in the Porphyridiaceae by Bourrelly (1985) but Gabrielson et al. (1990) and Garbary, Hansen, and Scagel (1980) have placed this genus in the Phragmonemataceae based on plastid type.

Cyanidiales

Regarding Cyanidium and Galdieria, Seckbach and Ott (1994) have proposed that these genera should be united in a separate order, the Cyanidiales, within the Bangiophycidae. Our analyses support this idea, although Cyanidium spp. SSU rDNA sequences are needed to determine whether the Cyanidiales are a monophyletic or a paraphyletic order (see Oliveira and Bhattacharya, 2000).

Florideophycidae

The Florideophycidae examined in this study form a well-supported clade in the SSU rDNA gene trees with the Bangiales on a basal branch of this clade. This result has strong to moderate bootstrap support (see Figs. 2 and 3). The Bangiales/Florideophycidae association is supported by the type of reproductive cells, or cysts, defined by Magne (1989), by the association of the Golgi apparatus with the mitochondrion in these groups (Garbary and Gabrielson, 1990), the presence of pit connections in the Florideophycidae and in the Conchocelis phase of the Bangiales, the presence of group I introns in the nuclear SSU rDNA (in the Bangiales and in the Hildenbrandiales from the Florideophyceae; Oliveira and Ragan, 1994), and the results of Rubisco (Freshwater et al., 1994) and plastid SSU rRNA sequence comparisons (Oliveira and Bhattacharya, 2000). Based on nuclear SSU rDNA analyses, Ragan et al. (1994) have shown that the Florideophycidae are monophyletic, with the Hildenbrandiales as the first divergence and the NAP complex (Nemaliales, Acrochaetiales, and Palmariales) as the second diverging lineage. This finding supports the conclusion of some researchers that the Rhodophyta should be classified into one class, the Rhodophyceae with no subclasses (e.g., Garbary and Gabrielson, 1990).

Secondary symbiotic origin of red algal plastids

The secondary origin of cryptophyte, haptophyte, and the heterokont plastids from a red algal endosymbiont has been previously documented on the basis of sequence comparisons and plastid gene order (Douglas et al., 1991; Bhattacharya and Medlin, 1995, 1998; Palmer and Delwiche, 1996; Daugbjerg and Andersen, 1997; Medlin et al., 1997; Douglas and Penny, 1999; Durnford et al., 1999; and see discussion in Oliveira and Bhattacharya, 2000). The Rubisco protein analyses support the origin of the heterokont plastid from a red alga in the Cyanidiales (Fig. 3), a result that is consistent with previous analyses of plastid coding regions (e.g., Bhattacharya and Medlin, 1995; Leblanc, Boyen, and Loiseaux-de Goër, 1995; Medlin et al., 1995, 1997; Oliveira and Bhattacharya, 2000). Our study provides, however, no clues to the bangiophyte source(s) of cryptophyte and haptophyte plastids. The position of the cryptophyte nucleomorphs as a separate lineage in Fig. 2 is consistent with the Rubisco analyses, which also show the plastids from these secondary symbionts to be an independent lineage within the Bangiophycidae (Fig. 3). The haptophyte plastids have moderate bootstrap support as a sister group of cryptophytes in the Rubisco tree. A rearranged tree in which this relationship is broken is, however, not significantly worse than that showing cryptophyte/haptophyte monophyly (Tree 3 in Table 2). These conflicting data may be explained by poor sampling of cryptophyte Rubisco sequences in our analysis. Limited sampling of cryptophyte and haptophyte plastid SSU rDNA, for example, also suggested monophyly of these organelles (compare Figs. 2 and 3 in Oliveira and Bhattacharya, 2000). Alternatively, our data may indicate that evolutionarily closely related bangiophytes may have given rise independently to the plastids in cryptophytes and haptophytes. We also find no support for the scenario that all bangiophyte-derived secondary plastids (and, thereby, their host cells) form a monophyletic group, an idea that was put forth largely on the basis of plastid characters, such as the existence of a plastid endoplasmic reticulum in these taxa (Cavalier-Smith, 1982, 2000). The rearranged Tree 6 that forced the monophyly of the cryptophyte, haptophyte, and heterokont plastids was significantly worse than the best MP Rubisco tree shown in Fig. 3 (P = 0.0038). This result agrees with analyses of plastid SSU rDNA sequences (Oliveira and Bhattacharya, 2000).

Consistent with the view of independent origins of the plastids in cryptophyte, haptophytes, and heterokonts is the separate evolutionary histories of the host cells containing these organelles. The cryptophytes, for example, are more closely related to glaucocystophytes than to the heterokonts or haptophytes (Bhattacharya and Medlin, 1995; Van de Peer et al., 1996; Gray et al., 1998). These results underscore the hypothesis that secondary symbiosis, and thereby the evolution of a plastid import system in each host lineage, may have evolved multiple independent times (McFadden, 1999). A broader sample of plastid genes needs to be studied to test the hypothesis of independent secondary plastid origins in cryptophytes, haptophytes, and heterokonts. In addition, more bangiophyte taxa need to be included in these analyses to identify possible plastid donors for the cryptophytes and haptophytes.

In summary, our expanded phylogenetic analyses help to improve the understanding of the evolutionary histories of the Bangiophycidae and the Florideophycidae. The subclass Bangiophycidae is found to be a paraphyletic assemblage that should be included in a single red algal class, the Rhodophyceae, which includes the Florideophycidae (Gabrielson, Garbary, and Scagel, 1985). In addition, we suggest that the cryptophyte, haptophyte, and heterokont plastids were derived from independent secondary endosymbiotic events involving members of the Bangiophycidae. This hypothesis is consistent with our present understanding of plastid and host phylogeny in these algal groups.

Table 1. Algae used in the present study

Table 2. Results of tests of tree congruence with the Kishino-Hasegawa test. The comparisons of the best maximum parsimony tree inferred from the Rubisco data set (see Fig. 3) with rearrangements 2–6 are shown

Figure 1

Phylogeny of nuclear SSU rDNA coding regions (1582 positions) inferred using the maximum likelihood (ML) method. This tree has been rooted on the branch leading to the chlorophyte rDNAs. The results of neighbor-joining bootstrap analyses (2000 replications) using LogDet and HKY-85 (Hasegawa, Kishino, and Yano, 1985) distances are shown above and below the internal nodes, respectively. Only bootstrap values ≥60% for nodes shared by the ML and either distance bootstrap tree are shown. The dashes identify nodes in the ML tree that are not resolved in the distance bootstrap analyses. Lines are used to show the position of bootstrap values when they do not fit on the branches. Sequences determined in this study have been underlined, and the GenBank accession numbers for all existing rDNAs are shown after the species names. The position and approximate branch length of Dixoniella grisea when it is included in the LogDet and distance analyses are shown with the arrow. Florideo. is Florideophycidae and Bangio. is Bangiophycidae

Figure 2

A maximum likelihood tree inferred from comparisons of concatenated nuclear and plastid SSU rDNA coding regions (2953 positions) showing the evolutionary relationships of Rhodophyta and the nucleomorph of cryptophytes. This tree has been rooted on the branch leading to the Mesostigma viride rDNA sequence. Bootstrap values (200 replications) from the ML analysis are shown above the branches on the left of the slash marks. The values shown to the right of the slash marks are from a distance (HKY-85; Hasegawa, Kishino, and Yano, 1985) bootstrap analysis (2000 replications). The bootstrap values (2000 replications) shown below the branches on the left of the slash marks are inferred from neighbor-joining using LogDet distances, whereas the values shown to the right of the slash marks are inferred from an unweighted maximum parsimony analysis. Only bootstrap values ≥60% for nodes shared by the bootstrap consensus trees inferred with the different phylogenetic methods are shown. The dashes identify nodes in the ML tree that are not resolved in the different bootstrap analyses. Lines are used to show the position of bootstrap values when they do not fit on the branches. GenBank accession numbers for all existing plastid SSU rDNAs are shown after the species names. Florideo. is Florideophycidae and Bangio. is Bangiophycidae

Figure 3

(A) An unweighted maximum parsimony tree inferred from comparisons of 494 amino acids of the Rubisco protein. This tree is rooted on the branch leading to the heterokont and Cyanidiales Rubisco sequences. Bootstrap values (2000 replications) from the parsimony analysis are shown above the branches on the left of the slash marks. The values shown to the right of these slash marks are from a protein distance (mean difference) bootstrap analysis (2000 replications). The values shown below the branches on the left of the slash marks are inferred from a protein quartet puzzling analysis (10 000 puzzling steps), whereas the values shown to the right of these slash marks are inferred from a DNA quartet puzzling analysis (10 000 puzzling steps) using the first and second positions of Rubisco codons (954 bp). Both quartet puzzling analyses used minimum evolution as the optimality criterion. Only bootstrap values ≥60% for nodes shared by the bootstrap consensus trees inferred with the different phylogenetic methods are shown. The different rearrangements of the Rubisco tree are shown with the arrows. GenBank accession numbers for all existing Rubisco sequences are shown after the species names