Fast-Evolving Mitochondrial DNA in Ceriantharia: A Reflection of Hexacorallia Paraphyly?

The low evolutionary rate of mitochondrial genes in Anthozoa has challenged their utility for phylogenetic and systematic purposes, especially for DNA barcoding. However, the evolutionary rate of Ceriantharia, one of the most enigmatic “orders” within Anthozoa, has never been specifically examined. In this study, the divergence of mitochondrial DNA of Ceriantharia was compared to members of other Anthozoa and Medusozoa groups. In addition, nuclear markers were used to check the relative phylogenetic position of Ceriantharia in relation to other Cnidaria members. The results demonstrated a pattern of divergence of mitochondrial DNA completely different from those estimated for other anthozoans, and phylogenetic analyses indicate that Ceriantharia is not included within hexacorallians in most performed analyses. Thus, we propose that the Ceriantharia should be addressed as a separate clade.


Introduction
''DNA barcoding'' is the usage of a standardized DNA region not only as a tool for fast and reliable identification of known species, but also to assist the detection of undescribed species [1]. Researchers utilize a short DNA sequence (,700 bp) from the mitochondrial protein-coding gene cytochrome c oxidase subunit I (COI) that differs by several percent between even closely metazoan related species as an adequate ''barcode'' to distinguish species. Since then, the COI region has been widely used for DNA barcoding in many Metazoa, including cnidarians (see more in [2]). However, within cnidarians, the effectiveness of this marker as a species-level barcode is purported to be limited to Medusozoa [3] as it has levels of divergence (Kimura 2-Parameter, K2-P) within congeners greater than 20% in Medusozoa [2], while this divergence hardly exceeds 5% within anthozoan congeners [3][4]. Furthermore, other mitochondrial DNA markers broadly usedsuch as mitochondrial 16S rDNA -also have divergence values among anthozoan congeners significantly lower than those verified for medusozoans [3]. As such, it appears that the anthozoan mitochondrial molecular clock, most probably as a result of a mismatch repair system -e.g. a MutS-like protein encoded by the mt genomes of a number of octocorals [5][6][7][8][9], is slower than in most Metazoa. This low evolutionary rate has been inferred as an ancestral condition [3]; [10][11], although there is no evidence of this mismatch repair system in Hexacorallia mitochondrial genomes, or in other eukaryotes [12].
Within the ''orders'' that traditionally compose the Hexacorallia, only Ceriantharia does not have a representative with a complete mitochondrial genome determined to date. Phylogenetic studies of cerianthids based on mitochondrial and nuclear markers have resulted in reconstructions divergent from those based on morphology [11]; [13][14][15]. Thus, Ceriantharia is an ''incertae sedis'' group [13] within Anthozoa. In general, the phylogenetic reconstruction of higher taxonomic levels (i.e. class, order) demands molecular markers with ''low'' mutation rates (e.g. less than 1% divergence/million years) [16]. Therefore, the long (28S) and short (18S) nuclear ribosomal genes have been purported to be adequate [17] if they are conserved enough to produce unambiguous alignments and to provide an appropriate phylogenetic signal to define basal relationships [18]. In this study, nearly complete sequences from 18 and 28S and partial 16S ribosomal genes were used to reconstruct the cnidarian evolutionary history, and specifically focus on the relative position of cerianthids within the phylum. In addition, the genetic divergences of mitochondrial markers (i.e. partial 16S and COI sequences) were defined, and compared between Ceriantharia and other cnidarian groups, indicating a ''fast-evolving mtDNA profile'' (based on medusozoan data) in the former. In general, this study adds to the development of a more cohesive evolutionary scenario for Ceriantharia, defines a phylogenetic framework for Ceriantharia systematics, and proposes a new evolutionary scenario for the mutation history of mtDNA in the non-bilaterian Metazoa.

DNA Extraction, Polymerase Chain Reaction and Sequencing
Total DNA was extracted from individual cerianthid specimen tentacles (Table 1) using InstaGeneH (Bio-Rad #732-6030) or the DNAdvanceH kit (AgencourtH #A48705). PCR reactions and conditions followed under predefined conditions ( [13]; primers used in the present study are listed in Table 2). Amplicons were purified using AMPureH kit (AgencourtH #A63881) following manufacturer's instructions, and made ready for sequencing using the BigDyeH Terminator v3.1 kit (Applied Biosystems #4337455; same primers and Tm temperature conditions as in PCR reactions). Sequencing was carried out on an ABI PRISMH3100 genetic analyzer (Hitachi), and resulting sequences were assembled and edited using Geneious TM 5.4.4 (Table 1). Ceriantharia sequences from mitochondrial (COI, 16S) and nuclear (18S, 28S) molecular markers were obtained from 12 of the 50 taxonomically recognized species [28].

Mitochondrial Evolutionary Divergence between Ceriantharia Species and Genera
Evolutionary distances of mitochondrial (16S rDNA and COI) genes from Ceriantharia were analyzed within congeners (Isarachnanthus -see Table 3) and between genera in MEGA5 software [29], using Kimura's two-parameter model of base substitution (K2-P) in order to calculate their respective genetic distances. In addition, K2-P genetic distances were also estimated for other anthozoan clades, and hydrozoans were considered as ''outgroup'' for discussion. The distances obtained for both mitochondrial markers are compiled in four graphs in two figures, of which one are from congeners (Figure 1), and another from species of different genera ( Figure 2). The complete datasets, including sequences obtained from GenBank, are compiled in the supplementary material (Table S1).
To analyze the phylogenetic signal profile along the recovered phylogenies in figure 3, the net approach test was performed using PhyDesign ( Figure 4; [44]). Finally, to test alternative systematic proposals and our own main results, considering the phylogenetic position of Ceriantharia, we computed AU and other phylogenetic hypothesis tests using consel ( [45]; same input datasets and results from the MAFFT+RAxML analysis; Table 4).

Mitochondrial Genetic Diversity
While the genetic distance estimates for COI and 16S within Ceriantharia congeners ranged from 3 to 17% and 3 to 12%, respectively (Figures 1 and 2), values for all other anthozoan groups excepting Octocorallia were estimated to be less than 1% (COI) to 2% (16S). Octocorallian congeners slightly exceeded these values, with approximately 2% and 4% divergences within COI and 16S sequences, respectively. In general, the divergence values estimated for ceriantharians showed a range most similar to those from hydrozoans (medusozoan example) (intrageneric divergence: COI/16S 3-17%/3-12% (Ceriantharia) and 4-23%/2-14% (Hydrozoa)). Furthermore, the estimated COI genetic divergence between genera in every major anthozoan group (Octocorallia and Hexacorallia) was lower than the divergence observed for ceriantharian congeners. Most anthozoans had intergeneric divergences below 10% while the estimated divergence in ceriantharians was between 14 and 22% ( Figure 3). However, the calculated 16S divergence between species of different genera showed less difference between Ceriantharia and other anthozoans, as observed values in Ceriantharia ranged from 7 to 23%, while other anthozoan groups ranged from 0 to 7%.
An example of the genetic divergence of related congeneric species is the divergence values of Isarachnanthus spp. in the Atlantic Ocean ( = 6-9%), which are higher than levels reported from similar comparisons in other anthozoans. The estimated divergence between several specimens of I. nocturnus and I. maderensis was based on multiple specimens and is considered reliable (Table 3). Both 18S and 28S genes recovered monophyletic Anthozoa and Meduzozoa with a sister relationship in almost all results (Figures 3  and 4), although the Hexacorallia monophyly suffered due to Ceriantharia's relative position. In most 18S reconstructions, all anthozoan and medusozoan lineages (i.e. Octocorallia, Hexacorallia, Scyphozoa, Staurozoa, Cubozoa, and Hydrozoa) were recovered as monophyletic groups. However, Ceriantharia was recovered as a sister group to Anthozoa, or in some cases, as a sister group of Cnidaria (Table 4). Ceriantharia, which has been taxonomically placed within Hexacorallia, was recovered as such only with dynamic homology with low CI and RI indexes if compared to other MP-POY topologies from the same dataset. A sister-like relationship between Ceriantharia and Octocorallia was also recovered in MP topologies where gaps were treated as missing data (TNT analysis), or in ML and MP with filtered datasets (RAxML and POY results; Table 4; except for these cases, Ceriantharia was recovered as a sister group to Cnidaria).

Phylogenetic Position of Ceriantharia
The monophyly of both Anthozoa and Medusozoa, and their respective major component groups, were also recovered using 28S rDNA ( Figure 3, Table 4). However, in general, the main topology differed from that of 18S rDNA as Ceriantharia occupied a sister-like position to Octocorallia, with high statistical support in all but non-parametric bootstraps. MP topologies that recovered Ceriantharia as sister group of Hexacorallia (traditional placing) had several unexpected results -for example, non-monophyletic Scyphozoa and Staurozoa as sister groups of Hydrozoa. As in 18S rDNA, these MP topologies had lower CI and RI values compared to results using different methods (Table S2). The 16S dataset including two ctenophore sequences had a loss of monophyly of the Medusozoa group and some well-recognized cnidarian clades (e.g., Hydrozoa), with Ceriantharia grouped with the rest of the anthozoan lineages (but not as sister group of the rest of Hexacorallia), possibly representing a case of basal long branch attraction for portions of the sampled Hydrozoa and Ctenophora sequences. For the 16S dataset we rooted the 16S phylogeny considering Medusozoa and Anthozoa as monophyletic groups, and this analysis had highly similar results as to the 18S ML analysis and rooted with ctenophore sequences (Figure 3). It is important to note that the late definition of rooting does not affect analyses in ML studies nor phylogenetic relationships between the terminal trees. Due to better support values and more similar clade relationships for major cnidarian lineages with 18S and 28S results, we focused on the 16S dataset and their ML result that did not include ctenophoran sequences for Consel and PhyDesign analysis. Consel recovered low support (negative significance value -less than 0.05 of p-value) for Ceriantharia as sister group of Hexacorallia compared to alternative systematic scenarios in both 18S and 16S analyses (Table 5). Based on PhyDesign outcomes, it is believed that the 28S rDNA has more phylogenetic signal than the 18S rDNA, and consequently is more adequate for Cnidaria higher rank reconstruction ( Figure 4). However, it is important to note that these results may be biased due to the different number of species included in each gene dataset (and consequently different number of epochs in each topology; Figure 4). Overall, our results from both nuclear and mitochondrial ribosomal gene analysis corroborate that Ceriantharia is a monophyletic and basal independent lineage, but indicates that it does not have a direct relationship to other hexacorallian orders. Thus, these results suggest that Ceriantharia should be ranked as a separate subclass within Anthozoa.  Table S2 for details in datamatrices, software and parameters). doi:10.1371/journal.pone.0086612.g003 Fast-Evolving Mitochondrial DNA in Ceriantharia PLOS ONE | www.plosone.org

Anthozoan Phylogenetics and Genetic Divergence among Cnidarian Species and Genera
According to [3], ''the tempo of evolution in anthozoan mitochondrial genes appears to be at least 10-20 times lower than the standard mitochondrial clock based on vertebrate sequences, which averages a sequence divergence of 1-2%/Myr [46][47][48][49]…''. With the use of a number of statistical methods, [50] also found that the anthozoan intra-and interspecific COI variations were #6% or invariant (in 98% of the analyzed species). The genetic divergence of ceriantharian mitochondrial markers would therefore be expected to follow the slow rate of differentiation as detected in other groups currently classified as anthozoans [51]. However, an overview of our results suggests a much faster rate. For both analyzed mitochondrial markers (COI and 16S rDNA), the genetic divergence rates estimated between Ceriantharia congeners are more similar to those from Hydrozoa than to Anthozoa, suggesting that the mitochondrial clock in Ceriantharia has almost the same divergence rate as in Medusozoa and other Bilateria groups (see also [13]). Therefore, the above statement of slow mitochondrial evolution in Anthozoa is accurate only if Ceriantharia is not included within Anthozoa classification. In fact, this hypothesis is not unusual. Based on nuclear molecular markers, [15] and [11] showed that Ceriantharia is not within Hexacorallia. Another study also stressed that Ceriantharia could be a hexacorallian outgroup [52], but its position was not further discussed. In addition, analyses of complete mitochondrial genomes indicate that the phylogenetic position of Ceriantharia is unstable and not well defined (e.g. sister group of Hexacorallia) [53]. For discussion on the historical context of molecular data and Ceriantharia's evolutionary position over the last twenty years see (Text S1).
More recently, however, a new body of data is changing the scenario previously proposed for the origin of the rapid evolution of mtDNA [50]. In summary, there are two most parsimonious explanations for the slow mtDNA evolution in anthozoans (see Figure 2 in [50]). However, both scenarios must be modified to accommodate the Ceriantharia results presented herein. The first scenario ( Figure 5, reconstruction A) hypothesizes independent origins for the rapid evolution of mtDNA in Medusozoa and Bilateria. However, it is necessary to include an additional step (origin of the rapid evolution of mitochondrial DNA) before Ceriantharia, and consequently this reconstruction is no longer most parsimonious. In the second scenario ( Figure 5, reconstruction B) there is no necessity to add an extra step, making it the most parsimonious scenario; namely that anthozoans (not including Ceriantharia) decreased their mtDNA rate of evolution.
Corroborating the findings from [15], the phylogenetic ''uniqueness'' of Ceriantharia in relation to anthozoans has major implications not only in basal metazoan mtDNA evolution but also in cnidarian evolutionary history. The inclusion of Ceriantharia data in molecular clock phylogenetic reconstructions may influence estimates for the appearance of Cnidaria. It may also have effects on the stability of the position of Cnidaria in broader studies, where the monophyly of the group (and major lineages, like Anthozoa and Medusozoa) are compromised by potential incomplete taxon sampling [53][54][55]. From the current state of knowledge, it is possible to infer that Ceriantharia does not belong to Hexacorallia and thus represents a new subclass. However, the exact phylogenetic position of Ceriantharia is still debatable. Our study indicates that, based on a broader sampling of the group, Ceriantharia is most likely a sister group to Anthozoa. However, a sister-like relationship between Ceriantharia and Octocorallia cannot be discarded and further analyses are needed to clarify this matter.

Futures Prospects on Cnidarian mtDNA Evolutionary Genetics
A better understanding of the mechanisms related to the shifts of mtDNA evolutionary rates is a challenging task in current cnidarian studies [56][57][58]. There are several possible historical influences on qualitative aspects of mtDNA on this group, such as genome linearization and fragmentation, horizontal gene transfer from a non-bilaterian species (e.g., HGT, mtMutS gene), and gene arrangement (e.g., [59][60][61]). Recombination phenomena have been invoked to explain gene rearrangements in Octocorallia, and plasmid insertion to explain mtDNA linearization in Medusozoa [60]. The connection between genome components and rearrangements with mutation rates is not a trivial matter [62][63][64][65], as it is proposed to have been involved as one of the main mechanisms on evolutionary genetics of linear genomes [66]. Nonetheless, the theory of a mtDNA mismatch repair gene in Octocorallia as the mechanism explaining slow rates of genetic divergence does not explain the slow rates exhibited in Hexacorallia (Figures 1-3). This consideration reinforces the ''classic'' relationship proposed between Octocorallia and Hexacorallia lineages, differing from phylogenetic analyses based on mitochondrial data linking Octocorallia as sister group of Medusozoa (Anthozoa not monophyletic: [53]; [67]. Thus, more extensive studies on Table 4. Phylogenetic results considering all the analyzed ribosomal datasets and methods, showing the recovery of major clades in Cnidarian systematics proposals in our topologies and literature.  mitochondrial genome arrangement and how it is related to mutation/evolutionary patterns are needed in Hexacorallia. Additional efforts should also be undertaken to increase ceriantharian taxon sampling, evaluation of different cnidarian outgroups (e.g., Porifera vs Ctenophora), together with a thorough evaluation of (1) data treatment to define homology, (2) reconstruction parameters and methods, (3) nodal support, (4) hypotheses testing and the (5) recovered phylogenetic signal (e.g. [68][69]; [34]).
Recently [70] suggested that almost 90% of the Ceriantharia species are already known. However the authors neglected to address the existence of cryptic species, species with disjunct distributions (e.g. Ceriantheomorphe brasiliensis, Ceriantheopsis americanus), and oceanic areas with no data (e.g. deep sea); additionally, the evolutionary position of Ceriantharia should be clarified because of their possible reflection of basal adaptive radiation in cnidarians, considering their asymmetric conditions on sister clades, based on (1) number of extinct derived species, (2) genetic divergence and (3) morphological-ecological traits, that altogether could represent signals of adaptative events [67] [71]. Finally, we emphasize that the inclusion of Ceriantharia in Hexacorallia should no longer be accepted and this group should be elevated to subclass.
Thus the class Anthozoa should be divided into three subclasses; -Hexacorallia, -Octocorallia -Ceriantharia subclass nov. Figure S1 ML cladogram of Cnidaria based on a combined dataset (18S+16S ribosomal molecular markers partitions). Supports values are aBAYES and aLRT (parametric, upper branch) together with BS and SH-like (non parametric, lower branch) in clockwise direction. Each dataset was aligned in MAFFT and phylogeny estimated in RAxML (independent branch length calculated for every molecular marker partition); support values were calculated in RAxML (BS) and PhyML. Data treatment and basic parameters (e.g., number of replicates) were similar to individual ML analysis for 18S dataset (MAFF+RAxML analysis; see Table S2 for details in datamatrices, software and parameters).  Figure B) were analyzed. Trying to overcome molecular saturation (non-phylogenetic related heterogeneity), the COI dataset was filtered at two different intensity levels: treating gaps as missing data (''Aliscore -N'' strategy; Figure C) and a more intense approach (''Aliscore -N -r -w4''; Figure D). The COI dataset (Genbank IDs presented in terminal's names) was originally aligned in MAFFT (codon frame checked); the root position was defined a posteriori (random position; no effect on ML analysis). (TIF)  Table S2 Software and analytic details (data treatment and parameters) for every analysis applied in this study for 18S and 28S datasets (see Table 4 for major results compilation).

Supporting Information
(XLSX) Text S1 Discussion on evolution and previous systematic considerations of Ceriantharia and Anthozoa. (DOCX)