Millions of RNA sequences (as DNA) have been deposited in public databases such as Genbank (NCBI), and over one hundred thousand of these correspond to genes such as 18S that are widely used in phylogenetic studies. Although functional RNA sequences fold themselves to form secondary structures that depend sensitively on the primary sequence, only a handful of phylogenetic studies have taken advantage of this secondary information. Most phylogenetic studies using current methods have focused on primary sequence information, but RNA secondary structures are particularly useful in systematics because they include characteristics, not found in the primary sequence, that give "morphological" information for reconstructing the complete tree of life 1. Ribosomal gene secondary structures have conserved zones that make it easier to reconstruct the structures of unknown homologous RNA sequences by direct comparison. Although molecular phylogenetic inference relies heavily on single-copy genes, ribosomal genes that show concerted evolution still offer one of the best alternatives in the genome. The aim of this paper is to demonstrate the value of RNA secondary structures in improving and enhancing phylogenetic inference for a group of marine invertebrates.

Despite progress towards the resolution of octocoral phylogeny using mitochondrial gene sequences 2, there is no consensus about a sequence region that provides enough resolution for closely-related species relationships. The most variable regions known so far in octocorals are the Internal Transcribed Spacers (ITSs), which have provided good resolution among a few soft coral species that would be very difficult to distinguish morphologically because their characters show excessive homoplasy 34. However, the ITS region shows saturation and in some cases many INDELS (insertions-deletions) in an alignment, which can produce misleading results for high taxonomic ranks 5. Opportunely, analyses of the predicted RNA secondary structures of some genes such as 16S 4 and ITS2 6 have provided an alternative way of increasing the phylogenetic signal of RNA sequences for octocorals. Here, we studied the predicted RNA secondary structures of ITS2 to resolve close-species relationships within the gorgonian genus Eunicea (candelabrum corals).

ITS2 (Internal Transcribed Spacer 2) sequences, found in the tandem arrays of the nuclear ribosomal RNA between the 5.8S and 28S genes, have not been considered useful for molecular systematics in certain invertebrate groups mainly because of their excessive INDELS and/or intragenomic variation 78. Nonetheless, they seem to afford a promising nuclear region for lower eukaryotes, particularly at the intra-familial level and down to closely-related species, where other known regions are nearly invariant 9101112. In scleractinian corals, except for the genus Acropora, ITS sequences give reliable phylogenetic information especially if the predicted RNA secondary structures are compared 6. These sequences vary considerably within octocoral genera, providing information for classifying a recently-described species, Alaskagorgia aleutiana Sánchez and Cairns, within the Plexauridae; this was uncertain on the basis of morphological characters, but the results were consistent with mitochondrial DNA analysis 135. Variation was also found within the Gorgoniidae family; ITS2 secondary structures retained and classified different gorgonian genera 5. Perhaps they are among the most variable genomic regions in these organisms, while mitochondrial DNA and other nuclear sequences are highly conserved 101114. Unfortunately, ITS alignment always produces multiple and variable INDELS depending on the different gap opening and/or extension penalties (e.g. ClustalW), which make phylogenetic reconstruction somewhat unreliable.

The known predicted ITS2 RNA secondary structure is a model with a common core in most eukaryotes 15. It is a large-scale marker that is not limited to a specific taxonomic level and has even been used in mega-systematics analysis 15. Among other advantages, molecular morphometrics makes it possible to identify characters informative about parsimony that are not found in the primary sequence 1617, and to correct primary alignments on the basis of secondary structure information 3. Could the predicted ITS2 RNA secondary structures help to overcome the great difficulties in the taxonomy of lower invertebrates such as octocorals?

Caribbean octocorals are very abundant in coral reefs with up to 60 species in a single location, and many species and genera show few or no morphological differences 18. Gorgonian octocorals from the genus Eunicea Lamouroux, 1816 (Cnidaria; Octocorallia; Alcyonacea [=Gorgonacea]; Plexauridae), known as "sea candelabrum", are endemic to the Caribbean region (Tropical Western Atlantic Ocean excluding the Brazilian region). There are 13 valid species, although more than 30 species have been described in, or assigned to, Eunicea 1819. No phylogeny for the Eunicea genus has been constructed on the basis of morphological or molecular characters, and this could be important for testing the monophyly of the internal subgenera 19 and other morphological characters that incur the risk of confusion with Plexaura octocorals 18. In this study, in order to differentiate between Eunicea and Plexaura gorgonian corals using molecular and morphological data, we present a phylogeny based on the predicted secondary structures of the ITS2 sequences (molecular morphometrics) and contrast it with traditional molecular systematics based on primary DNA alignment and Bayesian analysis.

The ITS2 nuclear region ranged from 252 bp in Eunicea laciniata to a minimum length of 185 bp in Plexaura kuna. Twelve new predicted RNA secondary structures were constructed for this region, all following the 6 helicoidal ring model except for E. pallida, which lacked the sixth helix (Fig. 1I). Two major common features were present in these structures: (1) 5' G-C, R-R, G-C, G-C, bulge on helix IV and (2) 5' GUGC, bulge, CAAGG with its complementary pair base on helix V (Fig. 1).

Three different phylogenetic analyses were obtained from the ITS2 sequence alignment. The molecular morphometrics analysis relied on 42 parsimony-informative characters and resulted in one most parsimonious tree using ordered Wagner parsimony (Length = 270; Retention index = 0.57). Both bootstrapping and decay indices showed strong support for Plexaura and Eunicea as reciprocally monophyletic groups (95% and 9, respectively) with the exception of E. flexuosa (new combination from Plexaura flexuosa, see details below), which was clearly a derived species of Eunicea (Fig. 2A). The topology based only on the predicted RNA secondary structure of the ITS2 region resolved most relationships among the species studied (Fig. 2A). Maximum likelihood analysis using traditional primary sequence alignment retained one tree with the assumed model (TVMef+G) selected by AIC. The tree searches had six substitution rates (A-C, 1.2431; A-G, 3.0064; A-T, 1.0809; C-G, 0.3376; C-T, 3.0064; and G-T, 1.0000), not assuming the proportion of invariant sites, and a gamma-shape parameter (1.413); node support was examined by 1000 bootstrap replicates (Fig. 2B). Bayesian analysis of the same alignment retained the same topology and supported the same branches (data not shown). This last topology differed in two main respects from the molecular morphometrics tree: the placement of E. mammosa within the outgroup P. kuna Lasker, Cofforth and Kim, with low bootstrap and Bayesian clade credibility values (< 60%), and of E. fusca Duchassaing and Michelotti as a derived instead of a basal species (Fig. 2A–B). The Bayesian inference in which helices were to evolve independently at different rates in a mixed matrix secondary structure was similar to the molecular morphometrics tree though less resolved within Eunicea (Fig. 2C). Eunicea flexuosa (Lamouroux), E. pallida Garcia-Parrado and Alcolado, E. laxispica (Lamarck, 1815) and E. mammosa Lamouroux formed a separate clade in the molecular phylogenies (except for E. mammosa in the maximum likelihood tree), most probably the Eunicea subgenus 19, and were reciprocally monophyletic with respect to other Eunicea (Euniceopsis subgenus : E. laciniata Duchassaing & Michelotti, E. tourneforti Milne Edwards and Haime, and Eunicea spp. 1–2), except for E. fusca in the molecular morphometrics tree (Fig. 2). The predicted ITS2 RNA secondary structure models showed that E. mammosa, E. pallida, E. laxispica and E. flexuosa exhibited a long helix IV (Fig. 1A, E, H, I) that could be a sinapomorphic feature of the Eunicea subgenus. Plexaura spp., in contrast, lacked that helix [see Additional file 1].

Traditionally, morphological identification of Eunicea and Plexaura species has been based mostly on microscopic characteristics 19. Morphological variation within species can overlap with variation among species, so molecular identification has become an important alternative for octocorals in general. ITS2 has proven to be a good molecular marker, resolving octocoral species at different taxonomic levels 5. In particular, the ITS2 sequences and predicted RNA secondary structures have afforded a new view of the phylogeny of the Eunicea genus. This approach supports the divergence of Eunicea flexuosa comb. nov. from Plexaura, although it has been considered part of this outgroup for nearly two centuries because of morphological resemblance. Nonetheless, the morphological observation of two subgenera in Eunicea 19 seemed congruent with the molecular results.

The ITS2 RNA secondary structures yielded homologous models that grouped conserved species features in the Eunicea subgenus (e.g. long helix IV), though it is important to note that these molecular structures can also produce homoplasies because of selective pressures acting on particular helix features such as the 6-helicoidal ring core, conserved among most eukaryotes. ITS2 structures are important for proper pre-rRNA maturation; any change in their helix conformation can affect the process of 25S production 21. Each helix was partitioned as if it had evolved at a different rate and substitution bias, because they do not all have the same role in this dynamic conformation 22. This approach resulted in a supported Bayesian inference phylogram very similar to the molecular morphometrics approach. The major difference between the topologies obtained by the Bayesian (e.g. molecular morphometrics using partitions) and primary sequence (e.g maximum likelihood) methodologies, was the placement of E. mammosa in and out of Eunicea, which could be due to the lack of phylogenetic signal and the likely saturation of ITS2 sequences after multiple alignment (e.g. 5).

The combined molecular-morphological Bayesian approach data matrix has not been widely used, mainly because until recently the methods for combining datasets were only applied to parsimony-based analysis, which has been shown to be insufficient when the combined data imply different evolutionary scenarios 23. This type of analysis, combining molecular data with "typical" morphological characteristics, has sometimes failed to yield a satisfactory model of character evolution 24. Here, we combined the corrected primary alignment (e.g. aligning sequences from each helix only) with the molecular morphometrics data. Traditionally, some problems in explaining changes in character state arise from assuming that each character has exactly the same state at a particular time, in opposition to punctuated equilibria. In molecular morphometrics, the possibility of change is based in the number of nucleotides present in a particular helix, which can be readily identified. The other issue, corresponding to the comparison between character states, was resolved using only the correction from the parsimonious-informative sites 24. Nevertheless, because there were inconsistencies in the placement of a few species (E. mammosa and E. fusca), this study needs to be extended, hopefully including all Eunicea representatives, in order to gain a better understanding of the systematics of this group as well as the evolution of their predicted ITS2 RNA secondary structures.

The molecular study of this genus, which is the most species-rich among shallow-water Caribbean gorgonian corals, is just beginning. Consequently, ITS2 RNA secondary structure analysis could be a valuable tool for distinguishing new species and completing Eunicea systematics: morphological phylogenetic reconstruction of octocorals is very difficult (e.g. 13), and ITS2 secondary structure contains more information than the usual primary sequence alignment 526.

AG performed most of the phylogenetic analyses and had a major role in writing the manuscript. CA reconstructed the ITS2 RNA secondary structure, collaborated in the phylogenetic analyses and had a major role in writing the manuscript. JAS collected all the specimens studied, obtained the DNA sequences in the laboratory and performed the SEM analyses of sclerites; he also contributed to write the manuscript. All authors read and approved the final manuscript.