Apical Groove Type and Molecular Phylogeny Suggests Reclassification of Cochlodinium geminatum as Polykrikos geminatum

Traditionally Cocholodinium and Gymnodinium sensu lato clade are distinguished based on the cingulum turn number, which has been increasingly recognized to be inadequate for Gymnodiniales genus classification. This has been improved by the combination of the apical groove characteristics and molecular phylogeny, which has led to the erection of several new genera (Takayama, Akashiwo, Karenia, and Karlodinium). Taking the apical groove characteristics and molecular phylogeny combined approach, we reexamined the historically taxonomically uncertain species Cochlodinium geminatum that formed massive blooms in Pearl River Estuary, China, in recent years. Samples were collected from a bloom in 2011 for morphological, characteristic pigment, and molecular analyses. We found that the cingulum in this species wraps around the cell body about 1.2 turns on average but can appear under the light microscopy to be >1.5 turns after the cells have been preserved. The shape of its apical groove, however, was stably an open-ended anticlockwise loop of kidney bean shape, similar to that of Polykrikos. Furthermore, the molecular phylogenetic analysis using 18S rRNA-ITS-28S rRNA gene cistron we obtained in this study also consistently placed this species closest to Polykrikos within the Gymnodinium sensu stricto clade and set it far separated from the clade of Cochlodinium. These results suggest that this species should be transferred to Polykrikos as Polykrikos geminatum. Our results reiterate the need to use the combination of apical groove morphology and molecular phylogeny for the classification of species within the genus of Cochlodinium and other Gymnodiniales lineages.


Introduction
Gymnodiniales is a major order of dinoflagellates, which are ecologically important because many of its species, such as Cochlodinium polykrikoides, Karenia mikimotoi, cause harmful algal blooms. While their correct classification is crucial for understanding their varying ecological characteristics, the traditional genus delimitation criteria in the unarmoured order of dinoflagellates have proven to be problematic [1], [2]. Many genera were described during the 1800s or early 1900s, and the taxonomy is sometimes artificial and misleading. For instance, Gymnodinium Stein, Gyrodinium Kofoid and Swezy, Amphidinium, and Katodinium Fott were defined on the basis of the relative sizes of the epicone and hypocone [1], which is not reliable. Genus delimitations can be further obscured by similar morphological characteristiccharacteristics shared between two or more genera [2]. Since the beginning of this century, some researchers proposed alternative classification schemes to address some of the apparent taxonomic problems and to improve our understanding of the phylogeny within the order [1][2][3][4][5][6]. As a result, some genera (e.g., Gymnodinium Stein, Gyrodinium Kofoid and Swezy, and Amphidinium Claparede et Lachmann) have been redefined [1], [4], [7][8]. Furthermore, several new genera have been erected (Akashiwo, Karenia, Karlodinium, Takayama) and moved out of the initial Gymnodinium genus (sensu lato) based on molecular phylogeny and the characteristics of the apical groove. The combination of apical groove morphological and molecular phylogenetic studies has led to the uncovering of many relationships that were previously unknown [1][2][3][4][5][6][7][8][9][10][11]. However, some genera, such as the genus of Cochlodinium, are still defined based on the old genus discriminating criteria [12], [13]. It is necessary to re-examine species in Cochlodinium based on the combination of apical groove morphology and molecular phylogeny.
C. geminatum presents an excellent case calling for the reexamination of Cochlodinium taxonomy. This species was first described as Gymnodinium geminatum with its cingulum reported to wrap around the cell body 1.5 or 2.0 turns [13], [14]. After Schütt (1896) defined cingulum surrounding the cell at least 1.5 turns as the delimitating criterion of the genus of Cochlodinium, it was transferred to the genus Cochlodinium. It was reported in River Derwent, Triabunna, Australia in 2010 [15] and has been found to cause blooms in Pear River Estuary, China in recent years [16], [17]. Unlike C. polykrikoides, which caused mass mortality of fish while forming blooms, C. geminatum has not inflicted such ecological devastation even in its massive blooms in Pearl River Estuary, China. It is important to ascertain the taxonomic relationship between C. geminatum and C. polykrikoides for understanding their different ecological behaviors. In this study, we analyzed the ultrastructure, the major accessory pigment, and multi-gene phylogeny of the population to determine its relationship with other species in the Cochlodinium genus and other Gymnodiniales lineages. Our results clearly indicated that C. geminatum should be placed in the genus Polykrikos, reiterating that the traditional cingulum turn number is unreliable and should be replaced by apical groove morphotyping combined with molecular phylotyping for the classification of Cochlodinium and related species.

Materials and Methods
Sampling from the bloom event and isolation of C. geminatum cells Water samples were collected on August 25, 2011 at 22u9.89N, 113u37.59E near Zhuhai in the Pearl River Estuary, South China Sea, during a bloom event that lasted for .7 days. No specific permissions were required for sampling at this location. Meanwhile, our field studies did not involve endangered or protected species. One subsample was transferred into a 500-mL clean plastic container and preserved with Utermöhl's solution [18]. The sample was then stored in the dark at room temperature until microscopic morphological analysis (within 2 months). Six 2-mL subsamples were preserved in 2% formaldehyde and immediately frozen in liquid nitrogen and stored at 280uC for microscopic pigment fluorescence examination. Besides, a 50-mL water sample was filtered onto a 20-mm membrane and preserved in liquid nitrogen for subsequent HPLC analysis of pigment composition. In addition, a live sample was brought back to the laboratory in Guangzhou for microscopic observation of the mode of cell movement. In the laboratory, from the original Utermöhl's-fixed samples, twenty C. geminatum cells were isolated under an inverted microscope. The isolated cells were rinsed carefully with 0.45 mmfiltered seawater and used for subsequent DNA extraction.

Culture of North American C. polykrikoides and sample collection
A culture of C. polykrikoides originally isolated from Gulf of Maine (kindly provided by Don Anderson, WHOI) was grown in L1 medium [19] at 25uC under a 14:10 L:D cycle with a photon flux of 125620 mE s 21 m 22 . Cells were harvested using centrifugation at 30006g for 10 min at 4uC. The cell pellet was immediately stored at 280uC for subsequent DNA analysis.

Light microscopy (LM) and cell sorting
Cells of C. geminatum fixed in formaldehyde and Utermöhl's solution were observed under an Olympus BX51 epifluorescence microscope. Morphological identification of the species was done following Kofoid and Swezy (1921) and Schütt (1895Schütt ( , 1896 [12], [13], [14]. A subsample of the formaldehyde-fixed sample was incubated in SYBR Green I solution (35149A, Molecular probes, Invitrogen Corporation, Carlsbad, CA, USA, with 10 4 -fold dilution) at room temperature in the dark for ten minutes to stain DNA in the nucleus of the cell. The stained samples were examined under the epifluorescence microscope [20].

Scanning electron microscopy (SEM)
The Utermöhl's-fixed cells were allowed to settle by gravity, the supernatant decanted, and 1 ml of the 5% buffered glutaraldehyde fixative was added to the pellet as a post fixative. After 4 h at 4uC, the glutaraldehyde-fixed cells were allowed to settle by gravity again, and washed three times in 0.1 M phosphate buffer solution (PBS, pH7.2). Then 0.5 ml of 2% osmium tetroxide solution in the cacodylate buffer was added to the pellet as a post stain and fixative. After 12 h at 4uC, 0.5 ml of 2% osmium-fixed cells were allowed to settle by gravity again, washed three times by PBS and dehydrated in a graded ethanol /aqueous series (30%, 50%, 70%, 85%, 95% and 100% [twice] for 20 min each), acetone and isoamyl acetate (100%, twice for 20 min each). The cells were critical point-dried with CO 2 using a Tousimis Samdri 795CPD  (Rockville, MD, USA). Dried coverslips were amounted to aluminum stabs and then sputter coated with gold using a Denton Desktop 2 sputter coater. The coated cells were observed with a Hitachi S-3400N scanning electron microscope (Hitachi, Japan).

Transmission electron microscopy (TEM)
Ten ml of the samples fixed in 2% formaldehyde (final concentration) and stored 280uC were used for TEM analysis. The samples were centrifuged at 60006g for 10 min and the pelleted cells were resuspended in 2 mL of 5% buffered glutaraldehyde fixative and allowed to fix at 4uC for 48-hours. The re-fixed cells were gently centrifuged at 60006g for 10 minutes and resuspended in 0.5 mL of 2% osmium tetraoxide solution prepared in the cacodylate buffer as a post stain and fixative. After 12 h at 4uC, the osmium-fixed cells were sedimented by gravity, washed in PBS buffer, dehydrated in a graded ethanol/aqueous series and acetone, and then imbedded in SciPon resin in beam capsules at 70uC for 18 h. Ultrathin sections, collected on uncoated copper grids, were obtained with a Porter-Blum MT-2 ultramicrotome fitted with a diamond knife, poststained with Reynold's lead citrate, and observed under a JEM-100CXII transmission electron microscope (Jeol Inc., Tokyo, Japan).

Pigments analysis by high performance liquid chromatography (HPLC)
The 50-mL blooms water sample was gently filtered onto 25mm diameter 20 mm-pore size fiber filters and immediately stored in liquid nitrogen. Filters were subsequently transferred to 2.5 mL of methanol, sonicated for 30 s, and filtered (0.2 mm pore size, Millipore). One milliliter of this filtrate was mixed with 250 mL of distilled deionized water immediately before analysis. HPLC analyses were performed on a Shimadzu LC 10A system with a Supelcosil C18 column (25064.6 mm, 5 mm) using the method of Wright et al. (1991) [21]. The types of major pigments were identified by typical retention times and absorption spectra of the separated peak fractions for dinoflagellates.

DNA extraction, PCR amplification and sequencing of nuclear and mitochondrial genes
Twenty C. geminatum cells and a cell pellet containing ,1000 cells of C. polykrikoides were put into separate 1.5 mL microtubes and re-suspended in 0.5 mL DNA lysis buffer (0.1M EDTA pH8.0, 1% SDS, 200 mg mL 21 proteinase K) and incubated for 48 hours at 55uC. DNA extraction of both samples followed a previously reported CTAB protocol [22] and further purified using Zymo DNA Clean and Concentrator kit (Zymo Research Corp., Orange, CA). The final DNA elution step in Using 1 mL of the extracted DNA as the template, PCR was carried out to amplify fragment of ribosomal RNA (rDNA) gene cluster containing the majority of the small subunit rDNA (18S), complete internal transcribed spacers (ITS) 1, 5.8S rDNA, and ITS2 (ITS1-5.8S-ITS2), and a 59end fragment of the large subunit rDNA (28S). Several pairs of primers were used amplify each of these rDNA regions (Table 1), [23][24][25][26]. PCR was performed under the following condition: one initial denature step at 94uC for 3 min followed by 35 cycles of 94uC for 30 sec, 56uC for 30 sec, and 72uC for 45 sec, finalized by 10 min at 72uC for one cycle. PCR products were resolved by agarose gel electrophoresis and the bands with expected sizes were excised. DNA was purified and directly sequenced as reported [20].

Alignment and phylogenetic analyses
The DNA sequences obtained were used to search against GenBank databases using Basic Local Search Tool (BLAST) to determine what organisms these sequences represented. Thirty to 50 sequences showing significant similarity in BLAST to the sequences obtained in this study were retrieved from the databases. Phylogenies based on partial 18S, partial 28S (D1-D4, 1, 264 bp) and the concatenation of these two (18S+28S) were used to investigate the phylogenetic position of C. geminatum. These datasets were separately aligned using ClustalX. The alignments were run through ModelTest to select the most appropriate evolutionary model. The selected General Time Reversible (GTR) model with gamma distribution was employed for Maximum Likelihood analysis using PhyML3.0 aLRT [27]. Categories of substitution rates were set at 4, and other parameters were estimated based on the datasets. The proportion of invariable sites and gamma shape parameter were 0.267 and 0.514, respectively for the 18S dataset, 0.000 and 0.436 for 28S, 0.185 and 0.345 for 18S+28S.

Morphological characters
Cells of C. geminatum collected during the bloom occurred in pairs (Figure 1a (Figures 2, 4). SEM showed that the two cells were connected through a connection pore (3.2-3.6 mm in diameter) at the attachment point (with small a hingelike protrusion, Figure 2b, c) which was located near the apical loop of the cell on the bottom and the antapical area of the cell on the top (Figure 2a, c, e-g). Hypotheca exceeded epitheca: the epitheca accounted for 39% of the cell length and the hypotheca 61% on average (Figure 2). Observation of live samples showed that the two flagella moved in concert to enable the two cells to rotate around the longitudinal axis in unison ( Figure 2). Examination of fixed samples indicated that in each cell, the cingulum wrapped around the cell body by ,1.2 turn (Figure 2a), with the first half turn being horizontal and the second half deflecting toward the antapical end of the cell, with a displacement of 38-51% (47% on average, n = 10) of the cell's length ( Figure 2). Accordingly, the sulcus turned slightly, with 0.2 turn between the position of the top end and that of the bottom end. The sulcus was deeper than the cingulum, and its top end invaded the epitheca. The apical groove started from the top end of the sulcus and looped anticlockwise on the apical area forming a kidney beanshaped loop, which was not closed at the end but extended ventrally along side the sulcus.
Under the LM, SEM, and TEM, numerous chloroplasts were observed in the C. geminatum cells (Figure 3), occupying most of the intracellular space (Figure 3a, c, d). The chloroplasts were round shaped, 2.3-7.9 mm in diameter, with a mean of 5.261.9 mm (n = 25). Under TEM, we observed that the chloroplasts had a pyrenoid and a three-layer envelope (Figure 3f). HPLC analysis showed a peak with retention time and absorption spectrum typical of peridinin ( Figure 4) [28]. These results indicate that this species possesses a typical dinoflagellate chloroplast. The oval shaped nucleus is located at the center of the cells (Figure 3d).

Molecular phylogeny
We obtained sequences of a long rDNA fragment from sorted cells of a natural population of C. geminatum (3.522 kb; GenBank accession number JX967270) and the culture of C. polykrikoides (3.634 kb; JX967271) respectively. These sequences contained partial 18S, ITS1-5.8S-ITS2 (abbreviated as ITS hereafter), and partial 28S (D1-D4). For C. geminatum rDNA fragment, the 18S region spanned 1.555 kb, the ITS region 0.703kb, and the 28S region 1.264kb. BLAST analyses showed that C. geminatum rDNA sequences were most similar to the counterparts of the Gymnodinium sensu stricto (s. s.) clade. Furthermore, the 28 S sequence was almost identical (99%, differing by 10bp out of 0.85 kb) to that of C. geminatum isolated from Triabunna, on the East Coast of Tasmania, Australia [15], [29]. In contrast, the sequences obtained from the cultured C. polykrikoides were almost identical (99%) to those previously reported for C. polykrikoides, confirming correct species identification. However, the sequences from C. geminatum and C. polykrikoides differed markedly (,80% identity). Phylogenetic trees based on 18 S, 28 S and 18 S+28 S regions of rDNA were inferred separately, including 38, 40 and 42 (26 species) sequences respectively, representing all known clades of athecate dinoflagellate lineages. For each sequence region, the topologies for Neighbor Joining (NJ) and Maximum Likelihood (ML) trees were similar, consistently placing C. geminatum deep within the Gymnodinium s. s. clade, in sister relationship with Polykrikos with strong bootstrap support. In all these trees, C. geminatum was far separated from Cochlodinium spp. (including the C. polykrikoides strain we analyzed), which formed a distinct clade. In the 28S tree ( Figure 5), the Pearl River Estuary C. geminatum and the Australian C. geminatum were clustered together; these two C. geminatum were then grouped with Polykrikos hartmannii albeit with weak NJ support. In the 18 S tree (Figure 6), C. geminatum was also closest to P. hartmannii and other species of Polykrikos with moderate NJ and ML supports. Similarly, in the 18 S+28 S ML and NJ trees (Figure 7, Figure S1), C. geminatum was placed within the clade of Polykrikos with strong bootstrap support, showing a close relationship to P. hartmannii, but a far distance from C. polykrikoides.

Discussion
Through a careful examination of morphological, cytological and phylogentic data achieved in this study, we have obtained strong evidence that C. geminatum has been misplaced in the genus of Cochlodinium and should be moved back to the Gymnodinium s.s. clade, specifically to the currently valid genus Polykrikos. The result suggests that a combination of apical groove characteristics in combination with molecular phylogenies, instead of cingulum turn number which is currently employed, should be used to discriminate species in the genus Cochlodinium from those in Polykrikos.

Morphological evidence: Apical groove versus cingulum turn number
Previous studies have shown that the shape of the apical groove path, which should be viewed from the apical perspective for consistent characterization, appears to be a stable character in unarmored dinoflagellates, such as genera of Gymnodinium, Gyrodinium, Karenia, and Takayama [1][2][3], [30]. In this study, we confirmed that this character is more stable than the turn number of the cingulum. In some of our samples, depending on how they were fixed and from which angle they were viewed under the light and electron microscopes, cingulum can appear to be between ,1.10 and 1.25 turns. However, the apical groove exhibited a consistent morphology in all the specimens we examined (Figure 2). In sharp contrast to the apical groove in C. polykrikoides, which has a U-shaped trajectory ( Figure 6) [31], the apical groove of C. geminatum forms an anticlockwise open-ended loop of kidney bean- shape. The presence of an anticlockwise apical groove has been proposed to be a signature character of species in Gymnodinium s. s. clade [1]. Consistent with this, the species in Polykrikos that were previously assigned to Gymnodinium s. s. clade, such as P. hartmannii [32] P. schwartzii [33], and P. kofoidii [1], [34] have similar anticlockwise apical grooves ( Figure 6, Table S1). This is despite the fact that within Gymnodinium s. s. clade, the shape of groove differs slightly between some lineages. For instance, in G. impudicum [35], G. trapeziforme [29], G. catenatum [36], G. aureolum [37], [38], G. microreticulatum [39], and Lepidodinium viride [40] the apical groove takes an anticlockwise horseshoe shape ( Figure 6, Table S1). The apical groove of P. lebourae is an anticlockwise nearly closed loop ( Figure 6, Table S1) [41]. Whether the subtle difference in the shape of the apical groove represents differentiation at the genus level or species level requires further investigation.
Compared to the apical groove path, cingulum length is less reliable. For example, under SEM, we observed that the cingulum of C. geminatum wrapped around the cell body ,,1.56 turns (Figure 2a) while the turn number was hard to estimate accurately under the light microscope. However, the length of C. geminatum shown. If only one of the two phylogenetic methods yielded significant support, the other is shown with ''-''. Oxyrrhis marina was used as the outgroup to root the tree. Note that P. geminatum is embedded within the Gymnodinium s. s. clade with strong support, grouped with Polykrikos hartmannii. doi:10.1371/journal.pone.0071346.g005 cingulum was first described as 1.5 or 2.0 turns under the light microscope (Figure 8b; Schütt, 1895, pl.23, Figures 75 2 , 75 3 ) [14]. Other than cingulum length, the originally described morphology of C. geminatum is very similar to what we observed in the Pearl River Estuary population (Figures 2, 8a, b). Probably the inconsistency was due to confusion between the apical groove and the obliquely turning cingulum because before the 1920's, apical groove ( = acrobase) was not yet documented for unarmored dinoflagellates. The ''tail'' of the apical groove path obliquely descends toward the sulcus on the ventral side of the cell, which could be misunderstood as the beginning part of the cingulum under the light microscope. As a consequence, the number of cingulum turns might have been overestimated in Schütt (1895) (Figure 8b) [14]. Additionally, sample preservation often alters the morphologies of unarmored dinoflagellates, causing inconsistent estimates of the turn number of cingulum under the light microscope depending on the angle from which the cells are viewed, as noted in the present study.
Some cytological features we observed further support C. geminatum to be placed in the Gymnodinium s. s. clade, closely related to the genus of Polykrikos. The location of the nucleus and many small oval golden-brown (due to peridinin as the major accessory pigment) chloroplasts of C. geminatum are consistent with the type description of P. hartmannii [32]. Meanwhile, both C. geminatum and P. hartmannii have been documented as ''the connected paired cells ''.
In the Gymnodinium s. s. clade, most genera are single-celled while some form pseudocolonies [33], [34]. Polykrikos form colonies of 2 to 16 pseudo-cells (zooids) [32][33][34]. In most of Polykrikos species, there is no clear separation between neighboring zooids [32]. However, in C. geminatum, the two cells are clearly separate cells which are only connected at the attachment point recognizable by the connection pore on each cell (Figure 2). This feature places C. geminatum in an ''intermediate'' state between singlecelled and pseudocolony-forming lineages, i.e. probably a basal species in the genus of Polykrikos.

Molecular phylogenetic evidence: nuclear rDNA
Our molecular phylogenetic analyses based on 18 S, 28 S and 18 S+28 S rDNA concatenated sequences have provided definitive evidence that C. geminatum falls within the genus of Polykrikos (Figures 5, 6, 7, Figure S1) and is only distantly related to the genus Cochlodinium (Figures 5, 6, 7, Figure S1). These results place C. geminatum deep within the Gymnodinium s. s. clade which also includes the genera Polykrikos, Warnowia, and Gymnodinium [10], [32], [41]. In the 18 S phylogenic tree (Figure 6), C. geminatum is at the base of the Polykrikos cluster, consistent with the basal position posited above based on the cell connection mode. In the 28 S phylogenic tree ( Figure 5), our C. geminatum strain is allied with the Australian strain also identified as C. geminatum (EF616462) and both are grouped with P. hartmannii. The lack of other molecular data on the Australian C. geminatum strain [15], [29] has precluded further analysis on its relationship with our Pearl River Estuary population; however, its 28 S sequence is almost identical to our C. geminatum sequence, which along with morphological resemblance (Figures 2, 8) unequivocally shows that these two populations are the same species. Similarly, in the phylogenetic tree inferred from 18 S+28 S sequences, C. geminatum was closely related to P. hartmannii and far separated from the genus of Cochlodinium with strong NJ and ML support ( Figure S1, Figure 7).
In conclusion, C. geminatum has cytological features typical of the genus Polykrikos (especially to P. hartmannii, which is also a photosynthetic species), especially the anticlockwise open-ended loop of kidney bean-shaped apical groove. The affiliation of C. geminatum with the genus Polykrikos is consistently supported by molecular phylogenies. Therefore, it should be re-classified into the genus Polykrikos.
Re-emphasis of the apical groove -molecular phylogeny approach The historical taxonomic criteria, based on morphological characters such as cingulum turn number and vertical displacement have been shown not to be always reliable for delimiting genera in Gymnodiniales (including Cochlodinium) [1]. In its place, characteristic different shapes of apical grooves have been documented for Polykrikos, Gymnodinium, Cochlodinium and other genera in the unarmoured dinoflagellates [1], [2], [13]. In this study, we have confirmed that this apical groove-based distinction is more stable and discriminative when it applies to the case of P. geminatum. The combination of this cytological character (supported by other cytological features) with molecular phylogenies has helped unequivocally separate P. geminatum from the genus of Cochlodinium, with compelling evidence prompt us to transfer it to Polykrikos. Based on this case and historical studies, we further propose that classification of species within the genus of Cochlodinium, and perhaps all other Gymnodiniales genera, should be based on the shape of the apical groove trajectory (alone or in combination with thecal plate pattern) coupled with molecular phylogeny. This is consistent to what has been practiced for many genera in Gymnodinium sensu lato (e.g. [1], [2], [30], [32]). Figure S1 Neighbor-joining (NJ) phylogeny of P. geminatum ( = C. geminatum) with other dinoflagellates inferred from 18S+28S rDNA concatenated data. Sequence obtained in this study is bold-typed. Support of nodes is based on bootstrap values of NJ with 500 resamplings. Only values greater than 60 are shown. Oxyrrhis marina was used as the outgroup to root the tree. Note that P. geminatum is closest to P. hartmannii but far separated from the genus of Cochlodinium with strong support. (TIF) Table S1.

Supporting Information
(DOC)