PHYLOGENY of DINOFLAGELLATES BASED on MITOCHONDRIAL CYTOCHROME B and NUCLEAR SMALL SUBUNIT Rdna SEQUENCE COMPARISONS1

J. Phycol. 41, 411–420 (2005) r 2005 Phycological Society of America DOI: 10.1111/j.1529-8817.2005.04168.x

PHYLOGENY OF DINOFLAGELLATES BASED ON MITOCHONDRIAL CYTOCHROME b AND NUCLEAR SMALL SUBUNIT rDNA SEQUENCE COMPARISONS1

Huan Zhang Department of Marine Sciences, University of Connecticut, Groton, Connecticut 06340 USA Debashish Bhattacharya Department of Biological Sciences and Roy J. Carver Center for Comparative Genomics, University of Iowa, 312 Biology Building, Iowa City, Iowa 52242-1324 USA and Senjie Lin2 Department of Marine Sciences, University of Connecticut, Groton, Connecticut 06340 USA

Despite their evolutionary and ecological impor- Key index words: cob; cytochrome b; dinoflagellates; tance, dinoflagellate phylogeny remains poorly phylogeny; rDNA; tertiary endosymbiosis resolved. Here we explored the utility of mitochondrial cytochrome b (cob) in inferring a dinoflagellate Abbreviations: AU, approximately unbiased; COB, tree and focused on resolving the relationship be- cytochrome b; cob, gene coding for COB; GAPDH, tween fucoxanthin-and peridinin-containing taxa. glyceraldehyde-3-phosphate dehydrogenase; ML, Trees were inferred using cob and small subunit maximum likelihood; MP, maximum parsimony; rDNA alone or in combination as concatenated data mt, mitochondrial; SSU, small subunit and including members of the six major dinoflagellate orders. Many regions of the cob DNA or protein and rDNA trees were congruent with support Dinoflagellates (subphylum Dinoflagellata, phylum for the monophyly of Symbiodinium spp. Freud- Dinozoa) are an evolutionarily diverse group of pro- enthal and of the Prorocentrales and the early di- tists that are closely related to the apicomplexan par- vergence of Crypthecodinium cohnii Seligo in Grasse. asites and together with the ciliates form the crown However, these markers provided differing support group Alveolata (Gajadhar et al. 1991, Cavalier-Smith for the monophyly of Pfiesteria spp. Steidinger et 1998). Dinoflagellates play pivotal roles in the marine Burkholder (only supported strongly by rDNA) and ecosystem and are of tremendous economic signifi- of the fucoxanthin dinoflagellates with Akashiwo sp. cance. They are important primary producers, with (Hirasaka) Hansen et Moestrup (Gymnodiniales, their abundance second only to diatoms; they support only supported strongly by the cob data). The ap- coral reef ecosystems through symbiotic associations; proximately unbiased (AU) test was used to assess and they include species that cause harmful algal these results using 13-and 11-taxon (excluding blooms. Dinoflagellates are considered unusual eukar- apicomplexans) backbone maximum likelihood yotes because, among other well-recognized unique trees inferred from the combined cob þ rDNA data. cytological features (Hackett et al. 2004a), they harbor The AU test suggested that our data were insuffi- a highly reduced plastid genome (approximately cient to resolve the phylogenetic position of Symbio- 14 genes, Hackett et al. 2004b) of 1–3 gene minicir- dinium spp. and that the ancestral position of cles (Zhang et al. 1999, Barbrook and Howe 2000, C. cohnii might have resulted from long-branch at- Koumandou et al. 2004), a nuclear-encoded form II traction to the apicomplexan outgroup. We found RUBISCO of alpha-proteobacterial origin (Morse significant support, however, for the association of et al. 1995, Rowan et al. 1996, Zhang and Lin 2003), fucoxanthin dinoflagellates with Akashiwo sp. The and a complex RNA editing machinery (Lin et al. monophyly and relatively derived position of the 2002). The ecological and economic importance of Gymnodiniales in our cob DNA and protein trees dinoflagellates, combined with their unusual genetic/ and in the cob þ rDNA tree is consistent with the genomic traits, renders them an evolutionary group of tertiary endosymbiotic origin of the plastid in high biological interest. fucoxanthin dinoflagellates. Despite the increasing research effort in recent years to study the phylogeny of dinoflagellates and their plastids, many questions remain open regarding 1Received 13 September 2004. Accepted 31 December 2004. the evolution of the ‘‘host’’ cell and the photosynthetic 2Author for correspondence: e-mail [email protected]. organelle (Hackett et al. 2004a). Resolution of these

411 412 HUAN ZHANG ET AL. issues would be greatly aided by the availability of a targeted glyceraldehyde-3-phosphate dehydrogenase resolved dinoflagellate host tree. The small subunit (GAPDH) (unpublished data) of haptophyte origin in (SSU) rRNA gene (rDNA) has been instrumental in the K. brevis nuclear genome. Despite these data, the advancing our understanding of dinoflagellate evolu- monophyly of fucoxanthin-containing taxa has had tion; however, the low resolving power of this gene varying levels of support from analyses of nuclear regarding lineage relationships has resulted in an in- and chloroplast SSU rDNA (Tengs et al. 2000, Sal- complete understanding of the major splits among darriaga et al. 2001, de Salas et al. 2003) and remains dinoflagellates despite a taxonomically broad sampling to be substantiated. The cob analyses presented here of data (Saunders et al. 1997, Saldarriaga et al. 2001). allowed us to address the position of fucoxanthin- Recent studies clearly show the advantage of using containing taxa in the dinoflagellate tree. multiple genes for phylogenetic analysis (Pryer et al. 2001, Mattern 2004, Yoon et al. 2004). Mitochondrial MATERIALS AND METHODS (mt) DNA is a useful candidate for phylogeny recon- Algal cultures. Cultures of dinoflagellates and other algae struction for several reasons. First, mt genes are gen- used in this study were obtained from several sources (Table erally conserved although are more variable than 1). The photosynthetic species were grown in f/2 medium, nuclear genes such as SSU rDNA (Avise 1994, Saccone whereas heterotrophic dinoflagellates (Pfiesteria shumwayae et al. 2000, Garesse and Vallejo 2001). Second, the Glasgow et Burkholder, CCMP1827, CCMP1828, and mitochondrion likely emerged at nearly the same time as CCMP1835) were grown with an algal prey (Rhodomonas sp. Karsten CCMP768). Seawater was adjusted to 28 psu for the nucleus (Gray et al. 1999), rendering mt genes de- most species and to 15 psu for Rhodomonas sp., K. micrum,and spite their organellar location appropriate for studying the heterotrophic dinoflagellates. The temperature was eukaryotic host cell phylogeny (Lang et al. 1998, maintained at 15 11 CforAlexandrium tamarense (Lebour) 1999). Third, some mt genes are closer to a molecu- Balech; 20 11 CforRhodomonas, K. micrum, Akashiwo sp., lar clock than SSU rDNA (i.e. they have a relatively Ceratium sp. Schrank, and the heterotrophic dinoflagellates; constant mutation rate across taxa) and thus are ideal and 25 11 CforK. brevis and the two Symbiodinium species for cross-taxa phylogenetic studies (Saccone et al. (CCMP830, CCMP832). Illumination was provided in a 12:12-h light:dark cycle with a photon flux of around 2000). Cytochrome b (cob) is one of the mt genes 100 mmol photons m–2 s–1. The growth rate was monitored most widely used for phylogenetic and population ge- by microscopic cell counts using a Sedgwick-Rafter chamber netic analyses (Conway et al. 2000, Taylor and Hel- (Phycotech, St. Joseph, MI, USA). lberg 2003). A combination of cob and nuclear genes Sample collection and DNA extraction. Samples were collect- has provided robust phylogenetic trees for alveolates ed when cultures were in the exponential growth phase. and other organisms (Serizawa et al. 2000, Rathore Samples of heterotrophic dinoflagellates were collected after feeding was discontinued for 2 days and few cells of the prey et al. 2001). To date, cob has been studied only for alga Rhodomonas sp. were present. Previous studies had three species of dinoflagellates (Lin et al. 2002), and its shown that Rhodomonas cob and SSU rDNA sequences were utility for inferring the dinoflagellate phylogeny has distinct from those of dinoflagellates. The PCR primers used yet to be explored. Here we cloned and sequenced cob to amplify the dinoflagellates genes did not therefore target for 14 species from six major dinoflagellate orders: the Rhodomonas homologs, even if contaminating DNA was Gonyaulacales, Gymnodiniales, Prorocentrales, Peri- present in the samples (Zhang and Lin 2002). The cells were harvested by centrifugation at 3000 g at 41 C for 20 min. diniales, Suessiales, and Dinamoebales (Pfiesteria and These cell pellets were subjected to DNA extraction in a its relatives). We then used the existing and newly ob- buffer containing 0.1 M EDTA, 1% SDS essentially following tained cob gene sequences to reconstruct the dino- Zhang and Lin (2002). flagellate tree. PCR, cloning, and sequencing. To PCR amplify cob from Based on cob and SSU rDNA analyses, we also at- dinoflagellates, a set of primers was designed from the highly tempted to address the phylogeny of the Gymnodini- conserved regions of this gene based on previous data ales. This clade includes taxa (e.g. Gymnodinium Stein (Zhang and Lin 2002). The primer sequences were Dino- cob1F (forward), 50-ATGAAATCTCATTTACAWWCATAT- sensu stricto, Akashiwo spp.) that, like most photosyn- CCTTGTCC-30, and Dinocob1R (reverse), 50-TCTCT- thetic dinoflagellates, possess peridinin as the main ac- TGAGGKAATTGWKMACCTATCCA-30. The PCR reaction cessory pigment as well as taxa (Karenia spp. Hansen et was done using approximately 50 ng of genomic DNA for Moestrup, Karlodinium micrum [Leadbeater et Dodge] each species, and amplification was carried out with a single Larsen, and Takayama spp. de Salas, Bolch, Botes et incubation for 1 min at 951 C, followed by 40 cycles of 20 s at 941 C, 30 s at 551 C, and 40 s at 721 C. For several dinoflag- Hallegraeff) that lack peridinin but contain chl c2 þ c3 0 0 ellates, the SSU rDNA was also amplified using the universal and 19 -hexanoyloxyfucoxanthin and/or 19 -but- primers (18ScomF1 [forward], 50-GCTTGTCTCAAAGAT- anoyloxyfucoxanthin, similar to the haptophyte algae TAAGCCATGC-3 0,18ScomR1[reverse],50-CACCTACGG- (Daugbjerg et al. 2000, Tengs et al. 2000, de Salas AAACCTTGTTACGAC-30) or the combination of the uni- et al. 2003). It has been postulated that a tertiary versal primers with the dinoflagellate-specific primers plastid endosymbiosis (involving a haptophyte alga) (dino18S5F1 [forward], 50-AAGGGTTGTGTTTATTAGN- TACAGAAC-30, dino18SR1 [reverse], 50-GAGCCAGATRC- in the common ancestor of fucoxanthin-containing 0 dinoflagellates explains this distribution of pigment WCACCCAG-3 ). The PCR products were purified using DNA Clean & Con- characteristics (Tengs et al. 2000). This hypothesis centrator (Zymo Research, Orange, CA, USA) and directly se- was reinforced by the finding of oxygen evolving en- quenced on both strands using BigDye reagents and an ABI hancer 1 (psbO) (Ishida and Green 2002) and plastid Prizm automated sequencer (Perkin Elmer, Branchburg, NJ, PHYLOGENY OF DINOFLAGELLATES 413

TABLE 1. Dinoﬂagellate species included in the phylogenetic analyses.

Trophic Taxa Strain and source modee SSU rDNA coba Alexandrium tamarense CB307; D. P AF022191f AY456116g (Lebour) Balech M. Anderson Akashiwo sp. (Hirasaka) Hansen et Moestrup LIS1b P AY456106g AY456105g Akashiwo sp. (Hirasaka) Hansen et Moestrup LIS2b P AY456108g AY456107g Ceratium sp. Schrank ML1c P AY460574g AY460573g Crypthecodinium cohnii Seligo in Grasse WHd; M. Gray H M64245f AF403220f Heterocapsa triquetra (Ehrenberg) Stein CCPM449d AJ415514f AY481575g Karlodinium micrum (Leadbeater et Dodge) Larsen D. K. Stoecker M AF272050f AY345908g Karenia brevis Hansen et Moestrup CCP P AF274259f AY456104g M2229d Lingulodinium polyedrum CCMP405d AJ415511f AY456109g Pfiesteria piscicida Steidinger et Burkholder CCMP1831d H AF330620f AF357519 P. shumwayae Glasgow et Burkholder P. Tester, North Carolina H AF218805f AF502593g Pfiesteria-like CCMP1835d H AY590477g AY456120g Pfiesteria-like CCMP1827d H AY456118g AY456117g Pfiesteria-like CCMP1828d H AY590476g AY456119g Prorocentrum micans Ehrenberg CCMP1589d P AY585526g AY585525g P. minimum (Pavillard) Schiller CCMP696d P Y16238f AY030286 Pyrodinium bahamense Plate Azanza-Corrales P AY456115g AY456114g Scrippsiella sp. Balech LISc AY743960g AY743961g Symbiodinium microadriaticum CCMP830d P AY456111g AY456110g Freudenthal Symbiodinium sp. CCMP832d P AY456113g AY456112g Freudenthal

aThe cob sequences from the remaining taxa in Figure 1 are publicly available from GenBank. bIsolated from Long Island Sound in April 2003. Isolates LIS1 and LIS2 of Akashiwo sp. were morphologically similar except that LIS1 is the smaller cell. Both resemble Akashiwo sanguinea based on the description in Steidinger and Tengen (1997) and SSU rDNA sequence. The LIS isolate of Scrippsiella sp. is most closely related to S. trichoidea based on morphology (Steidinger and Tengen 1997) and SSU rDNA identity (99% [1738bp/1754bp] to AY421792). cIsolated from Mirror Lake on the Storrs campus of the University of Connecticut; it is morphologically similar to Ceratium hirundinella based on Graham and Wilcox (2000). dProvasoli-Guillard National Center for Culture of Marine Phytoplankton, West Boothbay Harbor, Maine. eP: photoautotrophic; H: heterotrophic; M: mixotrophic. fSequences obtained from other strains of the same species. gSequences determined in this study.

USA). In some cases, the PCR products were cloned into a TA dom starting tree was run for 1,000,000 generations with plasmid vector and the plasmid was propagated using XL1 trees sampled each 1000 cycles. Four chains were run simul- blue cells (Stratagene, La Jolla, CA, USA). To identify potential taneously of which three were heated and one was cold, with PCR-related polymorphisms, plasmids were isolated from 5 to the initial 500,000 cycles (500 trees) discarded as the burn-in. 10 colonies, sequenced over both strands, and the sequences A consensus tree was made with the remaining 500 trees to compared with each other. determine the posterior probabilities at the different nodes Phylogenetic analysis. Sequences were aligned using the in the ML tree. CLUSTAL W (1.8) server at the DNA Data Bank of Japan Thereafter, we analyzed dinoflagellate cob DNA (834 nt), (DDBJ) (http://www.ddbj.nig.ac.jp/Welcome-e.html) using COB protein (237 amino acids), and SSU rDNA (1469 nt) the default values. The alignment was then manually adjust- alignments as individual data sets as well as a concatenated data ed using the SeaView program to maintain codon integrity set of cob þ rDNA (2303 nt). The apicomplexans, Plasmodium (Galtier et al. 1996). An initial protein (237 amino acids) tree spp., were used to root these trees. The COB protein analysis was inferred to test the utility of COB in reconstructing the was done as described above except that we also used the un- eukaryotic phylogeny as well as to determine the order of weighted maximum parsimony (MP) method to calculate boot- divergence within the dinoflagellates. This maximum likeli- strap (200 replicates) support values for nodes in the ML tree hood (ML) analysis was performed with proml in PHYLIP with PAUP (Swofford 2003). The COB protein sequence was ver. 3.61 (Felsenstein 2004) using the JTT þ G evolutionary deduced from the cob gene sequence with internal stop codons model. The alpha value for the gamma distribution (eight being treated as missing data; the internal stop codons, ranging categories) was calculated using TREE-PUZZLE ver. 5.1 0–2, occurred in some species and they are transcriptionally (Schmidt et al. 2002) and the same evolutionary model. Glo- edited to, or serve as, sensible codons (Lin et al. 2002, unpub- bal rearrangements and randomized sequence input order lished data). Ten heuristic searches were used in the MP anal- with five replicates was used to find the best tree. To test the ysis with random-addition-sequence starting trees (10 rounds), stability of monophyletic groups in the ML protein tree, 100 and tree bisection-reconnection branch rearrangements were bootstrap replicates were analyzed with proml as described done to find the optimal parsimony tree. Best scoring trees above. We also used Bayesian inference of the amino acid were held at each step. For each DNA data set, we implement- data (MrBayes ver. 3.0b4, Huelsenbeck and Ronquist 2001). ed the ML method in PAUP (GTR þ I þ G model) and used the The mtrev þ G model was implemented in this analysis, and search strategy described above for the MP analysis. The pa- Metropolis-coupled Markov chain Monte Carlo from a ran- rameters for the GTR þ I þ G model were calculated with 414 HUAN ZHANG ET AL.

PAUP using a starting distance (minimum evolution [10 ran- alone for additional assessment of the divergence point of the dom additions]) tree in each case that was built with a matrix of fucoxanthin dinoflagellates. Finally, we also generated the HKY85 distances. We analyzed 100 bootstrap replicates with same sets of topologies in an 11-taxon backbone ML tree that the ML method. The DNA data were also analyzed with boot- excluded the highly divergent Plasmodium spp. The site-by- strap ME-GTR þ I þ G and unweighted MP methods (2000 site likelihoods for these pools of trees were calculated using replicates). We also did Bayesian analysis of the cob,COB, the respective data sets and Baseml implemented in PAML rDNA, and cob þ rDNA data sets using the GTR þ I þ G model ver. 3.13 (Yang 1997) with the GTR þ G model and the de- for DNA and the mtrev þ G model for the COB protein data as fault settings. The approximately unbiased (AU) test was im- described above, except that 5 million generations were run plemented using CONSEL ver. 0.1f (Shimodaira and and the final 1000 trees were used to calculate the posterior Hasegawa 2001) to assign probabilities to the different trees probabilities. The third codon positions of cob were excluded in in each pool. a set of DNA analyses that were run as described above. Finally, analysis of the cob þ rDNA data set using the partition homo- geneity test (ILD test in PAUP, 1000 replicates) showed signif- RESULTS icant incongruence between these data sets (Po0.001). However, we chose to combine the data because of substantial Gene sequence. Dinoflagellate cob is highly con- controversy regarding the utility of this test (Barker and served (see branch lengths in Figs. 1 and 2), although Lutzoni 2002, Hipp et al. 2004). thegeneismoreenrichedinAþT content (70%) Testing the tree topology. To assess the positions of different than the other eukaryotes in Figure 1 (about 60% dinoflagellates in the cob þ rDNA ML tree, we generated an ML backbone tree of 13 taxa (from the cob þ rDNA data set) A þT). The cob sequences for the two Long Island that included representatives of the major dinoflagellate Sound isolates of Akashiwo sp. (Daugbjerg et al. 2000) groups but excluded the fucoxanthin-containing taxa (K. bre- were identical as were their SSU rDNA sequences. vis, K. micrum) and the closely related Akashiwo sp. These cells were indistinguishable morphologically, (AY456107), Symbiodinium spp., and C. cohnii.Theseclades although their cell sizes were quite different (except Akashiwo sp.) were then added separately using Mac- (length width, 48.5 5.0 35.6 5.0 mmvs. Clade (Maddison and Maddison 2002) to each of the 23 m available branches in the 13-taxon tree to generate three sets 78.2 7.9 61.4 6.9 m; data from 50 cells for of topologies that addressed all the possible divergence each isolate). The cob gene in the two Symbiodinium points for the three dinoflagellate clades (i.e. independent strains was surprisingly divergent (9% dissimilarity) of each other). We did the same analysis using the cob data in comparison with other congeners (e.g. 2% for

FIG. 1. The ML phylogenetic tree of dinoﬂagellates and other organisms based on the COB sequence. The results of an ML bootstrap analysis are shown above the branches, and the thicker branches denote a Bayesian posterior probability 40.95. Branch lengths are proportional to the number of substitutions per site (see scale bar). The COB sequences are available in GenBank; the accession numbers for those of dinoﬂagel- lates are shown in Table 1. PHYLOGENY OF DINOFLAGELLATES 415

FIG. 2. The evolutionary relationships of dinoflagellates based on cob. (A) ML phylogeny of the dinoflagellates using the cob gene (GTR þ I þ G model). The results of an ML bootstrap analysis are shown above the branches, whereas ME-GTR þ I þ G (left of slash mark) and MP (right of slash mark) bootstrap support values are shown below the branches in italic text. The thick branches nodes represent 495% Bay- esian posterior probability (GTR þ I þ G model). (B) ML phylogeny of the dinoflagellates using the deduced COB amino acid sequence. The results of an ML bootstrap analysis are shown above the branches, and MP bootstrap support values are shown below the branches in italic text. The thick branches nodes represent 495% Bayesian posterior probability (mtrev þ G model). Only bootstrap values 60% are shown. Branch lengths are proportional to the number of substitutions per site (see scale bar).

Pfiesteria piscicida–P. shumwayae and Prorocentrum tion occurred only in C. cohnii and Prorocentrum spp. A minimum [Pavillard] Schiller–P. m i c a n s Ehrenberg). 15-bp deletion in the middle of cob in a crypt- Anumberofcob indels were identified among the operidiniopsoid taxon (strain CCMP1828) was absent studied species. Three 6-bp insertions were found in all other dinoflagellates, even in CCMP1827, which within about 80 bp of the 30 terminus of cob: The first is thought to be closely related to CCMP1828. insertion was Symbiodinium spp. specific; the second was Phylogenetic analyses. The COB amino acid se- common to Symbiodinium spp., K. micrum, K. brevis, quence was used to obtain a global view of the dino- A. tamarense,andP.bahamense; whereas the third inser- flagellate radiation relative to other eukaryotes. This 416 HUAN ZHANG ET AL.

FIG. 3. The evolutionary relationships of dinoﬂagellates based on an ML analysis of SSU rDNA (GTR þ I þ G model). The results of an ML bootstrap analysis are shown above the branches, whereas ME-GTR þ I þ G (left of slash mark) and MP (right of slash mark) bootstrap support values are shown below the branches in italic text. The thick branches nodes represent 495% Bayesian posterior probability (GTR þ I þ G model). Only bootstrap values 60% are shown. Branch lengths are proportional to the number of substitutions per site (see scale bar).

tree (Fig. 1) confirmed that the apicomplexans (e.g. 3), although there was greater support for the mon- Plasmodium spp.) are the closest relatives of dinoflag- ophyly of the Pfiesteria clade and for the Gonyaulacales ellates and identified C. cohnii as the earliest diverg- with this data set. The combined cob þ rDNA tree (Fig. 4) ing dinoflagellate in our data set. A number of other merged the results of the two individual data sets, show- eukaryotic groups were resolved as monophyletic ingmoderatetostrongsupport for Gymnodiniales, lineages (animals, fungi, chlorophytes), suggesting Pfiesteria, and Gonyaulacales monophyly with Symbiodi- that COB is a reasonable marker of eukaryotic phylo- nium as sister to the latter clade (however, without boot- geny. strap support). Analysis of the first and second codon With Plasmodium spp. as the outgroup, cob ML trees positions in the cob data resulted in essentially the same were generated that focused on the dinoflagellates results as shown in Figure 2A (e.g. 89% MP bootstrap (Fig. 2). The two phylogenies, derived from DNA support for Gymnodiniales monophyly [200 replicates]). (Fig. 2A) and the deduced amino acid sequences AU test results. TheresultsoftheAUtestwereen- (Fig. 2B), respectively, were generally congruent tirely consistent with the ML tree topologies shown in with each other, showing for example an early diver- Figures 1, 2, and 3 and with the associated bootstrap gence of C. cohnii and the monophyly of Symbiodinium support values. The most highly favored position for spp. and the Prorocentrum clades. In both trees, the the fucoxanthin dinoflagellates in the 13-and 11-tax- Gymnodiniales formed a monophyletic group with on ML backbone trees in the analysis of the strong bootstrap support, with a close evolutionary cob þ rDNA and the cob data sets was as sister to Aka- relationship being resolved between the fucoxanthin- shiwo sp. Placing K. brevis and K. micrum on virtually containing taxa (K. brevis, K. micrum) and Akashiwo sp. all other branches resulted in trees of significantly The position of the Symbiodinium clade was, however, lower probability (Fig. 5, A, B, E, and F), except quite different in the cob DNA and protein trees with when they were placed at the base of Prorocentrum the former suggesting a later divergence position and spp. with the cob þ rDNA data and as sister to H. the latter an early divergence (with bootstrap support) triquetra with the cob data. Exclusion of the divergent after C. cohnii. In the rDNA tree, Symbiodinium spp. Plasmodium spp. markedly increased the probability appeared to be derived as well (Fig. 2). The Gymno- of the favored position as sister to Akashiwo sp. (e.g. diniales were polyphyletic in the rDNA phylogeny (Fig. from P 5 0.483 to P 5 0.984 in the cob data,Fig.5,B PHYLOGENY OF DINOFLAGELLATES 417

FIG. 4. The evolutionary relationships of dinoﬂagellates based on an ML analysis of cob þ SSU rDNA (GTR þ I þ G model). The results of an ML bootstrap analysis are shown above the branches, whereas ME-GTR þ I þ G (left of slash mark) and MP (right of slash mark) bootstrap support values are shown below the branches in italic text. The thick branches nodes represent 495% Bayesian posterior probability (GTR þ I þ G model). Only bootstrap values 60% are shown. Branch lengths are proportional to the number of substitutions per site (see scale bar).

and F). The early divergence of C. cohnii in our trees cate that this gene is also conserved in dinoflagellates. was very strongly supported with the AU test, with this Furthermore, dinoflagellate cob has several unique position having a probability of 0.968 when Plasmodium features. First, it is more highly enriched in As and Ts spp. were included in the analysis. However, exclusion (70%) than the homolog in many other organisms of this outgroup resulted in the change of the most (about 60%). The high A þT content of cob has also favored position of C. cohnii to sister to Pfiesteria spp. been found in insects, nematodes, and eumycotes (P 5 0.755, Fig. 5G). This surprising result leaves un- and is thought to result from high A þT pressure resolved the phylogenetic position of C. cohnii in our (Jermiin et al. 1994). The high A þTcontentofcob analyses and suggests that the apicomplexans provide potentially would reduce translation efficiency, but a potentially misleading signal when included in the the problem is ameliorated in dinoflagellates by RNA trees. Interestingly, divergence at the base of the Gony- editing, which results in an increase in G þ Ccontent aulacales was the second most highly favored position (Lin et al. 2002). for C. cohnii (P 5 0.583) in Figure 5G. Similarly, the Cob phylogeny. The cob and rDNA trees provide position of Symbiodinium spp. was essentially unre- congruent results in many respects, with both show- solved with our data, and there were many alternative ing an affiliation of dinoflagellates with other alveo- positions (i.e. to that favored as sister to Prorocentrum lates such as Plasmodium spp. and similar branching spp., P 5 0.883, P 5 0.807; Fig. 5, D and G) for this patterns within the Dinophyceae. This is in concert branch within the 5% confidence set of trees. with neighbor joining and MP (not shown). The comparison of cob and rDNA trees provides valuable in- sight, especially for resolving lineages for which previous analyses show conflicting results. Combina- DISCUSSION tion of the two genes can potentially resolve many Mitochondrial cob has proven to be useful for phylo- nodes in the dinoflagellate phylogenetic tree. genetic studies of various organisms (Serizawa et al. The different phylogenetic trees suggest an early 2000, Rathore et al. 2001). Our results suggest that divergence of C. cohnii within the Dinophyceae, but this this gene is a useful addition to SSU rDNA for infer- result is not supported by the AU test when the diver- ring dinoflagellate phylogeny. gent Plasmodium spp. are excluded from the analysis. Sequence conservation. The cob gene is highly con- Therefore, our data do not exclude kinship of C. cohnii served in general (Zhang et al. 1999, Conway et al. to the other Gonyaulacales in the tree; rather, a posi- 2000, Taylor and Hellberg 2003). Our results indi- tion as sister to the Gonyaulacales has a high, albeit 418 HUAN ZHANG ET AL.

FIG. 5. Results of the AU test for assessing the phylogenetic positions of the fucoxanthin dinoflagellates, Crypthecodinium cohnii, and Symbiodinium spp. in broadly sampled ML host trees with (A–D) and without (E–H) Plasmodium spp. as the outgroup. (A) Placement of Karenia brevis and Karlodinium micrum in the 13-taxon tree using the cob þ SSU rDNA data set. (B) Placement of K. brevis and K. micrum in the 13-taxon tree using the cob data set. (C) Placement of C. cohnii in the 13-taxon tree using the cob þ SSU rDNA data set. (D) Placement of Symbiodinium spp. in the 13-taxon tree using the cob þ SSU rDNA data set. (E) Placement of K. brevis and K. micrum in the 11-taxon tree using the cob þ SSU rDNA data set. (F) Placement of K. brevis and K. micrum in the 11-taxon tree using the cob data set. (G) Placement of C. cohnii in the 11-taxon tree using the cob þ SSU rDNA data set. (H) Placement of Symbiodinium spp. in the 11-taxon tree using the cob þ SSU rDNA data set. The probability for each taxon placement is indicated with the branch thickness (see legends). nonsignificant, probability of P 5 0.583. Interestingly, considered an early branch of dinoflagellate evolution the divergent apicomplexans appear to have a strong by other analyses (Saunders et al. 1997, Litaker et al. effect on the position of C. cohnii, with the latter being 1999). Conflicting results still remain in more recent attracted to the former when they are included in the studies based on multiple protein encoding genes trees as an outgroup (Fig. 5C) but essentially not hav- (Saldarriaga et al. 2003, Leander and Keeling 2004). ing a single highly supported position when they are Our analyses of cob do not convincingly resolve this is- removed (Fig. 5G). Crypthecodinium cohnii has been sue, suggesting that more taxa (and/or data) are need- placed within Gonyaulacales based on characteristics ed to determine the divergence point of C. cohnii. of thecal plate tabulation and some SSU rDNA data Symbiodinium is a highly complex lineage whose (Fensome et al. 1993, Saldarriaga et al. 2001) but is members are resolved to eight phylotypes (clades) on PHYLOGENY OF DINOFLAGELLATES 419 the basis of rDNA and chloroplast DNA (Ishikura et al. Green 2002) and a plastid targeted GAPDH (unpub- 2004, LaJeunesse 2004), but few species have been lished data) gene of haptophyte origin in the nuclear formally described. The two strains (CCMP830, 832) genome of K. brevis (Ishida and Green 2002). Under analyzed in this study form a monophyletic group this scenario, the haptophyte psbO and GAPDH se- strongly supported in all our trees, although SSU quences, which are presumably of tertiary endosymbi- rDNA analyses indicate they belong to two different otic origin, replaced the nuclear encoded genes of red clades (B and A for CCMP830 and 832, respectively; algal (i.e. chromalveolate) provenance that are nor- unpublished data). However, affiliation of this lineage mally found in peridinin dinoflagellates (A. tamarense) to other dinoflagellates remains unresolved (Fig. 5, (Hackett et al. 2004b). Furthermore, a broader multi- D and H). A close evolutionary relationship between gene phylogenetic analysis of five minicircle proteins Symbiodinium and Gymnodiniales has been suggested now also strongly supports the haptophyte origin of on the basis of a similar degree of thecal plate devel- the dinoflagellate fucoxanthin plastid through tertiary opment (Steidinger and Tengen 1997). In contrast, a endosymbiosis (unpublished data). more recent tree based on actin and b-tubulin indicated an early divergence of this taxon (Saldarriaga et al. We thank D. M. Anderson (WHOI), P. Tester (National Ocean 2003), as seen in Figure 2B in our study. These data Services), D. K. Stoecker (Horn Point Environmental Labora- suggest that similarity in thecal plate development may tory), and R. Azanza-Corrales (Marine Science Institute, Uni- not reflect the degree of genetic relatedness and calls versity of the Philippines at Diliman) for providing cultures of for additional analyses including more species of Sym- Alexandrium tamarense, Pfiesteria shumwayae, Karlodinium micrum, and Pyrodinium bahamense, respectively. Brett Branco and John biodinium and related dinoflagellates to determine the Bean collected water samples from Mirror Lake, Storrs cam- precise divergence point of this important lineage. pus of the University of Connecticut, from which Ceratium sp. More studies are also needed to address the relation- was isolated. W. Litaker and P. Tester provided SSU rDNA se- ship among the phylotypes. The greater difference in quences for strain CCMP1828 and CCMP1835. T. N. Feinstein cob sequence between the two Symbiodinium strains than and K. Perkins provided assistance with growing the cultures, in other congeneric dinoflagellate species suggests a and L. Miranda measured the cell size of Akashiwo sp. The critical comments of two anonymous reviewers are also deeply possibility of their affiliation to different genera. appreciated. This research was supported by NSF grant DEB The possibility of resolving Symbiodinium into different 0344186 (to S. L. and H. Z.), NOAA-ECOHAB grants genera has been suggested (Rodriguez-Lanetty 2003, NA860P0491 and NA16OP2797 (to S. L.), and NSF grants LaJeunesse 2005). DEB 0107754 and MCB 0236631 (to D. B.). This is ECOHAB Phylogenetic status of gymnodiniales and plastid evolu- publication no. 121. tion of Dinophyceae. The most interesting result in this study is the clear support for the monophyly of the fucoxanthin-containing taxa K. brevis and K. micrum Avise, J. C. 1994. Molecular Markers, Natural History and Evolution. with the peridinin-containing Akashiwo sp. (Daugbjerg Chapman & Hall, New York, 511 pp. et al. 2000) within the Gymnodiniales. It has been pro- Barbrook, A. C. & Howe, C. J. 2000. Minicircular plastid DNA in the dinoflagellate Amphidinium operculatum. Mol. Gen. Genet. posed that the fucoxanthin-containing taxa may share a 263:152–8. common origin (Tengs et al. 2000), but their mon- Barker, F. K. & Lutzoni, F. M. 2002. The utility of the incongruence ophyly has received variable degrees of bootstrap sup- length difference test. Syst. Biol 51:625–37. port from analyses of nuclear and plastid SSU rDNA Cavalier-Smith, T. 1998. A revised six-kingdom system of life. Biol. Rev 73:203–66. (Daugbjerg et al. 2000, Tengs et al. 2000, de Salas et al. Conway, D. J., Fanello, C., Lloyd, J. M., Al-Joubori, B. M., Baloch, 2003). The evolutionary position of these dinoflagel- A. H., Somanath, S. D., Roper, C., Oduola, A. M., Mulder, B., lates relative to peridinin-containing taxa was therefore Povoa, M. M., Singh, B. & Thomas, A. W. 2000. Origin of not firmly established. In a study using the minicircle- Plasmodium falciparum malaria is traced by mitochondrial DNA. encoded plastid gene psbA, Yoon et al. (2002) suggested Mol. Biochem. Parasitol 111:163–71. Daugbjerg, N., Hansen, G., Larsen, J. & Moestrup, Ø. 2000. Phylo- that peridinin and fucoxanthin dinoflagellate might geny of some of the major genera of dinoflagellates based on share an ancestral haptophyte plastid that resulted ultrastructure and partial LSU rDNA sequence data, including from a tertiary endosymbiosis in their common ances- the erection of three new genera of unarmored dinoflagel- tor. Our results (given that cob is a vertically inherited lates. Phycologia 39:302–17. de Salas, M., Bolch, C. J. S., Botes, L. & Hallegraeff, G. M. 2003. marker of host phylogeny) contradict this latter hy- Takayama gen. Nov. (Gymnodiniales, Dinophyceae), a new ge- pothesis (Inagaki et al. 2004) and provide clear molec- nus of unarmored dinoflagellates with sigmoid apical grooves, ular phylogenetic evidence for the position of including the description of two new species. J. Phycol. fucoxanthin dinoflagellates as a derived clade within 39:1233–46. peridinin-containing taxa. Felsenstein, J. 2004. PHYLIP. Version 3.61. Department of Genet- ics, University of Washington, Seattle. The derived position of the fucoxanthin-containing Fensome, R. A., Taylor, F. J. R., Norris, G., Sarjeant, W. A. S., Gymnodiniales relative to peridinin-containing taxa Wharton, D. I. & Williams, G. L. 1993. A Classification of Living (including members of this order) strongly suggests and Fossil Dinoflagellates. Micropaleontological Species Publica- that the haptophyte-type plastid is a derived trait that tion 7, Sheridan Press, Pennsylvania, 351 pp. Gajadhar, A. A., Marquardt, W. C., Hall, R., Gunderson, J., Ariztia- was acquired in Karenia spp., K. micrum, and, putative- Carmona, E. V. & Sogin, M. L. 1991. Ribosomal RNA se- ly, Takayama spp. (de Salas et al. 2003). These results quences of Sarcocystis muris, Theileria annulata and Crypt- are consistent with the finding of a psbO (Ishida and hecodinium cohnii reveal evolutionary relationships among 420 HUAN ZHANG ET AL.

apicomplexans, dinoflagellates, and ciliates. Mol. Biochem. Pryer, K. M., Schneider, H., Smith, A. R., Cranfill, R., Wolf, P. G., Parasitol. 45:147–54. Hunt, J. S. & Sipes, S. D. 2001. Horsetails and ferns are a Galtier, N., Gouy, M. & Gautier, C. 1996. SEAVIEW and PHY- monophyletic group and the closest living relatives to seed LO_WIN: two graphic tools for sequence alignment and plants. Nature 409:618–22. molecular phylogeny. Computer Applications in the Biosciences. Rathore, D., Wahl, A. M., Sullivan, M. & McCutchan, T. F. 2001. A 12:543–8. phylogenetic comparison of gene trees constructed from Garesse, R. & Vallejo, C. G. 2001. Animal mitochondrial biogenesis plastid mitochondrial and genomic DNA of Plasmodium and function: a regulatory cross-talk between two genomes. species. Mol. Biochem. Parasitol. 114:89–94. Gene 263:1–16. Rodriguez-Lanetty, M. 2003. Evolving lineages of Symbiodinium-like Graham, L. E. & Wilcox, L. W. 2000. Algae: Prentice Hall Press, dinoflagellates based on ITS1 rDNA. Mol. Phylogenet. Evol. Upper Saddle River, NJ., 650 pp. 28:152–68. Gray, M. W., Burger, G. & Lang, B. F. 1999. Mitochondrial evolu- Rowan, R., Whitney, S. M., Fowler, A. & Yellowlees, D. 1996. tion. Science 283:1476–81. Rubisco in marine symbiotic dinoflagellates: form II enzymes Hackett, J. D., Anderson, D. M., Erdner, D. L. & Bhattacharya, in eukaryotic oxygenic phototrophs encoded by a nuclear D. 2004a. Dinoflagellates: a remarkable evolutionary experi- multigene family. Plant Cell 8:539–53. ment. Am. J. Bot. 91:1523–34. Saccone, C., Gissi, C., Lanave, C., Larizza, A., Pesole, G. & Reyes, A. Hackett, J. D., Yoon, H. S., Soares, M. B., Bonaldo, M. F., et al. 2000. Evolution of the mitochondrial genetic system: an over- 2004b. Migration of the plastid genome to the nucleus in a view. Gene 261:153–9. peridinin dinoflagellate. Curr. Biol. 14:213–8. Saldarriaga, J. F., McEwan, M. L., Fast, N. M., Taylor, F. J. R. & Hipp, A. L., Hall, J. C. & Sytsma, K. J. 2004. Congruence versus Keeling, P. J. K. 2003. Three protein gene phylogenies suggest phylogenetic accuracy: revisiting the incongruence length dif- that Oxyrrhis marina and Perkinsus marinus are early branches of ference test. Syst. Biol. 53:81–9. the dinoflagellate lineage. Int. J. System. Evol. Microbiol. 53:355–65. Huelsenbeck, J. P. & Ronquist, F. 2001. MrBayes: Bayesian infer- Saldarriaga, J. F., Taylor, F. J. R., Keeling, P. J. & Cavalier-Smith, T. ence of phylogenetic trees. Bioinformatics 17:754–5. 2001. Dinoflagellate nuclear SSU rRNA phylogeny suggests Inagaki, Y., Simpson, A. G., Dacks, J. B. & Roger, A. J. 2004. multiple plastid losses and replacements. J. Mol. Evol. 53:204–13. Phylogenetic artifacts can be caused by leucine, serine and Saunders, G. W., Hill, D. R. A., Sexton, J. P. & Andersen, R. A. arginine codon usage heterogeneity: dinoflagellate plastid 1997. Small-subunit ribosomal RNA sequences from selected origins as a case study. Syst. Biol 53:582–93. dinoflagellates: testing classical evolutionary hypotheses Ishida, K. & Green, B. R. 2002. Second-and third-hand chlorop- with molecular systematic methods. Plant Syst. Evol. 11(suppl): lasts in dinoflagellates: phylogeny of oxygen-evolving enhanc- 237–59. er 1 (PsbO) protein reveals replacement of a nuclear-encoded Schmidt, H. A., Strimmer, K., Vingron, M. & von Haeseler, A. plastid gene by that of a haptophyte tertiary endosymbiont. 2002. TREE-PUZZLE: maximum likelihood phylogenetic Proc. Natl. Acad. Sci. USA 99:9294–9. analysis using quartets and parallel computing. Bioinformatics Ishikura, M., Hagiwara, K., Takishita, K., Haga, M., Iwai, K. & 18:502–4. Maruyama, T. 2004. Isolation of new Symbiodinium strains from Serizawa, K., Suzuki, H. & Tsuchiya, K. 2000. A phylogenetic view tridacnid giant clam (Tridacna crocea) and sea slug (Pteraeolidia on species radiation in Apedemus inferred from variation of llantina) using cultura me´dium containing giant clam tissue nuclear and mitochondrial genes. Biochem. Genet. 38:27–40. homogenate. Mar. Biotechnol. 6:378–85. Shimodaira, H. & Hasegawa, M. 2001. CONSEL: for assessing the Jermiin, L. S., Graur, D., Lowe, R. M. & Crozier, R. H. 1994. confidence of phylogenetic tree selection. Bioinformatics Analysis of directional mutation pressure and nucleotide content 17:1246–7. in mitochondrial cytochrome b genes. J. Mol. Evol. 39:160–73. Steidinger, K. A. & Tengen, K. 1997. Dinoflagellates. In To ma s, Koumandou, V.L., Nisbet, R. E. & Howe, C. J. 2004. Dinoflagellate C. R. [Ed.] Identifying Marine Phytoplankton. Academic Press, chloroplasts—where have all the genes gone? Trends Genet. New York, pp. 387–594. 20:261–7. Swofford, D. L. 2003. PAUP* V4.0b8. Phylogenetic Analysis Using LaJeunesse, T. C. ‘‘Species’’ radiations of symbiotic dinoflagellates Parsimony (*and Other Methods). Version 4.0. Sinauer Associates, in the Atlantic and Indo-Pacific since the Miocene-Pliocene Sunderland, MA. transition. Mol. Biol. Evol. (in press) Taylor, M. S. & Hellberg, M. E. 2003. Genetic evidence for local Lang, B. F., O’Kelly, C. J. & Burger, G. 1998. Mitochondrial gen- retention of pelagic larvae in a Caribbean reef fish. Science omics in protists: an approach to probing eukaryotic evolu- 299:107–8. tion. Protist 149:313–26. Tengs, T., Dahlberg, O. J., Shalchian-Tabrizi, K., Klaveness, D., Lang, B. F., Seif, E., Gray, M. W., O’Kelly, C. J. & Burger, G. 1999. et al. 2000. Phylogenetic analyses indicate that the 19’-hex- A comparative genomics approach to the evolution of eukar- anoyloxy-fucoxanthin-containing dinoflagellates have tertiary yotes and their mitochondria. J. Euk. Microbiol. 46:320–6. plastids of haptophyte origin. Mol. Biol. Evol. 17:718–29. Leander, B. S. & Keeling, P. J. 2004. Dinoflagellates and apicom- Yang, Z. 1997. PAML: a program package for phylogenetic analysis plexans (Alveolata) as inferred from HSP90 and actin phylo- by maximum likelihood. Comput. Appl. Biosci. 13:555–6. genies. J. Phycol. 40:341–50. Yoon, H. S., Hackett, J. D. & Bhattacharya, D. 2002. A single origin Lin, S., Zhang, H., Spencer, D., Norman, J. & Gray, M. 2002. of the peridinin-and fucoxanthin-containing plastids in Widespread and extensive editing of mitochondrial mRNAs in dinoflagellates through tertiary endosymbiosis. Proc. Natl. dinoflagellates. J. Mol. Biol. 320:727–39. Acad. USA 99:11724–9. Litaker, R. W., Tester, P. A., Colorni, A., Levy, M. G. & Noga, E. J. Yoon, H. S., Hackett, J., Ciniglia, C., Pinto, G. & Bhattacharya, D. 1999. The phylogenetic relationship of Pfiesteria piscicida Crypt- 2004. A molecular timeline for the origin of photosynthetic operidiniopsoid sp. Amyloodinium ocellatum and a Pfiesteria-like eukaryotes. Mol. Biol. Evol. 21:809–18. dinoflagellate to other dinoflagellates and apicomplexans. Zhang, H. & Lin, S. 2002. Detection and quantification of Pfiesteria J. Phycol. 35:1379–89. piscicida using mitochondrial cytochrome b gene sequence. Maddison, D. R. & Maddison, W. P. 2002. MacClade. Version 4.05. Appl. Environ. Microbiol. 68:989–94. Sinauer, Sunderland, MA. Zhang, H. & Lin, S. 2003. Complex gene structure of the form II Mattern, M. Y. 2004. Molecular phylogeny of the Gasterosteidae: Rubisco in the dinoflagellate Prorocentrum minimum (dino- the importance of using multiple genes. Mol. Phylogenet. Evol. phyceae). J. Phycol. 39:1160–71. 30:366–77. Zhang, Z., Green, B. R. & Cavalier-Smith, T. 1999. Single Morse, D., Salois, P., Markovic, P.& Hastings, J. W. 1995. A nuclear- gene circles in dinoflagellate chloroplast genomes. Nature 400: encoded form II Rubisco in dinoflagellates. Science 268:1622–4. 155–9.