Molecular Data and the Evolutionary History of Dinoflagellates

Abstract

We have sequenced small-subunit (SSU) ribosomal RNA (rRNA) genes from 16 dinoflagellates, produced phylogenetic trees of the group containing 105 taxa, and combined small- and partial large-subunit (LSU) rRNA data to produce new phylogenetic trees. We compare phylogenetic trees based on dinoflagellate rRNA and protein genes with established hypotheses of dinoflagellate evolution based on morphological data. Protein-gene trees have too few species for meaningful in-group phylogenetic analyses, but provide important insights on the phylogenetic position of dinoflagellates as a whole, on the identity of their close relatives, and on specific questions of evolutionary history. Phylogenetic trees obtained from dinoflagellate SSU rRNA genes are generally poorly resolved, but include by far the most species and some well-supported clades. Combined analyses of SSU and LSU somewhat improve support for several nodes, but are still weakly resolved. All analyses agree on the placement of dinoflagellates with ciliates and apicomplexans (= Sporozoa) in a well-supported clade, the alveolates. The closest relatives to dinokaryotic dinoflagellates appear to be apicomplexans, Perkinsus, Parvilucifera, syndinians and Oxyrrhis. The position of Noctiluca scintillans is unstable, while Blastodiniales as currently circumscribed seems polyphyletic. The same is true for Gymnodiniales: all phylogenetic trees examined (SSU and LSU-based) suggest that thecal plates have been lost repeatedly during dinoflagellate evolution. It is unclear whether any gymnodinialean clades originated before the theca. Peridiniales appear to be a paraphyletic group from which other dinoflagellate orders like Prorocentrales, Dinophysiales, most Gymnodiniales, and possibly also Gonyaulacales originated. Dinophysiales and Suessiales are strongly supported holophyletic groups, as is Gonyaulacales, although with more modest support. Prorocentrales is a monophyletic group only in some LSU-based trees. Within Gonyaulacales, molecular data broadly agree with classificatory schemes based on morphology. Implications of this taxonomic scheme for the evolution of selected dinoflagellate features (the nucleus, mitosis, flagella and photosynthesis) are discussed.

2004 Elsevier GmbH. All rights reserved.

Keywords: Dinoflagellates; Ribosomal RNA; Chloroplast evolution; Molecular phytogeny; Theca; Photosynthesis

Introduction

The importance of dinoflagellates in aquatic communities is hard to overestimate. They are ubiquitous in marine and freshwater environments, where they constitute a large percentage of both the phytoplankton and the microzooplankton, and in benthic communities as interstitial flora and fauna or as symbionts in reefbuilding corals, other invertebrates and unicellular organisms (Taylor, 1987). Both ecto- and endoparasitic dinoflagellate species are also common, infecting hosts ranging from other protists like ciliates, radiolarians or even other dinoflagellates, to crustaceans, cnidarians, appendicularians, polychaetes, fish and many others (Cachon and Cachon, 1987). Many species of dinoflagellates are notorious for producing toxins that can cause human illness through shellfish or fish poisoning (Steidinger, 1993); dinoflagellates are the ultimate cause of diseases like diarrheic shellfish poisoning (DSP), neurotoxic shellfish poisoning (NSP), paralytic shellfish poisoning (PSP) and ciguatera. Some toxic dinoflagellates (as well as other protists) can also cause fish kills and mortality of other marine fauna (Steidinger, 1993).

One recent definition of dinoflagellates is found in Fensome et al. (1993, p. 3): they are “eukaryotic, primarily single-celled organisms in which the motile cell possesses two dissimilar flagella: a ribbon-like flagellum with multiple waves which beats to the cell’s left, and a more conventional flagellum with one or a few waves which beats posteriorly”. Taxonomic treatments of the group have traditionally been based on two sets of cytological characters. One is the presence of a dinokaryon, a uniquely modified nucleus that lacks nucleosomal histones and contains fibrillar chromosomes with a typical ultrastructure that remain condensed throughout the cell cycle, and that divides through a special type of closed mitosis with an extranuclear spindle (review in Dodge, 1987). Dinokarya are present in most dinoflagellates, but not in the parasitic order Syndiniales or in particular life stages of the Blastodiniales (also parasitic) and Noctilucales (Fensome et al., 1993). The other character, applied to dinokaryotic dinoflagellates, is the arrangement of cortical alveoli, flattened vesicles immediately underneath the plasma membrane that often contain cellulose thecal plates (in dinoflagellate literature cortical alveoli are generally referred to as amphiesmal vesicles, review in Netzel and Durr (1984)). In thecate orders (Gonyaulacales, Peridiniales, Dinophysiales, Prorocentrales), the theca is contained in relatively few alveoli with a pattern that can be determined relatively easily (thecal plate tabulation). Athecate taxa, however, (notably the order Gymnodiniales, but also Syndiniales, Noctilucales, etc.) often contain hundreds of alveoli, making it difficult to determine homologies and locational relationships. As a consequence, thecate taxa are much easier to classify than athecate ones.

Thecal plate patterns are also easier to determine in species that are easily found as motile stages, the cell type that typically displays this feature. However, these motile stages are often very short phases of dinoflagellate life cycles; some species are most often found as cysts (Suessiales, Thoracosphaerales, Phytodiniales, a few Gonyaulacales), plasmodia (many Syndiniales) or as strongly modified trophonts that are not easily comparable to the typical dinoflagellate motile stages (Noctilucales or Blastodiniales). The tabulation of the motile stages is often reflected in cysts, a feature that has been used extensively to detect relationships between extant and fossil genera (Fensome et al., 1993; Fensome et al., 1999; most fossil dinoflagellates are cysts). Within some thecate orders, a (putative) radiation of forms can be followed remarkably well using extant species (e.g. in Dinophysiales, Gonyaulacales and Peridiniales, Taylor, 1980), even if the direction of the changes cannot. Nevertheless, this cannot be done between orders, for there appear to be few intermediate forms. As a consequence, except for some cases where informative intermediate fossil taxa have been found (Fensome et al., 1993), the mutual relationship of many dinoflagellate orders is still unclear. Also unclear is which groups of dinoflagellates are early or late diverging; different sets of characters support different hypotheses (discussion in Taylor, 1980; Fensome et al., 1993).

Early phylogenetic studies showed the monophyly of dinoflagellates (Maroteaux et al., 1985; Herzog and Maroteaux, 1986) and disproved notions that dinoflagellates are early branches of the eukaryote tree (the mesokaryotic theory, Dodge, 1965, 1966). A relationship between dinoflagellates and ciliates that had been postulated earlier (Corliss, 1975; Taylor, 1976) was also corroborated by these sequences, as was a newly discovered one to apicomplexans (Wolters, 1991; Gajadhar et al., 1991). In 1991 a new taxon, the Alveolata, was created encompasing ciliates, dinoflagellates, apicomplexans and their close relatives, the protalveolates (Cavalier-Smith, 1991), and numerous studies have repeatedly supported its validity (e.g. Cavalier-Smith, 1993; Van de Peer et al., 1996; Fast et al., 2002). The relationship of alveolates to other groups has been more difficult to resolve, but recent studies based on phylogenies of concatenated proteins and chloroplast-targeted genes (Baldauf et al., 2000; Fast et al., 2001) have supported the relationship between this group and chromists as predicted by the chromalveolate hypothesis (Cavalier-Smith, 1999, 2003) and by earlier taxonomic schemes (e.g. Taylor, 1976).

Within alveolates, dinoflagellates are more closely related to the apicomplexans than to the ciliates (Fast et al., 2002). Other close relatives of dinoflagellates include forms that share a number of features typical of all alveolates (e.g. cortical alveoli, mitochondria with tubular cristae, presence of trichocysts in diverse forms), but lack the synapomorphies that define ciliates, dinoflagellates or apicomplexans, the so-called protalveolates (Cavalier-Smith, 1991, 1993). The genus Perkinsus, for example, a parasite of oysters and other bivalves, and Parvilucifera, a parasite infecting dinoflagellates, often form a clade closely related to dinoflagellates (Siddall et al., 1997; Noren et al., 1999; Saldarriaga et al., 2003a); the genus Rastrimonas (formerly Cryptophagus, Brugerolle, 2003), a parasite of cryptomonads, could be a third member of this group (Brugerolle, 2002b). Other protalveolate taxa that seem to have close links to the dinoflagellates are the free-living genus Oxyrrhis, recently excluded from the group (Fensome et al., 1993), and the ellobiopsids, a group of parasites of crustaceans that are either derived from or very closely related to dinoflagellates (J. Silbermann, personal communication). The genus Colpodella, however, appears to be a basal branc\h to the apicomplexans (Cavalier-Smith, 2000; Brugerolle, 2002a; Kuvardina et al., 2002; Leander and Keeling, 2003). Other protalveolates have not been characterized at the molecular level, and so it remains to be seen where the phylogenetic affiliation of Colponema, Acrocoelus, and others may lie.

Nearly all molecular phylogenetic studies of the ingroup relationships of dinoflagellates have used rRNA, either partial sequences of the large-subunit (LSU) ribosomal RNA (rRNA) gene (LSU, e.g. Lenaers et al., 1991; Zardoya et al., 1995; Daugbjerg et al., 2000), or the small-subunit (SSU) rRNA gene (SSU, e.g. Saunders et al., 1997; Grzebyk et al., 1998; Gunderson et al., 1999; Saldarriaga et al., 2001). Phylogenies based on SSU are the only ones with data for phylogenetically important groups like the Syndiniales, Noctilucales or Blastodiniales. Relationships of orders to one- another are mostly unresolved (e.g. Saunders et al., 1997; Saldarriaga et al., 2001), but those at the base of the lineage are often well supported, as are some late-branching groups. Phylogenies based on the first two or three domains (D1-D3) of the LSU contain fewer taxa than SSU-based ones, but since the two molecules appear to evolve at different rates (Ben Ali et al., 2001; John et al., 2003) they have also proven very valuable since bootstrap support for certain groupings is greater. Protein-gene based phylogenies are still scarce, e.g. HSP90, actin, alpha- and beta-tubulin genes (Saldarriaga et al., 2003a; B. Leander, unpublished data), and plastid-encoded genes (e.g. psbA in Takishita and Uchida (1999), psaA in Yoon et al. (2002), Zhang et al. (2000)). None of these yet contain many taxa, and support for their in-group clades tends to be weak. Nevertheless, they have proven valuable for determining the position of some basal taxa (e.g. Saldarriaga et al., 2003a).

The objective of the present work is to clarify some key events in the evolutionary history of the dinoflagellates and their close relatives. We re-examine all available data, molecular, morphological, paleontological and biochemical, to produce a phylogenetic framework for the group, with special consideration given to rRNA trees. We explore the origin of the dinokaryon and of dinokaryotic dinoflagellates, the development of cortical alveoli in the group and the history of dinoflagellate photosynthesis. We also examine some smaller scale questions, e.g. the relationship of Perkinsua, Oxyrrhis, Noctiluca, Syndiniales and Blastodiniales to other dinoflagellates and the circumscription of the different groups of Gymnodiniales. Lastly, we consider whether the phylogeny of Gonyaulacales, generally better supported than other parts of the tree, is congruent with the proposals for dinoflagellate classification based on morphology put forward by Fensome et al. (1993).

Materials and methods

Organisms, DNA extraction, amplification and sequencing

Photosynthetic dinoflagellate species were obtained from non- axenic culture collections (Table 1) and cultured according to established protocols (e.g. Harrison et al., 1980). The heterotrophic Protoperidinium species were fed the diatom Ditylum brightwellii and maintained at 12C in F/2 medium at 30 mols photons m^sup -2^ s^sup -1^ on a plankton wheel at 1 rpm. Cells were harvested by centrifugation. DNA was extracted using the DNeasy Plant DNA Purification Kit (Qiagen). Whenever possible, the 18S (nuclear SSU) rRNA gene was amplified as a single fragment using a polymerase chain reaction with two eukaryotic universal SSU primers (5′-CGAATTCAACCTGGTTGATCCTGCCAGT-3′ and 5′-CCGGATCCTGATCCTTCTGCA GGTTCACCTAC-3′). However, in many cases two overlapping fragments had to be produced using internal primers designed to match existing eukaryotic SSU sequences (4F: 5′-CGGAATTCCAGTC-3′ and 11R: 5′- GGATCACAGCTG-3′). PCR products were either sequenced directly or cloned into pCR-2.1 vector using the TOPO TA cloning kit (Invitrogen). Sequencing reactions were completed with both the original PCR primers as well as 2-3 additional primers in each direction. When using cloned fragments, 2-4 clones were sequenced to detect and clarify possible ambiguities.

Phylogenetic analysis

New sequences were added to the SSU alignment of Saldarriaga et al. (2001); they are now available from GenBank. The final multiple alignment contained 98 dinoflagellate species, plus Perkinsus, Parvilucifera and several ciliate and apicomplexan sequences that were used as outgroups. The sequence for Oxyrrhis marina was excluded from the analyses: previous experience showed that this species has an extremely derived SSU sequence that distorts the topologies of SSU trees (Saldarriaga et al., 2003a). We also included seven sequences obtained from environmental samples of marine picoplankton by Lopez-Garcia et al. (2001) (GenBank accessions AF290066, AF290068, AF290077 and AF290078) and Moon-van der Staay et al. (2001) (GenBank accessions AJ402326, AJ402330 and AJ402354). Only unambiguously aligned sections of the molecule (1479 characters) were used in the phylogenetic analyses. A second set of analyses of SSU data was performed excluding all ciliate and apicomplexan taxa (Perkinsus was used as the outgroup); by doing this we were able to align confidently a significantly larger portion of the SSU molecule (1649 sites).

Table 1. List of strains examined in this study and GenBank Accession Numbers for their nuclear SSU rRNA sequences

SSU sequences were also concatenated with published sequences for sections of the LSU rRNA gene (LSU). Concatenated alignments that included SSU and domains D1-D3 of the LSU included 25 alveolate species (22 of them dinoflagellates) and 2418 nucleotides, while alignments with SSU and domains D1-D2 of the LSU included 34 species (31 of them dinoflagellates) and 2100 nucleotides. Phylogenetic trees based on LSU only were also calculated for comparison; in them the choice of sites was extremely conservative, only 447 sites for alignments of domains D1-D2, 718 sites for those of domains D1-D3.

Distances were calculated with PUZZLE 5.0. (Strimmer and von Haeseler, 1996) using the HKY substitution frequency matrix. Nucleotide frequencies and transition/transversion ratios were estimated from the data, and site-to-site variation was modeled by a gamma distribution with invariable sites plus 8 variable rate categories and the shape parameter alpha estimated from the data. Distance trees were constructed using BioNJ (Gascuel, 1997), Weighbor (Bruno et al., 2000) and Fitch-Margoliash (Felsenstein, 1993). One hundred bootstrap data sets were made using SEQBOOT and trees inferred as described for parsimony and corrected distances, where distances were calculated using puzzle-boot (by M. Holder and A. Roger) with the alpha shape parameter, nucleotide frequencies and transition/transversion ratio from the initial tree enforced on the 100 replicates. Maximum likelihood trees were calculated for the concatenated SSU/LSU (D1-D2) datasets and for a heavily reduced alignment of SSU sequences (40 species; 35 dinoflagellates). They were inferred under an HKY model incorporating a discrete gamma distribution (invariable sites and 8 variable rate categories; shape parameter, nucleotide frequencies and transition/transversion ratio estimated from the data, 5 jumbles, PAUP 4.0, Swofford, 1999). Maximum likelihood trees were also calculated from the 100 bootstrap data sets in the case of the concatenated data.

Results

SSU rRNA phylogeny

It is unknown whether the sequences from the environmental samples from Lopez-Garcia et al. (2001) and Moon-van der Staay et al. (2001) come from organisms that would be called dinoflagellates based on morphology, and for that reason it is very difficult to say whether the dinoflagellate clade as a whole was supported in our trees or not. Those environmental sequences, however, always grouped in two clades. One of them (group II in Lopez-Garcia et al., 2001) generally also included all known sequences of Syndiniales (Hematodinium and 3 species of Amoebophrya; in the Filch tree Hematodinium was outside of the group). The other clade (group I in Lopez-Garcia et al., 2001) included only environmental sequences, and in the BioNJ (Fig. 1) and weighbour trees branched basal to all other dinoflagellates but not to the Perkinsus/Parvilucifera grouping (in the Fitch tree this clade branched after the Syndinians and Noctiluca). If one assumes that all these environmental sequences come from true dinoflagellates, then the dinoflagellate clade is supported in all trees by bootstraps of 60-65%.

Placement of Noctiluca in all trees was very unstable. In BioNJ and Weighbor trees, it branched with negligible support at the base of all established dinoflagellates (including syndinians but not the members of the group I clade). Interestingly, SSU trees including only dinoflagellates and Perkinsus that utilized more sites invariably placed Noctiluca scintillons within the GPP complex (Fig. 2). All other dinoflagellates, including two species considered members of the Blastodiniales in Fensome et al. (1993) (Amyloodinium sp. and Haplozoon axiothellae), form a single clade in all trees examined. A large part of that clade is composed of very short- branched members of the orders Gymnodiniales, Peridiniales, Prorocentrales and Dinophysiales, the so-called GPP complex (Saunders et al., 1997), along with Thoracosphaera (Thoracosphaerales), Hemidinium (Phytodiniales), Amyloodinium, Haplozoon and the parasitic genus Pfiesteria. Only the order Dinophysiales, represented only by the genus Dinophyxix, groups strongly as a distinct clade within the GPP complex (Edvardsen et al., 2003). The Prorocentrales break into two groups, one containing benthic species (Prorocentrum lima, P. concavum), the other more planktonic species (P. micans, P. gracile, P. minimum, Grzebyk et al., 1998). The Gymnodiniales scatter throughout the tree, forming at lea\st five major subgroups. One, composed of several (but not all) species of the genus Amphidinium, lacks the characteristic short branches of the GPP complex and generally groups close to Gonyaulacales. A second group of Gymnodiniales always groups strongly with the only two extant genera of the order Suessiales, Symbiodinium and Polarella (bootstrap supports 97-99%). The last three strongly supported gymnodinialean clades are bona fide members of the GPP complex: one includes the type species of Gymnodinium (G. fuscum) and close relatives (including Lepidodinium viride); the second, members of Karenia and Karlodinium but also Amphidlnium herdmanii; and the third, three putative members of the genus Gyrodinium (G. instriatum and G. dor sum have identical SSU sequences that differ from that of G. uncatenum by only 3 nucleotides out of 1755). The sequences for Amphidinium cf. operculatum, Amphidinium massartii and “Amphidinium rhynchocephalum” are also identical, differing by 8 nucleotides (from a total of 1752) from that of A. carterae.

In some trees (e.g. Fitch), the majority of Peridiniales form a clade, albeit very weakly supported and interrupted by Haplozoon axiothelkie. It includes all members of Heterocapsa, Scrippsiella and Pcntapharsodinium, plus Lessardia, Roscoffia, Peridiniopsis, and two species of Peridinium, P. umbonatum and P. wierzejskii (in Weighbor and BioNJ trees this clade is interrupted by gymnodinialean and/or prorocentralean groups; Thecadinium dragescoi is probably a misnamed member of the peridinialean genus Amphidiniopsis, M. Hoppenrath, pers. comm.). Nevertheless, several peridinialean taxa never group with the bulk of the order. These include a well- supported clade of the three Proloperidinium species and a well- supported grouping of three Peridinium species (Peridinium sp., P. willei and P. bipes) that sometimes includes Glenodiniopxix xteinii. The diatom-bearing genera Kryptoperidinium and Durinskia form a weakly supported clade in BioNJ trees, as do Pfiesteria and the putatively blastodinialean Amyloodinium in the Fitch and Weighbor trees. None of these groupings ever branch with the bulk of the Peridiniales.

The Gonyaulacales generally have longer branches than other dinoflagellates (only Syndiniales, Haplozoon, Protoperidinium and the A. carterae clade have comparably long branches). They tend to form a clade to the exclusion of almost all other dinoflagellates (e.g. in the Fitch and BioNJ trees), although it is never well supported. The phytodinialean genus Halostylodinium consistently branches within the clade. Within the Gonyaulacales (Table 2), several groupings appear consistently, e.g. one containing Ostreopsis, Alexandrium, Fragilidium, Pyrophacus, Pyrodinium and Pyrocyslis (suborder Goniodominae, 50-60% bootstrap support) and another containing all Ceralium species (Ceratiineae, 75-90% bootstrap support). Members of the Gonyaulacineae (Proloceralium, Lingulodinium, Gonyaulax, Amylax and Ceratocoryx) consistently branch at the base of the Gonyaulacales, as a paraphyletic group that gives rise to the Ceratiineae and Goniodominae and that also contains Crypthecodinium and Halostylodinium.

LSU rRNA phytogeny

Phylogenetic trees based on LSU data were similar to those based on SSU. As LSU sequences for Perkinsus, Oxyrrhis, Syndiniales, Noctilucales or Blastodiniales are unavailable, the trees consisted of a large, badly resolved group of very short-branched taxa (the GPP complex, Gymnodiniales, Peridiniales, Prorocentrales and Dinophysiales) and a monophyletic grouping of longer-branched members of the order Gonyaulacales (Fig. 3). Within the GPP complex, groupings well supported in SSU trees are also well supported here, e.g. the Gymnodinium fuscum group (henceforth Gymnodinium sensu stricto, Daugbjerg et al. (2000)), the Karenial Karlodinium group, Suessiales (including several Gymnodinium species), and Dinophysiales. There are, however, several differences from SSU trees. In at least some LSU trees, all Prorocentrales do group together (e.g. in the Weighbor tree, Fig. 3), and while the A. carterae group still holds together with good support and a relatively long branch, it is not at the base of the Gonyaulacales (in Weighbor and Fitch trees it interrupts a badly supported clade of Peridiniales). The position of Woloszynskia is also different in SSU and LSU trees: while in LSU it branches with the Suessiales with 95-97% bootstrap support, in SSU its position is very unstable (the two alignments contain different species of the genus: W. pseudopalustris in LSU, W. leopoliensis in SSU).

Fig. 1. Phylogenetic tree constructed with BioNJ from a gamma- corrected distance matrix of SSU rRNA sequences (1479 nucleotides) from 117 species of alveolates, including 98 dinoflagellates and 7 undcscribed species from environmental samples identified by their GenBank accession numbers (Lopez-Garcia et al., 2001; Moon-van der Stay et al., 2001). Bootstrap values are shown when larger than 60%. Problematic names of taxa are given in quotation marks.

Fig. 2. Phylogenetic tree constructed with weighted neighbour joining from a gamma-corrected distance matrix of SSU rRNA sequences (1649 nucleotides) from 98 dinoflagellates, 7 undescribed species from environmental samples identified by their GenBank accession numbers, and Perkinsus marinus, used as the outgroup. Bootstrap values are shown when larger than 60%. Problematic names of taxa are given in quotation marks.

Table 2. Classification of the order Gonyaulacales according to Fensome et al. (1993), including only species for which SSU sequence data are available

Gonyaulacales also form a clade in most LSU trees, albeit with modest bootstrap support (in the ML tree, the genus Ceratium branches together with the Apicomplexan outgroup). The majority of the gonyaulacalean species with LSU data are members of the Goniodominae (Alexandrium, Fragilidium, Coolia and Ostreopsis), and they form a clade excluding all other taxa, with low bootstrap support (the sequence for Ceratium furca interrupts a very strongly supported clade of many Alexandrium species in all trees; we suspect this to be an error). The other Ceratium sequences, as well as those for Protoceratium and Gonyaulax, often make a paraphyletic group at the base of the Gonyaulacales that gives rise to the Goniodominae (not in the Fitch trees, where Goniodominae appear to give rise to Gonyaulacinae and Ceratium). Protoceratium and Gonyaulax were never sisters.

Combined rRNA phylogeny

Phylogenetic trees based on combined datasets (Fig. 4) generally show the basic structure discussed above: a badly supported backbone of short-branched taxa (the GPP complex) that includes some very well-supported subgroups, and the Gonyaulacales, longer-branched taxa that invariably form a clade, here very well supported (80- 100% bootstrap support). The well-supported groups in the GPP complex are identical to those discussed above, but their relative order is variable. Prorocentrales never group together, forming the same two clades as in SSU trees. Within Gonyaulacales, the Gonyaulacinae (Gonyaulax and Protoceratium) generally branch as sisters to a group that contains Ceratium and the Goniodominae (in the Fitch and Weighbor trees based on SSU/D1/D2/D3 concatenations the Gonyaulax/Protoceratium clade is not retained). One major difference between the concatenated and single gene trees is that in all concatenated trees the two Heterocapsa. species included are sisters to the bulk of the GPP complex with bootstrap support between 43% and 71%. In the Weighbor trees and in the Fitch and ML trees based on the SSU/D1/D2/D3 concatenation these two species make a clade, in the other trees they do not. Many nodes have better bootstrap support than in the single-gene trees. It is unclear whether this is a consequence of fewer taxa or the longer sequence.

Fig. 3. Phylogenetic tree constructed with weighted neighbor joining from a gamma-corrected distance matrix of domains D1 and D2 of the LSU rRNA gene (447 nucleotides) from 71 alveolates, 68 of them dinoflagellates. Bootstrap values shown when larger than 60%.

Fig. 4. Maximum likelihood phylogenetic tree constructed from concatenated LSU (domains D1 and D2) and SSU rRNA sequences (2100 nucleotides) from 34 alveolates, 31 of them dinoflagellates. Transition/transversion ratio 2.13; other trees found with slightly lower log likelihoods differed only in minor details. Bootstrap values are shown when higher than 60%.

Discussion

Rates of evolution, the structure of dinoflagellate phylogenetic trees and the mesozoic radiation of dinoflagellates

There is a striking asymmetry of evolutionary rates in the ribosomal genes of dinoflagellates, more pronounced in the SSU genes but also present in the domains of the LSU investigated here. As a consequence, both SSU- and LSU-based phylogenetic trees present a very characteristic structure: a large group of very short-branched GPP species, and a clade with medium- to very long branches. As far as these two groupings are concerned, the differences in evolutionary rate are certainly correlated with the phylogenetic history of the group: one clade of medium to long-branched species is composed exclusively of taxa classified in the order Gonyaulacales, and there are typically no Gonyaulacales elsewhere. Nevertheless, other species also have divergent sequences, including Oxyrrhis, Haplozoon, Proloperidinium and Amoebophrya in SSU trees and the A. carterae clade in both SSU and LSU. The fact that, with the exception of Oxyrrhis (Saldarriaga et al., 2003a) and in a few trees the A. carterae clade these long-branched taxa do not generally intrude into the Gonyaulacales is a sign that the grouping may reflect a real phylogenetic signal rather than long-branch attraction. It is interesting that no protein sequences known from the gonyaulacalean Crypthecodinium cohnii are particularly divergent; the asymm\etry of evolutionary rates in ribosomal genes of dinoflagellates may not extend to protein genes.

The “backbone” of all dinoflagellate rRNA trees is very weakly supported. This is consistent with a rapid dinoflagellate radiation into the major forms we see today. The palaeontological record gives a very similar picture (Fensome et al., 1999): although putative dinoflagellate fossils in the form of acritarchs and biogeochemical traces exist from as early as the Cambrian (e.g. Moldowan and Talyzina, 1998, discussion in Fensome et al., 1999), undisputed dinoflagellates appear for the first time in the early Mesozoic, and by the mid-Jurassic practically all variations of at least gonyaulacalean and peridinialean forms were already present. Nevertheless, the fossil record consists almost exclusively of groups that produce fossilizable cysts (ca. 15% of extant species of dinoflagellates, Head, 1996), other groups are very badly represented, making it unclear whether the rapid increase in gonyaulacalean and peridinialean morphological types in the early Jurassic was caused by a radiation of the whole group. The congruence between the patterns suggested by the fossil record and the rRNA trees, which include non-cyst-formers, implies a general radiation that included athecate forms.

A phylogenetic framework for understanding dinoflagellate evolution

Fig. 5 shows a hypothesis on the evolutionary history of dinoflagellates and their close relatives. It is based on features of molecular trees that are well supported and/ or congruent with one-another and on morphological and palaeontological information.

Perkinsus, Oxyrrhis and the Syndiniales

The relative positions of Perkinsus, the dinokaryotic dinoflagellates and the apicomplexans derived from molecular data are well supported by many different genes coding both for rRNAs and for proteins (e.g. Reece et al., 1997; Fast et al., 2001; Saldarriaga et al., 2003a). The relationship between Colpodella and the apicomplexans is based entirely on SSU rRNA phylogenies, it is recovered consistently and correlates well with morphological data (Kuvardina et al., 2002; Leander et al., 2003).

The phylogenetic position of Oxyrrhis has been more problematic. Phylogenies based on the SSU gene place it within the dinokaryotic dinoflagellates, with 76-81% bootstrap support for a clade of Oxyrrhis and Gonyaulax spinifera (Saldarriaga et al., 2003a). Protein-gene data give a very different result: all protein-gene phylogenies investigated to date (actin, alpha- and beta-tubulin in Saldarriaga et al., 2003a, also HSP90, B. Leander, personal communication) place Oxyrrhis at the base of the dinoflagellates. Because of the extreme divergence of the Oxyrrhis SSU sequence and the congruence amongst the protein-gene phylogenies (and also a short LSU fragment, Lenaers et al., 1991), it is likely that Oxyrrhis is a sister to the dinokaryotic dinoflagellates, but this will only be established by improved sampling of protein-coding genes from dinoflagellates. In any case, Oxyrrhis seems to have diverged later than Perkinsus; like dinokaryotes it lacks a curved ribbon at the apical end (regarded by some authors as a homologue to the apicomplexan conoid) that is typical of several protalveolates (e.g. Colpodella, Perkinsus, Parvilucifera and Rastrimonas) and the micronemes shared by them and apicomplexans.

Mitosis in Perkinsus is apparently similar to that in syndinians and dinokaryotic dinoflagellates in that channels containing an extranuclear spindle traverse the nucleus (Perkins, 1996), while Oxyrrhis has an intranuclear spindle (Triemer, 1982). Postulating that Oxyrrhis originated later than Perkinsus assumes a reversal in the organisation of the mitotic apparatus of Oxyrrhis (ciliates, like Oxyrrhis, have a closed mitosis with an internal spindle, apicomplexans a semiopen one). Nevertheless, other features strongly ally Oxyrrhis to the dinoflagellates, notably the lack of most real histones (Li, 1984) and several flagellar features (see discussion below). It is not known whether Perkinsus has histones, but electron microscopy shows nuclei with decondensed chromatin during the majority of their life cycle (Perkins, 1996), very different from those of either Oxyrrhis or the dinokaryotic dinoflagellates and similar to those of most other eukaryotes. All in all, because of the strength of the molecular data and the unlikelihood that histones were lost more than once, it seems clear that Oxyrrhis branches between Perkinsus and the dinokaryotic dinoflagellates, but that assessment could change if Perkinsus unexpectedly proved to lack histories.

Fig. 5. Hypothesis on the evolutionary history of dinoflagellates and their close relatives based on the features of molecular trees that are well supported and/or congruent with one-another and on morphological and palacontological information.

The order Syndiniales is very heterogeneous (e.g. Fensome et al., 1993). The only molecular data available for the group are SSU sequences from Hematodinium and Amoebophrya (different families), but no data are yet available for the morphologically most aberrant family of the order, the Duboscquellaceae. Hematodinium and Amoebophrya generally form a clade in phylogenetic trees (always with weak bootstrap), but in a few analyses they separate. Interestingly, the diversity of the order might be underestimated: many SSU sequences obtained from picoplanctonic environmental samples cluster with very high bootstrap support (up to 99%) around Amoebophrya. When these sequences were first presented (Moon-van der Staay et al., 2001; Lopez-Garcia et al., 2001), it could not be stated categorically that those sequences were actually from syndinians. The addition of Hematodinium to the data set greatly strengthens that assumption, as this syndinian branches at the base of the clade that contains several strains of Amoebophrya (another syndinian) and the picoplanctonic taxa. The monophyly of the order Syndiniales is controversial, and so it will be interesting to see whether Syndinium and the Dubosquellaceae also branch in this clade. It is also interesting to speculate whether those sequences from picoplanktonic cells represent free-living organisms (there are no named free-living syndinians) or the infective stages of parasitic forms. A second group of environmental marine sequences forms a well- resolved clade that is not closely related to any named alveolates. Since there is no morphological information for it, it is impossible to say whether the sequences are syndinian (or indeed dinoflagellate) or not. They always branch after Perkinsus, and so we consider them members of the dinoflagellate lineage, but little can be inferred about dinoflagellate evolution from them until their morphology is characterized.

The relative positions of the syndinians and Oxyrrhis cannot be determined from the available molecular data alone: only SSU sequences are known for syndinians, and the Oxyrrhis sequence for that gene is misleading (Saldarriaga et al., 2003a). However, syndinians and Perkinsus share an invagination of the nuclear membrane in interphase that houses centrioles (Ris and Kubai, 1974; Perkins, 1996) that does not occur in dinokaryotic dinoflagellates or in Oxyrrhis. For this reason, we weakly favour a topology where syndinians are sisters to a clade comprising Oxyrrhis and the dinokaryotic dinoflagellates. Nevertheless, more data is needed to confirm this.

Noctilucales and Blastodiniales

In the most recent general classification of dinoflagellates (Fensome et al., 1993) Noctilucales and Blastodiniales are basal classes of their own within the subdivision of dinokaryotic dinoflagellates. This is because members of both orders have non- dinokaryotic life stages: the trophonts of Noctiluca, Blastodinium, Amyloodinium and many others have nuclei that lack the typical fibrillar chromosomes of dinokaryotic dinoflagellates and stain brightly with alkali fast green, a chemical reagent that colors basic proteins (histones of typical eukaryotic nuclei are easily stained by it, dinokarya are not). Nevertheless, these species have life stages with real dinokarya: at certain phases of their life cycle trophonts start a series of divisions that produce ever smaller nuclei with chromosomes that gradually condense to produce the typical dinokarya (Soyer, 1969, 1971, 1972). The dinokaryotic cells thus produced have the typical appearance of biflagellate dinoflagellates.

Molecular sequences (only SSU) exist for three taxa of either Noctilucales or Blastodiniales: Noctiluca, Ainyloodinium and Haplozoon. Noctiluca branches basal to the dinokaryotic dinoflagellates (and usually also the syndinians) in many phylogenetic trees, but never with high bootstrap support (e.g. Fig. 1). However, in analyses with few outgroups and more aligned sites Noctiluca branches from within the GPP complex (Fig. 2). Moreover, Noctiluca chromatin may be more similar to that of dinokaryoles than to typical eukaryotes: electrophoretic gels of nuclear basic proteins extracted from the Noctiluca trophont produce a band pattern consistent with that of completely dinokaryotic dinoflagellates, not with eukaryotic, histone-containing nuclei (Li, 1984). In other words, Noctiluca may well lack typical core histones throughout its life cycle, suggesting that the alkali fast green stains other non-core-histone proteins in the trophont nucleus. As a consequence, the basal position of Noctiluca within the dinokaryotic dinoflagellates should be reexamined: the two main arguments for proposing such a basal position have been shown to be either very weak (SSU-based phylogenetic analyses), or probably wrong (the ostensible presence of histones in the nuclei of feeding stages). The fact that the vegetative stages (trophonts) of other Noctilucales, genera like Leptodiscus, Craspedotella, Petalodinium, Kofoidinium and Spatulodinium, appear to have typicaldinokarya (M. Elbrachter, personal communication) further strengthens this view. Three morphological features of Noctiluca and other Noctilucales argue for a relationship of the order to at least some groups of gymnodinialean dinoflagellates: young trophonts and/or dinospores of several of the less morphologically derived noctilucalean taxa (e.g. Kofoidinium and Spatulodinium) are practically indistinguishable from a number of athecate dinoflagellate genera, especially Amphidinium (Cachon and Cachon, 1968). More importantly, Noctiluca shares with members of the genus Gymnodinium senso stricto (Daugbjerg et al., 2000) two rare morphological features. First, Gymnodinium, like the gametes of Noctiluca, lacks a transverse striated flagellar root, a feature typical of most dinoflagellates (Hansen et al., 2000). Furthermore, the nuclear envelope of many (but not all) Gymnodinium (as well as Polykrikos) and the trophonl of Noctiluca have peculiar chambers (ampullae) in which the nuclear pores are situated (Afzelius, 1963; Soyer, 1969; Dodge and Crawford, 1969; Daugbjerg et al., 2000). These chambers disappear in Noctiluca as the dinospores are formed (Soyer, 1972), and so they may not be homologous to the ones in Gymnodinium, but if they are homologues they would provide an important morphological connection between the two groups. It is unknown whether other Noctilucales have ampullae around the nucleus.

The order Blastodiniales is almost certainly polyphyletic (e.g. Chatton, 1920; Fensome et al., 1993). Amyloodinium and Haplozoon never branch together in our trees, although both are typically members of the GPP complex (in some trees Amyloodinium may branch at the base of the dinokaryotic dinoflagellates as a whole, e.g. Fig. 1, but see also Fig. 2). Furthermore, although several members of the order have, like Noctiluca, non-dinokaryotic nuclei in some life stages (e.g. Blastodinium, Amyloodinium, Oodinium, Caryotoma and Crepidoodinium: Soyer, 1971; Lom and Lawler, 1973; Cachon and Cachon, 1977; Hollande and Corbel, 1982; Lorn et al., 1993), others do not: Dissodinium and Protoodinium are purely dinokaryotic (Cachon and Cachon, 1971; Drebes, 1981), as are probably Piscinoodinium and Haplozoon (trophonts in these two genera are dinokaryotic (Lorn and Schubert, 1983; Siebert and West, 1974), and dinospores have never been shown to have anything other than a dinokaryon in dinokaryotic dinoflagellates). Other genera are understudied, e.g. Apodinium, Cachonella, Sphaeripara; the true phylogenetic affinities of these taxa are unclear. The derived position of Amyloodinium in SSU trees is strongly supported by morphology: Amyloodinium (and also Pfiesteria, its sister taxon in most trees) has dinospores with a thecal plate pattern like that of Peridiniales (Landsberg et al., 1994; Steidinger et al., 1996; Fensome et al., 1999). Could other Blastodiniales also be Peridiniales? The trophont of Oodinium fritillariae has thecal plates like Amyloodinium, but also has ampullae around the nucleus (Cachon and Cachon, 1977), the Protoodinium trophont even has peridinialean tabulation (Cachon and Cachon, 1971). Other Blastodiniales share more similarities with athecate dinoflagellates: Haplozoon, Crepidoodinium and probably also Piscinoodinium have many polygonal alveoli in surface view (Lorn, 1981; Lorn and Schubert, 1983; Lorn et al., 1993; Leander et al., 2002). It is important to keep in mind, however, that many features known for Blastodiniales (e.g. the small, polygonal alveoli) have been observed in their trophonts, an often heavily modified life stage. The morphology of their dinospores will surely be much more helpful in determining their true phylogenetic affinities, as shown by the example of Amyloodinium ocellatum.

Gymnodiniales, Suessiales and the search for the first dinokaryotic dinoflagellates

The branching order of extant groups at the base of the dinokaryotic dinoflagellates is proving to be very difficult to determine using molecular methods: phylogenetic trees calculated through different algorithms and based on different genes place very different taxa at those basal positions, and bootstrap supports are never strong. Nevertheless, there are tendencies that warrant comparison with the morphological and paleontological data available.

Yoon et al. (2002), for example, propose Karenia and Karlodinium as sister taxa to the rest of the dinokaryotic dinoflagellates. They used three plastid-encoded genes (psaA, psbA and rbcL) to test the phylogenetic relationships between haptophytes and the plastids of peridininand 19-hexanoyloxyfucoxanthin-containing dinoflagellates (photosynthetic dinoflagellates contain different types of chloroplasts, the type that contains peridinin is by far most common, see below). They found, as expected, a strong phylogenetic relationship between haptophytes and the plastids of Karenia and Karlodinium, the two genera with 19- hexanoyloxyfucoxanthincontaining plastids (see also Tengs et al., 2000; Ishida and Green, 2002). Surprisingly, they also found that in psaA and psbA-based trees peridinin-containing dinoflagellates group either as a sister-taxon to Karenia and Karlodinium, or embedded within a clade with 19-hexanoyloxyfucoxanthin-containing ancestors. They proposed an early tertiary endosymbiosis event for the dinoflagellate lineage, and a later transformation of that plastid into the peridinin-type after the divergence of Karenia and Karlodinium.

The Karenia/Karlodinium clade is one of the groupings that does sometimes branch at the base of the dinokaryotic dinoflagellates in SSU-based phylogenetic trees (e.g. Fig. 2). Both of these genera are athecate taxa currently classified in the order Gymnodiniales, the grouping proposed by Fensome et al. (1993) as the most basal of the wholly dinokaryotic dinoflagellates. However, there are reasons to question this model for the origins of peridinin-containing plastids and the phylogenetic position of the 19-hexanoyloxyfucoxanthin- containing dinoflagellates. First, as Yoon et al. (2002) point out, the divergence rate of the dinoflagellate genes they examined is noticeably accelerated. Consequently, the dinoflagellate sequences may be attracted to one-another because they share long branches. The authors attempted to correct for this attraction, but nevertheless the concern remains. Second, an analogous study of the relationships between Karenia, the haptophytes and the peridinin- containing dinoflagellates using a nuclear-encoded but plastid- targeted gene (psbO, Ishida and Green, 2002) produced different results: the one sequence for a peridinin-containing dinoflagellate (Heterocapsa triquetra) was strongly excluded from a Karenia/ haptophyle grouping. This finding is probably more reliable because the divergence rates in the nuclear-encoded dinoflagellate psbO genes appear to be much slower than the plastid-encoded genes.

The genus Heterocapsa occupies a basal position within Dinokaryota surprisingly often in phylogenetic trees, especially those based on LSU (maximum likelihood) and alpha-tubulin; in combined SSU/LSU trees Heterocapsa consistently occupied such a position, although with a low bootstrap support (40-50%). Heterocapsa also has a somewhat atypical sulcal tabulation that could be interpreted as primitive with respect to that of the rest of the Peridiniales and the Gonyaulacales (discussion in Fensome et al., 1993). It is thus not unreasonable that this group may have diverged before the split between those two orders. Palcontological data also agree with this hypothesis: the earliest fossils from the family Heterocapsaceae are early Jurassic (Wille, 1982), prior to the radiative explosion of all other peridininialean and gonyaulacalean forms in the Mesozoic (Fensome et al., 1999).

The phylogenetic history of gymnodinialean dinoflagellates is particularly difficult to discern for several reasons. First, although the order is well defined as a group in which the cellular cortex contains relatively numerous amphiesmal vesicles arranged non- serially (Fensome et al., 1993), several of the species that have historically been classified here have been shown to contain tabulations that make them obvious members of other orders (e.g. Hansen, 1995 for Katodinium rotundatum/Heterocapsa rotundata, Montresor et al., 1999 for Polarella glacialis, Saldarriaga et al., 2003b for Gymnodinium elongatum/Lessardia elongala). These tabulations are very difficult to discover using light microscopy alone, so it is a virtual certainty that several (perhaps many) taxa currently classified in Gymnodiniales are really members of other orders. This makes the evaluation of phylogenetic trees where putatively gymnodinialean clades intrude into thecate orders very difficult; a stringent evaluation of the tabulational patterns of putatively gymnodinialean taxa is needed before strong statements can be made about their phylogenetic history. Furthermore, small, non-serially arranged amphiesmal vesicles do not necessarily imply the absence of a theca (Netzel and Durr, 1984), many gymnodinialean taxa have either a full-fledged theca (e.g. the genus Woloszymkia: Crawford et al., 1970; Crawford and Dodge, 1971), or an incipient one (several members of Gymnodinium, e.g. G. fuscum and G. cryophilum: Hansen et al., 2000; Wilcox et al., 1982); others have flocculent material or just liquid and no signs of a theca (e.g. Karlodinium micrum, Leadbeater and Dodge, 1966; A. carterae, Dodge and Crawford, 1968). These features can only be studied by electron microscopy, and because relatively few species have been investigated in such detail, the degree to which presence and type of intraalveolar material in Gymnodiniales is phylogenetically informative remains unknown.

Molecular data only exists for some gymnodinialean families, there are still no data available for Polykrikaceae, Warnowiaceae, Actiniscaceae, Ptychodiscaceae and others. Even \so, molecular phylogenies always show a number of separate gymnodinialean clades originating from within the GPP complex, generally separated from thecate forms by very weak bootstrap supports and not necessarily basal to them. But are Gymnodiniales sensu stricto (i.e. dinoflagellates with numerous small alveoli arranged non-serially) polyphyletic or not? Or is the reason for the polyphyly of the Gymnodiniales sensu lato only the fact that it contains species with unrecognized non-gymnodinialean tabulations? Molecular data seem to suggest that even Gymnodiniales sensu stricto are polyphyletic: well- studied taxa with small alveoli (e.g. A. carterae, Karenia brevis, G. fuscum) never group together in phylogenetic trees.

The fact that in virtually all molecular trees gymnodinialean species arise from within the GPP complex, separated from thecate taxa by very weak bootstrap values, suggests that most, if not all, groups of Gymnodiniales had thecate ancestors; the different types of alveolar inclusions in the group, from thecae to flocculent material to nothing at all would therefore represent intermediate stages of thecal loss. The alternative would be that the Gymnodiniales (or at least some of their subgroups) are the sister group to the other dinokaryotic dinoflagellates. This view is supported by the fact that the small alveoli of Gymnodiniales are shared with more basal members of the dinoflagellate lineage, e.g. the syndinians (plasmodial life stage), Oxyrrhis and even Colpodella. Molecular data cannot distinguish between these possibilities at present; it cannot determine whether some Gymnodiniales are ancestral and others derived. Palaeontology is not very helpful in this regard either: gymnodinialean cysts are often very difficult to ally to identifiable motile stages, so fossil cysts of this type are particularly likely to be considered acritarchs, microfossils without known taxonomical affinities (Fensome et al., 1993); the earliest certain gymnodinialean fossils (skeletal elements from Actiniscaceae and Dicroerismaceae) are relatively recent, from Tertiary formations.

Palaeontological data suggest an early origin for another order of dinokaryotic dinoflagellates, the Suessiales. They comprise organisms with amphiesmal vesicles arranged in 7-10 latitudinal series, fewer than in typical athecate dinoflagellates and more than in thecate ones, a feature that suggests an interesting position for the order between thecate and athecate forms. Much more interesting, however, is the fact that suessialean fossils are known from the mid- Triassic, prior to the emergence of most (if not all) peridinialean and gonyaulacalean forms (the gonyaulacalean Shublikodiniaceae, like the earliest suessialean fossils, are from the mid-Triassic, Fensome et al., 1999; the identity of any Palaeozoic fossils as true dinoflagellates is still controversial). Molecular trees presently do not support any particular position of the Suessiales: although the group rarely appears at the base of the dinokaryotic dinoflagellates (e.g. in Edvardsen et al., 2003), its position in other parts of the tree is never supported either. One additional feature of the Suessiales is becoming clearer as the SSU gene of more species of dinofiagellates is sequenced: the group is likely to be larger than previously assumed. Montresor et al. (1999) described the first extant member of the family Suessiaceae, a group until then known only from fossils, and since then many species of putatively athecate dinofiagellates have been shown to group in the same clade in both SSU- and LSU-based trees (e.g. G. beii, G. simplex, G. corii and Woloszynskia pseudopalustris). It will be interesting to see whether close morphological examination of more of these species will agree with the molecular data.

In summary, the problem of the earliest-branching dinokaryotic dinofiagellates is very far from being resolved: morphological and palaeontological data could be interpreted as pointing towards gymnodinialean and suessialean taxa for these positions, molecular data towards Heterocapsa. What is clear is that Gymnodiniales are a polyphyletic group; many gymnodinialean taxa are more closely related to thecate forms than to other gymnodinialeans and thecal loss in dinoflagellates seems to be common. Suessiales may form an intermediate stage between the primitively athecate gymnodiniales (if they exist) and thecate forms, but this is particularly uncertain as phylogenetic trees do not usually put them in a basal position. If dinokaryotic dinofiagellates indeed underwent an event of rapid evolutionary radiation early in their history, it will be very difficult to determine the phylogenetic order of the groups that originated in that explosion.

Phytodiniales

The order Phytodiniales (also Dinococcales, Dinocapsales or Dinamoebales, see Fensome et al. (1993) for a nomenclatural discussion) contains dinoflagellates in which the principal life stage is either a non-calcareous coccoid cell or a continuous- walled multicellular stage. It is a polyphyletic grouping of convenience used to contain species that are poorly understood; the only criterion for determining whether a species belongs to this order is a shift in life cycle that has also been seen in many dinoflagellate genera with well-known tabulations, e.g. in Symbiodinium (Suessiales), Pyrocystix (Gonyaulacales) and Thoracosphaera (probably Peridiniales). SSU rRNA sequences exist for three dinoflagellate species formally classified in the Phytodiniales: Halostylodinium arenarium, Hemidinium nasutum, and Gloeodinium viscum. H. nasutum and the type species of Gloeodinium, G. montanum, have extremely similar coccoid stages, the two species have even been proposed to be identical (Popovsky, 1971). A fourth species in our trees, Glenodiniopsis steinii, is currently classified in the Peridiniales, but has a coccoid life stage reminiscent of the Gloeodinium-like stage of Hemidinium nasutum (Popovsky and Pfiester, 1990).

Halostylodinium arenarium groups with gonyaulacalean taxa in all phylogenetic trees examined, a placement that is congruent with most (but not all) tabulational features of the species as interpreted by Horiguchi et al. (2000). Glenodiniopsis, Hemidinium and Gloeodinium on the other hand, consistently branch within the GPP complex, although only in the Weighbor trees do the three species weakly branch close to one-another (clades consisting of two of the three species are common in many trees). This placement is congruent with the peridinialean tabulation of the motile stage of Glenodiniopsis, and suggests that once the tabulations of Hemidinium and Gloeodinium are fully determined (only a partial tabulation is known for Hemidinium nasutum, no tabulational data exists for G. viscum) they will show peridinialean affinities. Molecular data do not strongly support a phylogenetic relationship between Hemidinium nasutum and Gloeodinium viscum, but do not disprove it either.

Peridiniales, Gonyaulacales, Dinophysiales and Prorocentrales

Members of the Peridiniales, Gonyaulacales, Dinophysiales and Prorocentrales are likely to have a common ancestor. The relative positions of the thecal plates in Peridiniales and Gonyaulacales are so similar that a close relationship between the two orders has never been doubted, and paleontological and morphological evidence points to a close relationship between Peridiniales, Dinophysiales and Prorocentrales. Palaeontological data yielded very strong evidence linking Dinophysiales to peridinialean ancestors: the fossil genus Nannoceratopsis, found as dinosporin cysts in marine strata of jurassic origin, has distinctly dinophysialean features in its lateral compressed shape, hyposomal sagittal features and hyposomal pseudotabulation, but its epitheca has distinct peridinialean traits, very different from those of other Dinophysiales (Piel and Evitt, 1980; Fensome et al., 1993). Within Peridiniales, the groups with the most similarity to Nannoceratopsis are the fossil Comparodiniaceae and the extant Oxytoxaceae (Fensome et al., 1993). No molecular data exist for Oxytoxum, but a close relationship between Peridiniales and Dinophysiales is weakly apparent in molecular phylogenetic trees: in our trees Dinophysiales are always embedded in the GPP complex (alternative placings for the group also exist, e.g. Edvardsen et al., 2003). Prorocentrales also branch within the GPP complex, even if in SSU trees the order splits into 2 groups (or more: Grzebyk et al., 1998). Despite this split, we consider that this morphologically very cohesive order is monophyletic, as tabulation patterns within it are both radically derived and homogenous. Interestingly, at least some LSU trees (notably Weighbor) show Prorocentrales as monophylctic. The phylogenetic origins of the group are more difficult to discern. The fact that Prorocentrales are members of the GPP complex in molecular trees weakly argues for a relationship to Peridiniales and Dinophysiales, as well as to many Gymnodinialcs. Two large lateral plates (valves) that contact each other along a sagittal suture are a common feature of Prorocentrales and Dinophysialcs; they may be closely related to each other (Taylor, 1980). Nevertheless, no intermediate fossil forms exist to shed light on this.

Thus, a relationship between Peridiniales and the Dinophysiales/ Proroccntralcs on the one hand, and Oonyaulacalcs on the other, seems likely. But which appeared first? The earliest dinoflagellate fossils (except for the controversial Palaeozoic forms already mentioned) are, in addition to the Suessiales, members of the gonyaulacalean Shublikodiniaceae (Fensome et al., 1999). This is an exclusively fossil family with very characteristic tabulation: more than five climactal plates and at least three fundital plates. Fensome et al. (1993) points out tabulational resemblances between this family and two other groups: early cladop\yxiineans and living members of the genus Glenodinum. What is interesting is that whereas Shublikodiniaceae and Cladopyxiineae are early lineages within the Gonyaulacales (as classified by Fensome et al., 1993), Glenodiniaceae are undoubtedly peridinialean forms. Thus, the lines between the two orders blur at this level. Molecular data tend to give trees where a paraphyletic Peridiniales is ancestral to the holophyletic Gonyaulacalcs, although bootstrap support for this branching order is generally low. Nevertheless, molecular data do not exist for many putativcly basal groups of cither Gonyaulacales or Peridiniales, genera like Cladopyxis, Acanthodinium, Amphidoma or Palaeophalacroma (Gonyaulacales), or Glenodinium (Peridiniales). Heterocapsa (Peridiniales) is an exception to this, and as discussed above, it tends to take a basal position to other thecate dinoflagellates in many phylogenetic trees, particularly those based on combined data sets of small- and large subunit rRNA genes. The implication of this position would be that Peridiniales are indeed ancestral to Gonyaulacaleans, a thesis that runs contrary to palaeontological data: no true peridinialean fossils are known from before the appearance of the earliest gonyaulacaleans, the Shublikodiniaceae. Nevertheless, just as it is dangerous to give too much credence to the branching order of Heterocapsa in phylogenetic trees because of mediocre supports of the relevant nodes, the present lack of peridinialean fossils from the Triassic does not necessarily mean that the group was completely absent then.

One more dinoflagellate “order” appears to be very closely related to Peridiniales: the Thoracosphaerales. The principal life- stage of Thoracosphaera, the only genus in the order, is a coccoid cell surrounded by a calcareous wall, very similar to calcareous cysts of a subgroup (subfamily Calciodinelloideae of the Peridiniaceae) within the Peridiniales that includes Scippsiella, Ensiculifera, etc. Nevertheless, the motile stage of Thoracosphaera is apparently athecate and the archeopyle of the cyst quite atypical, so a separate order was created for it (Tangen et al., 1982). Molecular data tends to support a relationship between Thoracospahera and several genera of Peridiniales, including Peridiniopsis and, interestingly, Scrippsiella. This position in molecular trees (as well as the calcareous cyst wall) would predict a peridinialean tabulation of the motile stage. If this turns out to be the case, the order Thoracosphacrales should be abolished and Thoracosphaera made a member of the Peridiniales and of the Calciodinelloideac.

If the scenario presented above is correct, the Peridiniales would occupy a very important place for the phylogeny of the dinoilagellates as a whole. They would be a complexly paraphyletic group that gave rise not only to the other obviously thecate taxa (Dinophysiales and Prorocentrales, and perhaps also Gonyaulacales), but to many athecate and putatively athecate forms as well (many lineages of Gymnodiniales, Thoracosphaerales, as well as, perhaps, also Noctilucales and Blastodiniales).

Whereas branching orders within Peridiniales are not resolved in any of our trees, within Gonyaulacales the rate of evolution of both the large- and the SSU rRNA genes is faster, and as a consequence branches of the resulting phylogenetic trees are longer; lineages are also generally separated by better bootstrap supports. Felicitously, the Gonyaulacales are also a group with a very good fossil record, and the tabulational patterns of extant and fossil members are well known. For this reason, the group provides a good model to contrast taxonomic schemes based on morphology (i.e. tabulation) with molecular data.

The two sets of data correlate well within this group. Two of the three gonyaulacalcan suborders for which there are SSU sequences are normally recovered in phylogenetic trees with decent bootstrap support: Goniodomineae (50-60%) and Ceratiineae (75-90%). The third suborder, Gonyaulacineae, usually forms a paraphyletic group that gives rise to both Goniodomineae and Ceratiineae, as well as to taxa of uncertain taxonomic position like Thecadinium and Crypthecodimium (and the formally phytodinialcan genus Halostylodinium). Large subunit gene trees also generally support the monophyly of Goniodominae, although with weak bootstrap support and one important caveat: it is always intruded by the sequence from Ceratium furca (SSU and morphological data agree that this species is a Ceratiineae, the veracity of the identity of this LSU sequence should be reexamined). The other Ceratiineae (i.e. the genus Ceratium) group strongly, and Protoceratium and Gonyaulax, the only Gonyaulacineae in the trees, make a paraphyletic group at the base of the order. One difference between molecular trees and taxonomic schemes based on morphology (i.e. Fensome et al., 1993) is the position of Protoceratium: SSU-based phylogenies never place it with Linyulodinium, Gonyaulax and Amy lax in the family Gonyaulacaceae, but rather with Ceratocorys (Ceratocoryaceae, 100% bootstrap support).

Tracing morphological characters onto molecular trees: from protalveolates to dinoflagellates

Having proposed a putative framework for the phylogenetic history of dinoflagellates, we now examine the evolutionary history of certain morphological features in its light.

The nucleus

Although undoubtedly eukaryotic (e.g. Minguez et al., 1994; Salamin Michel et al., 1996), dinokaryotic nuclei show important biochemical differences with respect to the nuclei of other eukaryotes: they lack histones (e.g. Rizzo, 1991), can contain very large amounts of DNA (2-200 pg DNA per haploid nucleus, nuclei of human cells have ca. 5.6pg DNA per cell, Sigee, 1986) and up to 70% of the thymine in their DNA is replaced by 5-hydroxymethyluracil (Rae, 1976). They also divide through a type of mitosis characteristic for the group: the nuclear membrane remains intact, chromosomes remain attached to the nuclear membrane, and, during mito