Systematics and Molecular Phylogeny of the Family Oscarellidae (Homoscleromorpha) with Description of Two New Oscarella Species

The family Oscarellidae is one of the two families in the class Homoscleromorpha (phylum Porifera) and is characterized by the absence of a skeleton and the presence of a specific mitochondrial gene, tatC. This family currently encompasses sponges in two genera: Oscarella with 17 described species and Pseudocorticium with one described species. Although sponges in this group are relatively well-studied, phylogenetic relationships among members of Oscarellidae and the validity of genus Pseudocorticium remain open questions. Here we present a phylogenetic analysis of Oscarellidae using four markers (18S rDNA, 28S rDNA, atp6, tatC), and argue that it should become a mono-generic family, with Pseudocorticium being synonymized with Oscarella, and with the transfer of Pseudocorticium jarrei to Oscarella jarrei. We show that the genus Oscarella can be subdivided into four clades, each of which is supported by either a small number of morphological characters or by molecular synapomorphies. In addition, we describe two new species of Oscarella from Norwegian fjords: O. bergenensis sp. nov. and O. nicolae sp. nov., and we compare their morphology, anatomy, and cytology with other species in this genus. Internal anatomical characters are similar in both species, but details of external morphology and particularly of cytological characters provide diagnostic features. Our study also confirms that O. lobularis and O. tuberculata are two distinct polychromic sibling species. This study highlights the difficulties of species identification in skeleton-less sponges and, more generally, in groups where morphological characters are scarce. Adopting a multi-marker approach is thus highly suitable for these groups.


Introduction
Sponges (phylum Porifera) are now formally regarded as being composed of four lineages: Demospongiae, Calcarea, Hexactinellida and Homoscleromorpha [1,2,3,4,5]. The latter group, Homoscleromorpha, contains about 78 species and is subdivided into two families: Plakinidae and Oscarellidae [1,6]. Currently, family Plakinidae encompasses five spiculate genera (Plakina, Plakortis, Plakinastrella, Placinolopha and Corticium) while Oscarellidae includes two aspiculate genera (Oscarella and Pseudocorticium) [1,6]. Family Oscarellidae [6] was established by Lendenfeld in 1887, but was rejected in 1995 [7] following the description of Pseudocorticium, which is morphologically similar to Corticium though lacking spicules. Oscarellidae was restored as a family only recently as the result of studies of molecular phylogeny [6] and metabolomic fingerprints [8] of Homoscleromorpha. The results of these studies have demonstrated that morphological similarities found in spiculate Corticium and aspiculate Pseudocorticium (cortex, aquiferous system organization, and outer morphology) are either plesiomorphic or homoplasic characters [6] and that Pseudocorticium is, in fact, more closely related to Oscarella species with very different internal and external morphology.
Although the subdivision of Homoscleromorpha into two families is now formally accepted by the sponge scientific community [1,6,9], the relationships between the two oscarellid genera and consequently, the monophyly of Oscarella, remain contentious. Indeed, in our previous study [6], the analyses of complete mtDNA genomes and 18S rDNA data supported the paraphyly of the Oscarella, which encompassed Pseudocorticium jarrei Boury-Esnault et al., 1995. In contrast, the 28S rDNA sequences supported the monophyly of Oscarella with P. jarrei as its sister group. In order to resolve this issue, we conducted a further molecular study including additional Oscarella species.
Oscarella lobularis (Schmidt, 1862) [10], the type species of the genus, was long considered to be a single abundant cosmopolitan species displaying a high polymorphism of both consistency (soft and cartilaginous) and color (purple, blue, yellow and green) [11]. In 1992, Boury-Esnault and colleagues investigated the relation-ships among four color morphotypes of O. lobularis from the Marseille area by analyzing their allozymes and cytological features [12]. They showed that two species were present rather than one: O. lobularis (the soft purple/ivory specimens, [13]) and O. tuberculata (Schmidt, 1868) (the yellow, green or blue cartilaginous specimens). Since then, however, the picture has become more complex and is rife with ambiguities. Recent studies have found soft specimens of Oscarella which do not have the habitual purple/ ivory coloring, but which are blue, entirely purple or pink. Similarly, cartilaginous specimens of Oscarella may also be purple or pink in addition to the green, blue and yellow morphotypes [13,14]. Subsequent, finer histological studies revealed additional differences among various color morphs of Oscarella, and the 'cosmopolitan' O. lobularis turns out to be different species (10 new species of Oscarella have been described during the last 20 years) [15,16,17,18]. The absence of a skeleton (the main morphological character for sponge taxonomy) and thus, the paucity of available morphological characters for Oscarella systematics, is largely responsible for difficulties associated with species delimitation in this genus, as well as in other genera of sponges without skeletons (e.g. Halisarca [19,20,21]). At present, Oscarella comprises 17 species, listed in the World Porifera Data Base (http://www. marinespecies.org/porifera/index.php), including seven Mediterranean species. However, this is certainly an underestimate and several new species are currently under description (this study) or have yet to be described. The relationships among these 17 Oscarella species and the phylogenetic position of Pseudocorticium jarrei relative to them are also largely unknown [6]. In addition, the relationships between the different color morphs of the two putative sibling species O. lobularis and O. tuberculata [14] have not yet been fully resolved and more loci from more color morphs are needed to elucidate them. This is crucially important, especially because O. lobularis is being developed as a new model species for evo-devo studies [13,22,23,24,25].
Thus the aim of this paper is to investigate the principal uncertainties in the phylogeny of Oscarellidae described above: (i) the position of Pseudocorticium jarrei and the monophyly of genus Oscarella, (ii) the relationships among common Oscarella species, (iii) the relationships between different color morphs of Oscarella lobularis and O. tuberculata. For this purpose, we collected a diverse dataset relating to 22 Oscarellidae specimens from different geographical areas. Our dataset includes four molecular markers (two nuclear (18S rDNA, 28S rDNA) and two mitochondrial (atp6, tatC)), predicted secondary structures features for nuclear rDNAs [26] and multiple non-molecular characters (in particular, histological and cytological). We also incorporated two new species of Oscarella from Bergen Fjords in our analysis, and we provide their morphological descriptions and formal diagnoses. We discuss our results from an integrative taxonomic point of view [27].

Specimen Collection
Specimens of Oscarellidae from the Mediterranean Sea, the Norwegian Fjords, the East Atlantic and the North Pacific were collected using SCUBA diving by members of our team (AVE, EG) or were provided by colleagues (see Acknowledgments). Locations of the collection sites are shown on Figure 1. The samples used in this study and their current taxonomic status are summarized in Table 1. The different color morphs of Oscarella lobularis and O. tuberculata are presented in situ in Figure S1. We obtained molecular data from 9 of the 17 officially described species of Oscarella. In addition, we formally describe two new species in this paper and discuss two specimens with uncertain systematic positions (Oscarella sp. (pink) and Oscarella sp. (purple)). Despite substantial efforts, we did not succeed in obtaining DNA from museum specimens of O. nigraviolacea Bergquist [28,29]. We choose not to include in our sampling O. imperialis Muricy et al., 1996 [30], as we were unable to find this species with certainty in situ.

Morphological Studies
2.1. Taxonomy. The identification of all specimens has been carefully checked on the basis of morphological characters by the taxonomists in our team (AVE and JV).
2.2. Description of new species. Specimens from Norway were collected using SCUBA diving on June 23 2009 from vertical walls of granite rocks (North Sea, Norway, Skarvoysundet +60u 27' 34.74" N, +4u 56' 2.16" E) at depths of 3 to 9 m. Vouchers for transmission electron microscopy (TEM) and scanning electron microscopy (SEM) were fixed according to [31]. Sections were cut with a diamond knife on a Leica Ultracut UCT Ultramicrotome. Semi-thin sections were stained with toluidine blue, observed using light microscopy (LM) and photographed with a Leica DMLB digital camera. For SEM, the specimens were fractured in liquid nitrogen, critical-point-dried, sputter-coated with gold-palladium, and observed under a Hitachi S570 SEM. Type specimens have been deposited in the Muséum National d'Histoire Naturelle (MNHN, Paris, France) and the new species have been registered in ZooBank.
2.3. Nomenclatural Acts. The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature, and hence the new names contained herein are available under that Code from the electronic edition of this article. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix "http://zoobank.org/". The LSID for this publication is: urn:lsid:zoobank.org:pub: 9766D4A6-D2B5-4311-BFDB-8B3262B401F1. The electronic edition of this work was published in a journal with an ISSN, and has been archived and is available from the following digital repositories: PubMed Central and LOCKSS.
2.4. Non-molecular characters. To identify putative nonmolecular synapomorphies for clades revealed by molecular analyses, several characters belonging to five major categories -(i) ecology/geography, (ii) external morphology, (iii) histology/ cytology, (iv) microbiology and (v) embryology -are described for each species. Observed characters and their states are listed in Tables 2 and 3. Most of these characters are non-informative for phylogenetic reconstruction and thus this table of characters has not been used as a matrix for phylogenetic analyses. Instead, parsimony reconstruction of character evolution (the matrix of characters is provided in Text S1) based on the consensus molecular tree ( Figure S2) was performed using Mesquite software version 2.72 [32].

Molecular Methods
3.1. Rationale for the choice of molecular markers. Our decision to use 18S rDNA and 28S rDNA markers in our analysis was based on their prior efficacy in solving phylogenetic relationships at the genus and supra-generic levels for various sponge groups [2,6,33,34,35,36,37]. In addition, we developed two new mitochondrial markers for Homoscleromorpha based on our previous data [6] and on new complete mtDNA sequences from Oscarella tuberculata yellow and Oscarella sp. (purple). The regions chosen for these markers are located within tatC and atp6 and contain the largest number of parsimony informative sites per kb of sequence in a whole mt-genome alignment for five closelyrelated Oscarella species: O. tuberculata green, O. tuberculata yellow, Oscarella sp. (purple), O. lobularis, and O. viridis. Together, these two markers encompassed ,30% of such sites (10/35) in fewer than 8% of mt-genome sequences.
Interestingly, atp6 has recently been used in another study where it has been shown to be suitable for alpha-level systematics in sponges [38]. To circumvent the pitfalls of using single-gene trees, which can tell a biased story of the species relationships [38,39], we combined these four genes in our analysis.
3.2. DNA sequence acquisition. Procedures used for genomic DNA extraction, cloning and DNA sequencing are standard laboratory protocols described in previous publications [6,37]. PCR primers for full-length/partial 18S rDNA, partial 28S rDNA, atp6 and tatC amplification are provided in Table S1. It was necessary to adapt reaction conditions for each species from previous studies [6,37]; the exact conditions of amplification can be provided by the authors upon request. The poriferan origin of the sequences was checked by a BLAST search [40] against the NCBI GenBank collection (http://www.ncbi.nlm.nih.gov/). All new sequences/genomes were deposited in GenBank under accession numbers listed in Table 1. 3.3. Sequence alignment. To achieve a reasonable trade-off between representativeness of outgroup taxa and ease of alignment, and because our prime interests were relationships within the Oscarellidae, we restricted our sampling to Homoscleromorpha. Several species of Plakinidae (three Corticium and two Plakortis species), the sister group of Oscarellidae, were used as an outgroup. Initial sequence alignment was performed using the software MUSCLE online (http://www.ebi.ac.uk/Tools/muscle/ index.html) [41,42], and was subsequently optimized by eye using the Bioedit Sequence Alignment Editor v5.09 [43]. Ambiguously aligned regions were determined by Gblocks v0.91 b software [44] for nuclear markers only (mitochondrial ones were partitioned by codon position). A relaxed selection of blocks is better for short alignment [45], thus the settings were the following for the 18S rDNA [1:13; 2:13; 3:8 4:2; 5: all] and the 28S rDNA [1:13; 2:13; 3:8; 4:2; 5: all]. The treatment by Gblocks resulted in the removal of 1% and 4%, for the 18S rDNA and 28S rDNA alignments, respectively. The character exclusion sets based on Gblocks are available upon request from the corresponding author.
3.4. Phylogenetic analyses. Phylogenetic analyses were performed using maximum likelihood (ML) and Bayesian inference (BI) methods.
For ML analyses, we used the Akaike Information Criterion (AIC) in JModelTest [46] to determine the best fitting nucleotide substitution model for each data set. The following models were chosen for 18S rDNA, 28S rDNA, 18S rDNA +28S rDNA, atp6, tatC, tatC+atp6, 18S rDNA +28S rDNA +tatC+atp6 datasets respectively: TIM2+G; GTR+G; TIM2+G; TPM2uf+I+G; TIM1+I; GTR+G; GTR+G. ML phylogenetic analyses were performed with PhyML software v.3 [47,48] using the previously estimated parameters. Among sites, rate heterogeneity was estimated using a discrete approximation of the gamma distribution with six rate categories. Gaps were treated as missing data and the statistical robustness of the tree topology was assessed by nonparametric bootstrap resampling (1000 replicates) [49]. Bootstrap values .80 were considered high enough to support clades in ML reconstructions.
BI analyses were performed with MrBayes 3.2.1 [50] under the best-fit evolutionary model estimated for each independent gene or partition under the AIC criterion with MrModeltest 2.3 [51]. The models selected for 18S rDNA and 28S rDNA were GTR+I+G and GTR+G, respectively. For mitochondrial markers atp6 and tatC, we partitioned the dataset according to codon position and the models selected are as follows;   Four Markov Chains were run for 2 million generations and sampled every 100 generations. The chains converged significantly and the average standard deviation of split frequencies was ,0.01 at the end of the run. The trees of the early generations (5000 trees) were discarded until the probabilities reached a stable plateau (burn-in) and the remaining trees were used to generate a 50% majority-rule consensus tree. Only posterior probabilities .0.90 were considered to robustly support clades.
All trees except the tatC and tatC+atp6 were rooted on Plakinidae species, the exceptions arising as the tatC gene is specific to Oscarellidae [6]. Based on the nuclear and atp6 phylogenetic analyses (this study) and previous complete mitochondrial analyses [6], these trees were instead rooted on a robust internal clade. The trees were visualized and edited using FigTree v.1.3.1 [52].   [54,55]. The default settings for all parameters were used as in a previous study [37]. When multiple secondary structures were found, we chose the structure with the lowest free energy (DG in kcal/mol) [26].We decomposed the rDNA secondary structures into elements (framed by different colors) and defined binary characters (i.e. presence/absence of each element). A matrix (Text S2) was constructed and parsimony reconstructions of character evolution were performed using Mesquite software version 2.72 [32] on the 18S rDNA molecular tree. 3.6. Sequences identity and nucleotide diversities. We investigated the percentage of molecular divergence (Table 4) on the mitochondrial marker sequences (atp6 and tatC) by using the identity matrix option of the Bioedit software [43]. Nucleotide diversities (p) between Oscarella tuberculata, O. lobularis, D3 sequences, as well as congeneric species of Oscarella, were estimated using DnaSP 5.10 software [56] for the atp6 marker and compared to available demosponge data [38] (Table 5).
3.7. Consensus tree and diagnostic characters. A consensus tree based on tatC+atp6 molecular phylogenies was manually drawn to map the diverse molecular and non-molecular characters that are diagnostic of some Oscarella clades. All robust nodes (BP.50+ PP.0.5) were conserved. Polytomy was prioritized for weakly supported nodes (BP,50 or PP,0.5).  Table 2. Three sets of non-molecular characters-states for each species/specimen: ecology/geography, external morphology and associated microbes. In summary, analyses of nuclear markers allow to clearly identify two main clades A and B, the latter being itself composed of two monophyletic groups: C and D. The relationships among the D clade (containing most of our samples) are not always clear and congruent, thus calling for the use of other markers (or more taxa).
The position of Oscarella viridis was resolved as either the sister group of C (in tatC tree, Figure 3A) or of D (atp6 tree, Figure 3B) with moderate support in both cases (65/0.72, 77/0.84, 2/0.65). In all cases, the positions observed here are not congruent with those found using nuclear markers.
Furthermore, in contrast to the results obtained with the nuclear genes, both mt markers have partly resolved the relationships between the members of the D clade. The latter can be subdivided into three sub-clades (variably supported): D1, D2 and D3. D1, containing all color morphs of O. lobularis, is moderately supported by the analyses of tatC and the combined dataset, but is not supported by atp6 (2/0.71, not found, 84 Although the position of the third species from Bergen, O. bergenensis sp. nov., was not precisely determined, we noted that with both nuclear and mt markers, it is unrelated to Oscarella sp. (pink). The mt combined dataset was more powerful than both nuclear markers and separated mt markers in resolving the relationships among D ( Figure 4B). rDNA +28S rDNA topologies. As in the case for tatC marker, the sub-clade D3 is found, its robustness being weakly supported in ML and well supported in BI analyses (56/1).
1.4. Predicted secondary structures of 18S rDNA. The predicted secondary structures for the V4 region are presented in Figure 5. We have encoded the molecular morphology of the derived secondary structures as elements (loop, semi-loop and helix) and consider them as binary characters. For this purpose, we use a color code to map them and their appearance. The Oscarellidae species included in this study present different secondary structures, but all except one share a specific element of an internal loop and a terminal loop (in orange, absent in Plakinidae). The presence of this specific structure appears to be diagnostic of Oscarellidae. The absence of this element in Oscarella lobularis (purple/ivory) is a derived feature of this morph. Among  internal loop lined by three base pairs on each side (in red), which appears to be a synapomorphy of this clade. Three species from A+O. microlobata share a common feature (two small terminal loops, in light blue) for which no evolutionary history can be proposed.

Morphological Descriptions and Systematic Justification for Two New Species
According to molecular data, it appears that at least two Oscarella samples from Bergen can be considered as distinct species (see discussion), and are thus morphologically and formally described herein.
Phylum DIAGNOSIS: Red-orange Oscarella at the apical parts and patchy yellow at the basal parts, with folded surface and average resilience consistency; containing two particular cell types with inclusions (vacuolar and granular cells), archaeocytes in low number in the mesohyl and one morphotype of endobiontic bacteria.
DESCRIPTION: Moderately large, encrusting, size from 2x1 cm to 6x4 cm, thickness 4-8 mm. Easy to detach from its substrate. Smooth surface but with small folds. Oscula at the end of small conical lobes 2-3 mm in height, not transparent. Consistency: average resilience. Color in vivo red-orange at the apical parts of sponge and patches of yellow and orange in inner parts, not bright ( Figure 6A).
SOFT TISSUE ORGANISATION: Spicule and fiber skeleton absent. Ectosome from 9 to 20 mm thick ( Figure 6B). Inhalant canals (12 mm in diameter) running perpendicular to the surface ( Figure 6B). Choanocyte chambers are eurypylous, roughly spherical to ovoid, about 32 mm in diameter ( Figure 6B). Choanocyte chambers are arranged around inhalant and exhalant canals in a sylleibid pattern. Exhalant canals about 40 mm in diameter running toward well-developed system of basal cavities, leading to the oscula. Ostia regularly distributed, 16-21 mm in diameter.
CYTOLOGY: Exopinacocytes ( Figure 6C) are flat (7.1 mm wide by 1.9 mm high), flagellated. Nucleus is ovoid (1.6 mm in diameter), often with a visible nucleolus. Cytoplasm contains inclusions, phagosomes (from 0.2 to 0.8 mm in diameter) and vacuoles (0.1-0.9 mm in diameter). Endopinacocytes ( Figure 6D) are flat (7.9 mm wide by 2.4 mm high), flagellated, often with thin cytoplasm projections in their basal part. Nucleus is ovoid (2.2 mm). Cytoplasm contains numerous osmiophilic inclusions as well as phagosomes (from 0.3 to 0.9 mm in diameter) and vacuoles (0.1-1.1 mm in diameter). Apopylar cells ( Figure 6E Figure 6H) are ovoid to irregular, 6.4 to 8.8 mm long with ovoid or slightly irregular nucleus 1.9 mm in diameter without nucleolus. Their cytoplasm has one to four large, irregular vacuoles (about 1.161.6 mm to 4.165.2 mm) with clear, filamentous contents. Vacuoles are often brought together. Granular cells ( Figure 6I) are ovoid to irregular (7.8 mm long and about 4.4 mm width). The nucleus is 2.2 mm in diameter. Cytoplasm is filled with 10 to 16 electron-transparent spherical vacuoles about 0.8 mm in diameter with loosely dispersed material. Rare electron-dense homogenous granules (0.4-0.9 mm in diameter) are present in the cytoplasm. Symbiotic bacteria belonging to a single morphotype are dispersed extracellularly in the mesohyl ( Figure 6J). They are elongated, rod-like (length 1.1-1.8 mm and diameter 0.25-0.35 mm). The cell wall is Gram-negative and consists of two layers. A filamentous network of the nucleoid is irregular with thick elements in the center and thin filaments closer to the periphery of the cell. A small layer of granular cytoplasm is observed near the cytoplasm membrane. Short, radially-arranged filaments are present outside the cytoplasmic membrane.
REPRODUCTION: Only rare early oocytes (before vitellogenesis) were observed in the material (collected from mid-to late June).
HABITAT: Depth 3-10 m; abundant as epiphyte on the basal parts of thalli of Laminaria digitata, on granite rock, vertical walls.
ETYMOLOGY: The species name is derived from the site where it was discovered (Bergen fjords).
TAXONOMIC REMARKS: Regarding color, Oscarella bergenensis sp. nov. is unusual: red-orange at the apical parts with patches of yellow at the inner parts. This coloring is almost unique compared to all described Oscarella species: only some specimens of O. rubra [58] from Roscoff have the same color pattern (AE, personal observation). The surface of O. bergenensis sp. nov. is smooth with small folds, and is thus similar to that of O. rubra. Cells with inclusions of the mesohyl are important discriminating characters in Oscarella taxonomy. These cells are diverse and abundant in Oscarella species, each species being characterized by a distinctive set of cells with highly variable inclusion types [12,15,16,17,29,59]   basal parts. The cells have short pseudopodia. Choanoderm and pinacoderm are lined with a basement membrane, which is a continuous, 12-19 nm thick layer of condensed collagen microfibrils ( Figure 7D, E, F). A thin, irregular layer of glycocalyx covers the surface of exopinacocytes, endopinacocytes, choanocytes and apopylar cells. Archaeocytes ( Figure 7H) are amoeboid (7.7 mm wide by 5.1 mm length). A large nucleolated nucleus is spherical or ovoid (3.862.9 mm). Cytoplasm includes small phagosomes and rare electron-transparent vacuoles (from 0.6 to 1.6 mm in diameter). Two types of cells with inclusions occur within the mesohyl: (i) Granular cells ( Figure 7I) Figure 7C). The cells are present during embryogenesis inside the embryos ( Figure 7C). Symbiotic bacteria: One morphological type of endosymbiont, extracellular bacteria occurs in the mesohyl ( Figure 7D, H, K). This type of bacteria is elongated to oval (0.8-1.5 mm long and 0.37-0.8 mm in diameter). Under the cell wall a layer of dense filaments can be observed. The nucleoid zone consists of an almost regular filamentous network. Some bacteria have internal mesosome-like structures. A developed glycocalyx is present at the surface of bacteria.
REPRODUCTION: Oscarella nicolae sp. nov. is ovoviviparous and simultaneously hermaphrodite: in the same individuals male and female reproductive elements are present. The spermatic cysts have different diameters (from 20 to 95 mm) and are randomly distributed in the sponge mesohyl ( Figure 7C). Spermatogenesis is generally asynchronous inside spermatic cysts. Oogenesis and embryogenesis are asynchronous, all stages from oogonia to egg were observed within the same specimen. Mature eggs are isolecithal and polylecithal, with a cytoplasm full of yolk granules ( Figure 7C). Embryogenesis is also asynchronous. All stages from cleaving embryos to prelarva were observed from mid-June to mid-July ( Figure 7C

Evolutionary Histories of Non-molecular Characters
The ecological, morphological, cytological, and embryological characters of Oscarellidae species examined in this study are presented in Tables 2 and 3.
Some of the cytological characters mentioned in these tables have often been mislabeled in previous publications and need to be clarified. According to the Thesaurus of Sponge Morphology [60] spherulous cells are cells filled with large round spherules that occupy almost the entire cytoplasm. Unfortunately, Oscarella's cytology descriptions in publications are sometime ambiguous. Indeed, the cell type designations used are confusing especially concerning granular, spherulous and globular cell terms. For our analysis, we did a complete revision of previously published descriptions and photos of Oscarella spp., as well as a comparative analysis of TEM pictures from our collection. The results of this analysis are provided in Table 3 [7,16,17,30]. In all these species, spherulous cells have characteristic paracrystalline inclusions which have never been described in other homoscleromorphs and which are very uncommon in other sponges. Furthermore, one additional type of spherulous cell "with granular inclusions" and without paracrystalline elements was described in O. microlobata [30].
Evolutionary histories of characters were examined and some were found to be not informative, as they were constant in all sampled specimens ( Figure S2A). This is the case for the basement membrane and several embryological characters that are present in all Homoscleromorpha. Other characters are highly variable in each species, and thus contain little phylogenetic signal. None of the recorded ecological, geographical and external morphological characters reflects phylogenetic relationships of Oscarella species. Furthermore, no color, type of consistency or sort of surface can be linked to any molecular clade. Nine histological and cytological characters were compared. It is noteworthy that one cytological character, the presence of spherulous cells with paracrystalline inclusions, is found in clade A species and also in O. microlobata which has an unclear position in phylogenetic analyses ( Figure  S2B). Three characters, concerning the aquiferous system and the cortex, are constant for all but one species: Pseudocorticium jarrei, now Oscarella jarrei, which has character-states similar to the Plakinidae (outgroup) ( Figure S2C). The sylleibid aquiferous system and the eurypylous choanocyte chambers appear diagnostic for Oscarellidae, however, these characters have also been described in some Plakina species (Plakinidae): P. trilopha, P. monolopha, P. crypta, P. endoumensis, P. jani [61]. In addition, we also looked at four other cellular types (archaeocytes, vacuolar cells, granular cells and spherulous cells). For the most part, their absence or presence in different states cannot be related to any relationship supported by molecular data ( Figure S2D). Nevertheless, we noticed that the vacuolar cells are absent in all clade A species with the exception of O. balibaloi. Our analysis suggests that the ancestral state for A members is the ''absence'' of vacuolar cells and that the state of O. balibaloi is due to reversal. In conclusion, we found very few morphological and cytological diagnostic characters supporting the clades defined based on molecular data. Clade A appears to be characterized by the absence of vacuolar cells and the presence of spherulous cells with paracrystalline inclusions, but with the exceptions of O. balibaloi (the member of the clade having vacuolar cells) and of O. microlobata (outside clade A -although with a poorly defined position -and having spherulous cells with paracrystals).

Suitability and Limits of Molecular Markers Used
All four markers used for this study were informative for understanding phylogenetic relationships within Oscarellidae and produced mostly congruent trees (Figures 2, 3 and 4). However, some differences were observed in the performance of individual markers and are discussed below.
The 18S rDNA analyses resolved the deeper nodes of the phylogenies. The suitability of this marker for reconstructing relationships at family level has often been demonstrated in sponge phylogenies [27]. Nevertheless, as previously noticed, its power of resolution is insufficient for deciphering relationships between closely-related species: this is the case here -and was expected given our previous data -for Oscarella tuberculata and O. lobularis color morphs, and more generally within the clade D. The 28S rDNA mostly confirms the main topology obtained with the 18S rDNA but did not provide higher resolution for the D clade. Moreover, the inferred relationships within D were different between the two nuclear markers. These observed discrepancies could be explained either by scarcity of the phylogenetic signal or by the fact that these markers constitute multigene families. Although it is generally assumed that paralogous copies of rDNA genes are homogeneous [62,63,64] because they evolve by concerted evolution, this is not always the case, and conflicting phylogenies have been inferred when using different copies of ITS [65,66] or 18S rDNA [67,68]. Nevertheless, one can suppose that the action of concerted evolution may not be sufficient to compensate paralogous evolution for recently diverged species.
Mitochondrial markers provide several advantages for phylogenetic reconstructions including their higher rate of sequence evolution and the rarity of gene duplication. Assuming the uniparental inheritance (common for most metazoans) and the effective absence of recombination (characteristic for Metazoa), the whole mtDNA should evolve as a single locus.
Thus, we selected two regions that showed the highest diversity in Oscarellidae in our preliminary results. Not surprisingly, these two regions did not include cox1, which is one of the most conserved regions of the mitochondrial genome and often fails to resolve relationships among closely-related sponge species [69,70] and, in general, often performs worse in resolving animal relationships than other genes. The atp6 marker has already been selected for alpha-level systematics in Demospongiae because it was more polymorphic than cox1 in the studied species [38]. In our case, despite comparable nucleotide diversity (P, Table 5), atp6 appears not to be powerful enough to clearly resolve phylogenetic relationships among members of D clade. TatC is the gene that encodes the subunit C of twin-arginine translocase. The Twin-Arginine Translocation Pathway is involved in transfer of folded proteins across biological membranes in bacteria, chloroplast and, possibly, mitochondria [71]. TatC gene is not found in any other sponge or animal mtDNA [72,73] but is commonly present in mitochondrial genomes of other eukaryotes [74]. It was first reported in the O. carmela genome [73] and then found in other Oscarellidae species ( [6], this study). This gene appears to be more variable than atp6 and is the most variable of our four markers. The tatC marker thus helped to better resolve the D clade relationships and led us to propose the D1, D2 and D3 sub-clades. We were surprised that despite their geographical distance and cyto-morphological differences, the three samples of sub-clade D3 have identical sequences for both tatC and atp6 genes. This observation suggests either the presence of three different morphotypes of only one species (see section 4.4) or the fact that both mt markers are non-resolving at this scale. The limit of resolution issue for mt markers was already known for non bilaterian species [75]. A study at a finer scale using more variable markers (e.g. introns) will be necessary in the future to understand the evolutionary history of D3 individuals.

Monophyly of Oscarella and Synonymy with Pseudocorticium
Our results confirm that the Oscarellidae, i.e. the Homoscleromorpha without skeletons, are separated into two clades A and B. Pseudocorticium jarrei is clearly included in clade A, and this is congruent with chemical data obtained by metabolic fingerprints [8]. The species also shares with all but one Oscarella spp. a character of the secondary structure of the 18S rDNA V4 region (internal and terminal loop), which can be considered as diagnostic for this group. Pseudocorticium jarrei, however, differs from Oscarella spp. by the development of the mesohyl, the aquiferous system and the thickness of the cortex. Two taxonomic interpretations are possible: (i) to maintain the genus Pseudocorticium for clade A, with transfer of Oscarella balibaloi, O. nicolae sp. nov. and O. kamchatkensis to Pseudocorticium, (ii) to synonymize Pseudocorticium with Oscarella, with transfer of P. jarrei to Oscarella. We prefer the second alternative involving fewer changes, considering that the outer morphology, aquiferous system and development of the mesohyl of P. jarrei can be considered as homoplasic features [6]. We thus propose the abandonment of the genus Pseudocorticium, which is presently made up of only one species. Pseudocorticium jarrei is thus transferred to Oscarella. Consequently, the Oscarellidae becomes a monogeneric (Oscarella) family, with the following definition: Homoscleromorpha without skeletons, with a variable degree of ectosome development. The aquiferous system has a sylleibid-like or leuconoid organization, with eurypylous or diplodal choanocyte chambers. Their mitochondrial genomes encode a gene absent in other animal mitochondrial genomes: tatC [6].

The Two Sibling Species Oscarella Lobularis and O. tuberculata
Our molecular results, including several color morphs for both Oscarella lobularis and O. tuberculata, mainly confirm and extend the previous study based on allozymes [12]. Indeed, phylogenetic analyses, molecular divergence and nucleotide diversity calcula-tions suggest that O. lobularis and O. tuberculata are two different and distinct species that live in sympatry. This is also corroborated by the presence of diagnostic positions in both mt DNA markers for each species ( Figure S3). Cell composition [12], symbiotic microbe diversity [21], as well as life history traits such as sex ratios and reproductive cycles, also differ between O. lobularis and O. tuberculata, consistent with the presence of two distinct species [14]. Concerning color polymorphism, the type species O. lobularis displays not only the classical purple/ivory color arrangement but can also be blue, pink and purple. O. tuberculata specimens also present a variety of colors from yellow to purple. The high polychromism observed in these two species is quite unusual, compared to other sponges, although color variability is often observed between specimens of sponges living in different local environmental conditions (shady side vs exposed to light) [76] and non-ecophenotypical polychromism in sponges can also be found (e.g. Mycale species [77]). All color morphs of both O. lobularis and O. tuberculata from our study live in the same areas and environments. A wide comparison within sponges is hampered by the fact that very few studies combining molecular and morphological data on color morphs of sponges have been conducted so far.
In several genera, it has been shown that color morphs are, in fact, distinct species: e.g. Suberites ficus from Great Britain using allozymes [78], or Latrunculia from New Zealand, using allozymes and chemical studies [79]. An almost comparable situation to what we found in Oscarella species is the case of Callyspongia vaginalis. This species has three morphotypes that vary in both color and shape, but which have identical sequences for four molecular markers, thus revealing that they all belong to the same species [80] (or that the markers do not evolve fast enough).
We cannot unequivocally exclude that each color morph (for both O. tuberculata and O. lobularis) may represent populations in the course of speciation; each might become a new species in the future. This is particularly the case of the O. tuberculata yellow morph, which is the most divergent in this species. In addition, these color morphs may also already be different recent species that were not discriminated by the markers used. A population genetics study using more variable markers with more specimens of various localities is needed to conclusively solve this question (assuming that ''true'' species exist in nature: see [81] for some insights). Nevertheless, the absence of distinguishable morphological characters between these color morphs, and particularly the fact that they are sympatric has led us now to prioritize, the polychromism hypothesis, and hence the presence of only two species.
As the homoscleromorph sponge O. lobularis is now emerging as a sponge model for evo-devo studies [13,22,23,24], clear criteria for sampling are indispensable. As color is not a valid diagnostic character to distinguish this sponge from O. tuberculata, only the in situ evaluation of the specimen consistency (soft vs cartilaginous) is useful during collection. This could be complemented by sequencing atp6 and/or tatC and looking for diagnostic positions (see Figure S3) and for the presence of vacuolar cells in LM.

One or Several Species in sub-clade D3? An Open Question
Our results based on phylogenetic analyses, molecular divergence and nucleotide diversity calculations have revealed no divergence/diversity between Oscarella sp. (purple), O. rubra and O. sp. (pink) in mitochondrial sequences. However, conventional taxonomy based on morphological characters has revealed that these samples have very distinct features (such as outer morphology, cytology and microbial composition, see Tables 2 and 3).
These three specimens could represent a single, morphologically variable species, Oscarella rubra, or may represent three distinct species. This can be resolved only by a more complete study based on more numerous specimens.

At least Two New Oscarella Species from Bergen Fjords
In this study, three different species/specimens of sponge, living in sympatry and belonging to the Oscarella have been sampled in Bergen fjords. Molecular data, especially the mitochondrial data, have revealed that these specimens are not closely-related but belong to three distinct species. Consequently, two of them have been morphologically described and formally named (see results). Oscarella nicolae sp. nov. belongs to clade A and is more closelyrelated to O. jarrei, O. kamchatkensis and O. balibaloi than to the other species from Bergen fjords. Its external morphology and, in particular, its microlobate surface, distinguish this species from the two others. Cytological characters are also informative: archaeocytes are often present in its mesohyl, while no vacuolar cells have been found, contrary to the two other species. Moreover, O. nicolae sp. nov. contains spherulous cells with paracrystalline inclusions and has no vacuolar cells as is the case for the other members of clade A. The affinity of O. nicolae sp. nov. with O. balibaloi is also supported by a shared secondary structure element in their 18S rDNA. Oscarella bergenensis sp. nov. is included in clade D, with uncertain relationships regarding D1, D2 and D3 sub-clades. It is clearly distinct from the two other samples from Bergen, as shown by phylogenetic analyses. Furthermore, its external morphology (color and consistency) is different to the two other specimens from Bergen, while its histological and cytological features are quite similar to Oscarella sp. (pink). The last Bergen specimen, Oscarella. sp. (pink) belongs to D3 sub-clade and it is unclear whether it is O. rubra or a new species (see 4.4). Present data are insufficient, preventing firm conclusions from being drawn, as a result, we propose no morphological description for the moment.

Conclusions
The taxonomy of sponges is usually based on the characters of the skeleton, fibers and spicules. Due to the absence of skeleton and to the presence of mainly invariable histological characterstates, the identification of Oscarella at the species level is very difficult. The differences among species are mostly in external traits: color, consistency, and aspect of the surface [12,15,16,17,18,28,29,57], but these characters must be evaluated with care; they can be highly subjective and also very polymorphic. At the same time, some cytological characters, such as the presence of types of cells with inclusions and symbiotic microbes morphology, have been proposed to be more informative for some Oscarella species identification [15,16,17,18,21,29]. These findings are corroborated by the present molecular study because species previously described using morphology are confirmed here with molecular data. Nevertheless, except for two (absence of vacuolar cells and possibly spherulous cells with paracrystalline inclusions), these morphological character-states are not powerful tools for reconstructing species relationships in this group (Figure 8). Mapping morphological characters on molecular trees to identify synapomorphies which support clades has been a successful approach for diverse groups (e.g. Demospongiae, Calcarea [2,34,82,83]) although it has also been poorly indicative or unsuccessful for some demosponge taxa: Haplosclerida, Halichondrida, Axinellidae [37,84,85,86], as well as for Oscarellidae (Figure 8). This study offers additional evidence that when there are few available morphological features to study and compare, molecular biology offers a powerful tool to provide insights into phylogenetic relationships [55], but this also requires the discovery of an efficient marker for the question under investigation. A more in-depth exploration of the microbial diversity of the species (e.g. identification of associated species rather than bacterial morphotypes) may also be a successful alternative path to follow for Oscarellidae phylogeny and systematics [87].  Text S1 Mesquite matrix for some morphological characters from Tables 2 and 3. Characters and characters-states are detailed. (PDF)

Supporting Information
Text S2 Mesquite matrix for V4 secondary structures for 18S rDNA. (PDF) Figure 8. Simplified consensus tree based on tatC+atp6 molecular phylogenies. All robust nodes (BP.50+ PP.0.5) were conserved. Polytomy was prioritized for weakly-supported nodes (BP,50 or PP,0.5). Molecular and non-molecular characters that are synapomorphies/ diagnostic characters of the clades are indicated on the corresponding nodes. The absence of vacuolar cells (1) is diagnostic of A. The presence of cortex (2), leuconoid aquiferous system (3) and diplodal chambers (4) are specific characters of Oscarella jarrei. The presence of vacuolar cells in O. balibaloi is a reversal character (5) The two new species names are indicated in bold in the tree. In light grey are specimen/species for which uncertainties remain (new species or not). Schemas of morphological characters are modified from [60] or are new. doi:10.1371/journal.pone.0063976.g008