J. Phycol. 42, 655–668 (2006) r 2006 by the Phycological Society of America DOI: 10.1111/j.1529-8817.2006.00228.x THE EVOLUTION OF ELONGATE SHAPE IN DIATOMS1 Andrew J. Alverson2 Section of Integrative Biology and Texas Memorial Museum, The University of Texas at Austin, 1 University Station, Austin, Texas 78712, USA Jamie J. Cannone, Robin R. Gutell Section of Integrative Biology and Institute for Cellular and Molecular Biology, The University of Texas at Austin, 1 University Station, Austin, Texas 78712, USA and Edward C. Theriot Section of Integrative Biology and Texas Memorial Museum, The University of Texas at Austin, 1 University Station, Austin, Texas 78712, USA Diatoms have been classified historically as ei- Key index words: 18S rDNA; Bacillariophyceae; ther centric or pennate based on a number of fea- centric; diatoms; Pennales; pennate; secondary tures, cell outline foremost among them. The structure; small subunit rDNA; Toxarium consensus among nearly every estimate of the dia- Abbreviations: BPP, Bayesian posterior probabili- tom phylogeny is that the traditional pennate dia- ty; GTR, General Time Reversible model of se- toms (Pennales) constitute a well-supported clade, quence evolution; I, proportion of invariable sites; whereas centric diatoms do not. The problem with ML, maximum likelihood; MP, maximum parsimo- the centric–pennate classification was highlighted ny; NJ, neighbor joining; TBR, tree bisection re- by some recent analyses concerning the phyloge- connection; C, gamma distribution netic position of Toxarium, whereby it was conclud- ed that this ‘‘centric’’ diatom independently evolved several pennate-like characters including an elon- gate, pennate-like cell outline. We performed sev- eral phylogenetic analyses to test the hypothesis Interest in the classification of diatoms dates back to that Toxarium evolved its elongate shape indepen- at least 1896 when diatoms with a round cell outline dently from Pennales. First, we reanalyzed the orig- (centrics) were distinguished from those with a long inal data set used to infer the phylogenetic position and narrow cell outline (pennates) (Schu¨tt 1896). Be- of Toxarium and found that a more thorough heu- yond cell outline, centric diatoms generally are ooga- ristic search was necessary to find the optimal tree. mous, and pennate diatoms generally are isogamous Second, we aligned 181 diatom and eight outgroup or anisogamous (Edlund and Stoermer 1997, Chepur- SSU rDNA sequences to maximize the juxtaposi- nov et al. 2004). These sexual characteristics have re- tioning of similar primary and secondary structure inforced the traditional centric–pennate split, and in of the 18S rRNA molecule over a much broader some cases, have taken precedence over cell outline in sampling of diatoms. We then performed a number the classification of taxa that are not unambiguously of phylogenetic analyses purposely based on dispa- centric or pennate (Hasle et al. 1983). The distinction rate sets of assumptions and found that none of between centrics and pennates persists largely out of these analyses supported the conclusion that convenience, and despite the fact that early specula- Toxarium acquired its pennate-like outline inde- tions about the diatom phylogeny suggested that cen- pendently from Pennales. Our results suggest that trics were not a natural evolutionary lineage (Simonsen elongate outline is congruent with SSU rDNA data 1979, Kociolek et al. 1989). Molecular phylogenetic and may be synapomorphic for a larger, more in- analyses consistently have shown that centric diatoms clusive clade than the traditional Pennales. grade into pennate diatoms, so only the ‘‘true’’ penn- ate diatoms (Pennales) are a monophyletic group (Medlin et al. 1993, 1996a, b, Medlin and Kaczmarska 2004, Sorhannus 2004, see Alverson and Theriot 2005 for a review). Because of this grade-like nature of relationships, 1Received 31 August 2005. Accepted 8 March 2006. some diatoms have a combination of ancestral 2Author for correspondence: e-mail [email protected]. (plesiomorphic) ‘‘centric’’ characters and derived 655 656 A. J. ALVERSON ET AL. (apomorphic) ‘‘pennate’’ characters. This perceived support the hypothesis that Toxarium acquired its elon- character conflict has obscured the higher level classi- gate outline independently from Pennales. fication of a number of elongate centric diatoms. For example, in establishing Cymatosiraceae, Hasle and ANALYSIS Syvertsen (1983) struggled as to whether the family Summary of Kooistra et al. (2003a) analysis. The should be considered centric or pennate. They exam- alignment combined 98 diatom SSU rDNA sequences ined numerous characters and ultimately concluded from Toxarium and other unspecified diatom taxa. that a majority of them affiliated Cymatosiraceae with Bolidomonas mediterranea and B. pacifica were used as other centrics. Two features traditionally associated outgroup taxa. This 100-taxon matrix was aligned with centric diatoms, flagellated gametes and develop- manually, and ModelTest 3.0 (Posada and Crandall ment from an annulus, were particularly important to 1998) was used to determine that the GTR þ G þ I their decision. model provided the best fit to their data set. A non- Toxarium undulatum Bailey is another diatom with parametric bootstrap analysis (1000 pseudoreplicates) features of both centrics and pennates. Toxarium is a on the 100-taxon matrix was performed. For each monotypic genus with a distinctly elongate cell shape, pseudoreplicate of the bootstrap analysis, the optimal though it lacks many of the other features traditionally tree was built with the neighbor joining (NJ) algorithm used to circumscribe Pennales. For example, poroids on distances corrected with the GTR þ G þ I model. are scattered on the valve face rather than being or- The resulting bootstrap consensus tree was used as the ganized around a longitudinal sternum and associated basis for an unspecified number of Kishino–Hasegawa transapical ribs (Round et al. 1990, Kooistra et al. tests (Kishino and Hasegawa 1989), with the goal of 2003a). Together, these two structures impose the or- identifying taxa that could be deleted without ‘‘im- ganization of poroids into striae aligned perpendicular pairing recovery of the phylogenetic position of to the longitudinal axis of the cell in Pennales (Round Toxarium’’ (Kooistra et al. 2003a, p. 191). These et al. 1990, Kooistra et al. 2003a). Sexual characteris- Kishino–Hasegawa tests were used to justify deletion tics of Toxarium have not been observed. Kooistra et al. of most radial centrics and pennates from the 100- (2003a) sought to resolve the classification of Toxarium taxon matrix because tree topologies with constraints (i.e. whether it is centric or pennate) through phylo- of Toxarium þ radial centrics and Toxarium þ Pennales genetic analysis of SSU rDNA sequences. They per- gave likelihood scores significantly worse than that of formed a number of phylogenetic analyses and the initial bootstrap consensus tree. All radial and mul- concluded that Toxarium represented independent evo- tipolar centrics separated by pairwise distances greater lution of several ‘‘true’’ pennate-like characteristics, than 0.2 were also removed (Kooistra et al. 2003a, including valve outline, which was described as ‘‘elon- p. 191). In the end, these two criteria were used to gate,’’ ‘‘pennate-like,’’ and ‘‘pennate’’ (Kooistra et al. delete 60 diatom taxa from the original 100-taxon 2003a, p. 186). We will use this convention, likewise matrix. ModelTest 3.0 again was used to determine recognizing that some members of phylogenetically that the GTR þ G þ I model provided the best fit to the isolated lineages have independently evolved elongate 40-taxon matrix. They performed a ML analysis and shape (e.g. Stephanodiscus rhombus Mahood). All refer- fixed the values of their model parameters to those ences to ‘‘pennate shape,’’ ‘‘pennate outline,’’ or ‘‘elon- output by ModelTest. No description of their heuristic gate outline’’ in this article describe an elongate, search was provided. They stated that a bootstrap biangular, and/or oval valve outline (Round et al. 1990). analysis on their initial 100-taxon matrix using full In this article, we test the hypothesis that Toxarium heuristic searches in ML ‘‘would take years’’ and cited evolved its elongate cell outline independently from this constraint as the primary reason for deleting 60 Pennales. First, we reanalyzed the 40-taxon data set of taxa from the matrix. Ultimately, however, NJ was Kooistra et al. (2003a). Second, because a growing lit- used to find the optimal tree in each pseudoreplicate erature emphasizes the critical importance of taxon of the bootstrap analysis, rather than heuristic searches sampling on the accuracy of phylogenetic inferences with ML. (Hillis 1998, Pollock et al. 2002, Zwickl and Hillis 2002, METHODS Hillis et al. 2003), we also analyzed 189 SSU rDNA se- quences (181 diatom, eight outgroup) aligned by max- Multiple sequence alignment. All SSU rDNA sequences from imizing the proper juxtaposition of homologous diatoms available before April 19, 2004 were obtained from GenBank for alignment (Table 1). Additional information primary and secondary structure. We analyzed this about the sequences, sequence alignment, and secondary alignment using two methods based on distinctly dif- structure analysis are available at: http://www.rna.icmb.utex- ferent assumptions as a rough control over robustness as.edu/PHYLO/SSU-DIATOM/. of our results to analytical details. We first used