Introduction

The Mediterranean Sea, the largest semi-enclosed sea on Earth, is characterized by unique oceanographic conditions (e.g. limited water exchange, distinctive subsurface-water circulation and deep overturning circulation; see Pinardi et al. 2006; Tanhua et al. 2013) and strong environmental gradients (Danovaro et al. 1999). This wide range of climatic and hydrological conditions within the Mediterranean Sea have been used to define up to 10 biogeographical regions based on the combination of different geological, physical and biological parameters (Bianchi and Morri 2000) and results in the coexistence of temperate and sub-tropical organisms in this basin (Coll et al. 2010).

The Mediterranean is considered a hot spot of marine biodiversity with a high percentage of endemic species (Bianchi and Morri 2000; Boudouresque 2004). The Mediterranean marine biota is closely related to that of the Atlantic Ocean (Sarà 1985; Bianchi and Morri 2000), and its evolutionary history has been shaped by episodes of isolation and reconnection with the Atlantic. In this regard, evidence for a radiation after the Messinian Salinity Crisis (5.2 Ma)—the isolation and restriction of the Mediterranean due to the closure of marine gateways between the Atlantic ocean and the Mediterranean Sea—has been reported for invertebrates (Sanna et al. 2013) and vertebrates (e.g. Carreras-Carbonell et al. 2005). A second event, the Gelasian Crisis occurred 2.6 Ma (Dov Por 2009) likely affecting the Mediterranean biota and changing its general composition and diversity patterns. Although a wealth of oceanographic and ecological data exist for the Mediterranean Sea (e.g. Sarà 1985; Pinardi et al. 2006; Bas 2009), little is known about the evolutionary history of different Mediterranean taxonomic groups. Among octocorals in particular, the few studies so far published have been mostly focused on the population genetics of economically and/or ecologically relevant species in the region such as the precious coral Corallium rubrum (Costantini et al. 2013; Pratlong et al. 2015) and the red gorgonian, Paramuricea clavata (see Agell et al. 2009; Mokhtar-Jamaï et al. 2011).

The genus Paramuricea Koelliker 1865 (Holaxonia, Plexauridae) has been reported from different latitudes including the Gulf of Mexico (Doughty et al. 2014), North-Western Atlantic (Thoma et al. 2009), Mediterranean (Bo et al. 2012; Angeletti et al. 2014) and South Eastern Australia (Alderslade et al. 2014). The genus currently includes more than 20 species, two of which, P. clavata and P. macrospina, are endemic to the Mediterranean (Grasshoff 1992; Vafidis 2010). P. clavata is one of the most common gorgonians dwelling the coralligenous assemblages of the western Mediterranean. Paramuricea macrospina has a more restricted distribution occurring mainly in detritic or sandy deep-sea bottoms up to 200 m deep (Carpine and Grasshoff 1975) and is also common in the oligotrophic Sea of Marmara (Topçu and Öztürk 2015). Despite their ecological importance and their widespread distribution across the Mediterranean basin, the phylogenetic affinities of the Mediterranean Paramuricea and their historical biogeography remain largely unstudied. In order to shed light on the molecular (phylo) diversity of the Mediterranean Paramuricea and on the evolutionary processes leading to the diversification of these species, here, we (I) sequence the complete mitochondrial genome of the two Mediterranean endemic species, (II) assess their phylogenetic relationships with other members of this widespread genus and (III) infer the biogeographic history of Paramuricea to clarify the processes involved in the diversification of the Mediterranean species and shed light on the possible historical events that shaped the evolutionary history of this genus in the Mediterranean.

Materials and methods

Study area, sampling and taxonomic identification

P. macrospina (LT576168, LT576169) and P. clavata (LT576167) were collected in the Mediterranean, whereas Paramuricea grayi was sampled in Galicia (LT576170) and the Bay of Biscay (LT576171) (North-Eastern Atlantic). All the samples were collected by SCUBA diving at depths ranging from 10 to 90 m and were preserved in absolute ethanol for molecular analyses. Species assignments done in the field were further corroborated by the morphological analysis of the colony and sclerome.

Molecular analyses

Genomic DNA was extracted following three different protocols. The Macherey-Nagel NucleoSpin® Tissue kit (M&N, Düren, Germany), a modified CTAB phenol-chloroform (Doyle and Doyle 1987) and a salting out extraction method (Aljanabi and Martinez 1997) were used to obtain high molecular weight DNA. The quality of the DNA was assessed on a 1.5% agarose gel. Mitochondrial DNA fragments of different sizes (500–4000 bp) were amplified using published octocoral specific primers (i.e. France and Hoover 2002; Sánchez et al. 2003; McFadden and Hutchinson 2004) as well as additional primers designed to bind regions conserved in the octocoral mitochondrial genomes available to date (see Supplementary Table 1). For P. grayi (LT576170-LT576170) and P. macrospina (LT576169), the 5′ end of mtMutS was amplified using the primers ND42599F (France and Hoover 2002) and Mut-3458R (Sánchez et al. 2003) following previously published protocols. PCR products were purified using a polyethylene glycol (PEG)-NaCl precipitation and were sequenced with different internal primers (see Supplementary Table 1) using the BigDye® Terminator v3.1 chemistry in an ABI PRISM® 3700 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). The mitochondrial genomes were assembled, edited and annotated using Geneious 6.0.5 (Kearse et al. 2012). This program was also used to estimate GC-content and to annotate and create a graphical map of the genomes. Nucleotide diversity (π) between genomes was calculated in DnaSP 5.10.1 using a sliding window of 500 bp (Librado and Rozas 2009). The sequences obtained in this study were deposited in the European Nucleotide Archive (LT576167-LT576171).

Extended phylogenetic analyses

A fragment of about 750 bp of the octocoral mitochondrial gene mtMutS was extracted from the mitochondrial genome of the two Mediterranean species and was aligned with the sequences of P. grayi here generated and with sequences of other plexaurids deposited in public sequence repositories (GenBank). A dataset consisting of 166 taxa of which 150 belong to the genus Paramuricea was used to assess the phylogeny of the genus Paramuricea. This dataset contained many identical genotypes. Thus a second, reduced dataset (see Supplementary Table 2 for the list of species used) was generated including only 26 Paramuricea representative species. This dataset was used for phylogenetic inference and biogeographic analyses (i.e. S-DIVA, DEC; see below). All sequences were aligned in MUSCLE (Edgar 2004) with the default options available through Seaview 4.5.3 (Gouy et al. 2010). Alignments used in this study are available at OpenDataLMU (DOI: http://dx.doi.org/10.5282/ubm/data.89).

A maximum likelihood tree was inferred in RAxML 7.2.8 (Stamatakis 2006) under the GTR + Γ model (General Time Reversible + Gamma). Branch support was assessed through a rapid bootstrap analysis (Stamatakis et al. 2008) with a random seed number and 1000 pseudo-replicates. For the Bayesian analyses we used BEAST 2.4.4 (Bouckaert et al. 2014) under the best-fit substitution model selected using the AIC (i.e. GTR + Γ + I) in the program jModeltest 2.1.3 (Darriba et al. 2012). The Markov chain Monte Carlo was run for 10,000,000 generations sampling every 1000 generations. Convergence of the Markov chains was assessed in Tracer 1.6 (Rambaut and Drummond 2009). One tenth of the sampled trees (i.e. 1000 trees) were discarded as burn-in and the species Eunicea fusca, Muricea elongata, Plexaura homomalla and Pseudoplexaura porosa were used as outgroup in both the ML and Bayesian analyses.

Biogeographic analyses

The geographic distribution of the plexaurid species was based on collection localities gathered from the literature (Sánchez et al. 2003; Wirshing et al. 2005; McFadden et al. 2006, 2011; Thoma et al. 2009; Doughty et al. 2014; Vargas et al. 2014). Seven biogeographical areas were considered: (A) Eastern Pacific (Panama and Galapagos), (B) Caribbean, (C) North-West Atlantic, (D) North-Eastern Atlantic (Galicia, Bay of Biscay), (E) Mediterranean Sea, (F) Central Indo-Pacific (Indonesia, Papua New Guinea, Palau, Western Australia and Philippines) and (G) South-West Pacific (Supplementary Table 2). For simplicity, the Gulf of Mexico and the Caribbean were included in a single geographic area (B). Ancestral distributions were reconstructed using the statistical Dispersal-Vicariance (S-DIVA) method (Yu et al. 2014) and the Dispersal Extinction Cladogenesis (DEC) model (Ree and Smith 2008), both implemented in RASP 3.2. S-DIVA is an extension of classical DIVA (see Ronquist 1997, 2001) that reconstructs ancestral areas using sets of trees (e.g. those samples by the Markov Chain in Bayesian analyses) instead of a fixed topology (Yan et al. 2010). The DEC model estimates the geographic range of a given species taking into account stochastic processes such as range expansion and contraction (Ree and Smith 2008). Hence, these methods treat tree topology as a nuisance parameter and allow to account for the effect of uncertainty in tree topology on ancestral area optimization. Both methodologies are of special interest in groups, such as octocorals (e.g. Wirshing and Baker 2015; see also Supplementary Fig. 1 and Supplementary Fig. 2), where phylogenetic signal is weak at shallow scales and topological uncertainty is high. In these groups, integrating over a set of possible topologies allows to extract the biogeographic signal that is present in the set of trees independently of whether the summary of the trees (e.g. the consensus) is highly supported or not.

Using either S-DIVA or DEC, the reconstruction of the biogeographic history of Paramuricea was performed using a set of 10,000 trees generated for the Bayesian analysis and sampled 1000 random trees from the posterior tree space. Ten percent of the total sampled trees were discarded as burn-in, and the maximum clade credibility (MCC) tree from the BEAST analysis was used as the summary tree to display ancestral area reconstruction. The maximum number of areas allowed per node was four.

Molecular dating

For many soft-bodied organisms such as octocorals, with little or no fossil remains, the estimation of divergence times among lineages rely on the use of geological events or mutation rates to calibrate the molecular clock and provide taxon-ages. We used the Bayesian MCMC analysis implemented in BEAST 2.4.4 (Bouckaert et al. 2014) under a GTR + Γ + I substitution model, and a strict or a relaxed molecular clock model to date the mtMutS tree here inferred (see above). Due to the lack of fossils or of information on the substitution rates of Mediterranean gorgonians, we used the only rates so far available for octocorals, calculated for the genus Leptogorgia, and inferred dated trees using diverse strategies to explore how different analytical settings affected molecular dating. We assumed a general mutation rate of mtMutS ranging between 0.14 and 0.25% Myr−1 (see Lepard 2003 in Thoma et al. 2009 and Pante et al. 2015) and tested a conservative (assuming 0.14% mutation rate) and a speculative (assuming 0.25% mutation rate) molecular dating for Paramuricea. In addition, in order to prevent biases derived by the use of a single (fixed) molecular mutation rate, we modelled the clock rate using a log-normal prior with a mean 0.195% ± 0.055 Myr−1 and a uniform prior ranging between 0.14 and 0.25%. We tested strict and relaxed clock (uncorrelated log-normal) models, and used a Yule model as this is suitable for trees describing species-level processes. For each molecular clock analysis, the Markov chain was run for 10 million generations, sampling every 1000 generations. The effective sample size (ESS) of each parameter was assessed using Tracer 1.6 (Rambaut and Drummond 2009). Ten percent of the trees were discarded as burn-in, and the remaining set of trees was visualized using DensiTree 2.0 (Bouckaert 2010) (Supplementary Fig. 1 and Supplementary Fig. 2). The MCC tree with mean divergence times and 95% highest posterior density (HPD) interval was summarized in TreeAnnotator. In order to check the influence of the priors on the posterior distribution estimates, each analysis was re-run using ‘empty’ alignments, allowing to sample from the prior (s) only. Each analysis was run for 100 million generations sampling every 1000 generation. A summary of the trace estimates and a comparison between the marginal prior and posterior densities obtained using different clock models are reported in the Supplementary Fig. 3 and Supplementary Fig. 4.

Results

Mitochondrial genomes

The mitogenomes of P. clavata and P. macrospina were 18,669 base pairs (bp) and 18,921 bp, respectively (Fig. 1a). The architecture of the two genomes, in terms of genome content and gene order, was consistent with that of other octocorals with genome arrangement A (see Brockman and McFadden 2012). Both genomes contained all 14 mitochondrial protein-coding genes reported in octocorals, the 12S rDNA (SSU) and 16S rDNA (LSU) and one tRNA gene (tRNAMet). GC-content was 37% in P. macrospina and 37.3% in P. clavata. Except for Nad2 and Nad4, which overlapped by 13 bp, the remaining genes were separated by intergenic regions (IGRs) of variable length. In both species, the shortest IGRs were 12S rDNA-Nad1 and 16S rDNA-Nad2, which were 4 bp long. The longest IGR was located between Cox1 and 12S rDNA (150 bp) in P. clavata and between Nad5 and Nad4 (208 bp) in P. macrospina. In total, non-coding regions spanned about 4% of the genome in P. clavata and 5.4% in P. macrospina. Among the intergenic regions, the highest variability in terms of sequence length was observed between Cob and Nad6 (13 and 178 bp in P. clavata and P. macrospina, respectively). Between Nad5 and Nad4, we found a variable IGR with single base substitutions and insertion/deletion (Fig. 1b). Sequence variability across the two genomes was 2.6%, excluding gaps. The highest level of nucleotide diversity (π) was found within Nad4 (3.59%), mtMutS (3.46%) and Nad5 (2.75%). The least variable gene was Atp8 (0.46%) and other conserved regions were identified within Nad1, Cox3 and Atp6 (Fig. 1b).

Fig. 1
figure 1

Mitochondrial genome structure and genes variability. a Mitogenomes of Paramuricea clavata and Paramuricea macrospina with genome size and gene annotation. GC-content and AT-content are shown in blue and green on the inner and outer surface of the ring, respectively. b Sliding window analysis of the complete mitochondrial genomes of P. clavata and P. macrospina. The black line indicates nucleotide diversity across the miogenome in a window of 500 bp (10 bp steps). Coding protein genes are in grey; white boundaries represent rDNA genes, tRNA gene and 14 intergenic regions (IGRs). The IGRs between Cob-Nad6 and Nad5-Nad4 are highlighted in yellow and blue, respectively; alignments are showed in the two panels below

Full size image

Molecular phylogeny

The phylogenetic tree of the partial mtMutS gene (~750 nt) showed that the 17 Paramuricea species sampled grouped in a clade supported by high bootstrap (BP = 89) and posterior probability (PP = 0.95) values and that the closest related genera were Echinomuricea, Menella and Placogorgia (Fig. 2). Among Paramuricea, most of the shallow nodes are poorly supported (BP <70; PP <0.95) and the phylogenetic relationships within NW Atlantic specimens were not resolved. Twenty-six Paramuricea specimens were divided in three clades: the first (I) includes P. clavata and P. grayi; the second (II) groups Paramuricea sp. (DQ297420), P. nr. biscaya (KF856184) and three different Paramuricea sp. types (C, G, H) and the third (III) consists of P. macrospina, P. multispina, P. biscaya, P. nr. biscaya (KF856209), P. nr. grandis and five different Paramuricea sp. types (e.g. A, B, D, E, F).

Fig. 2
figure 2

Phylogenetic relationships, divergence times and ancestral area reconstructions of the genus Paramuricea. The tree is a chronogram inferred using a strict molecular clock and conservative mutation rate (0.14% per million year) based on a BEAST Markov chain Monte Carlo (MCMC) analysis. Numbers above and below the nodes represent posterior probability (PP) and bootstrap (BP) values, respectively. Dashed lines indicate branches which lead to nodes with PP <0.95 and BP <70. Roman numbers indicate the three main groups within the genus Paramuricea. The Eastern Atlantic and Mediterranean species analysed in this study are shown in bold. The pie charts indicate the ancestral area reconstruction of a given node based on seven pre-assigned geographic distribution (A-G) for the S-DIVA analysis. The biogeographic regions considered for the analysis are the following: (A) E Pacific (Panama and Galapagos), (B) Caribbean, (C) NW Atlantic, (D) NE Atlantic, (E) Mediterranean Sea, (F) Central Indo-Pacific (Indonesia, Papua New Guinea, Palau, Western Australia and Philippines) and (G) South-West Pacific. Node bares indicate 95% HPD age ranges. The time-event curves for dispersal (black) and vicariance (red) events obtained with RASP 3.2 are shown on the bottom side of the chronogram. The scale bar is in million years

Full size image

Historical biogeography of Mediterranean Paramuricea species

According to our S-DIVA analysis, a total of six dispersal and eight vicariance events occurred during the evolutionary history of the genus Paramuricea (Fig. 2). Node 1 included all 26 Paramuricea specimens from four main geographic areas (e.g. Caribbean, North-West Atlantic, North-Eastern Atlantic and Mediterranean Sea). S-DIVA assigns this node, the ancestor of Paramuricea, to four different ancestral area reconstructions with similar frequency: NW Atlantic + Mediterranean (30%), NW Atlantic + NE Atlantic (25%), NW Atlantic + NE Atlantic + Mediterranean (25%) and Mediterranean (19%). The most probable ancestral range at node 2 (i.e. P. clavata + P. grayi ancestor) is NE Atlantic + Mediterranean with an occurrence frequency of 100%. This reconstruction suggests that P. clavata and P. grayi originated through vicariance of this Atlanto-Mediterranean ancestor. S-DIVA reconstructs a Caribbean + NW Atlantic as the ancestral area of node 3 with an occurrence frequency of 100%. This node is further split in two lineages with distributions in the North Atlantic (node 4) and the Caribbean (Paramuricea sp. H). For node 5 (i.e. the ancestor of the clade including P. macrospina), S-DIVA postulates a NW Atlantic + Mediterranean ancestral range with an occurrence frequency of 93%. This ancestral distribution implies a vicariance led to the separation of P. macrospina in the Mediterranean and the ancestor of clades 6 and 7 in the NW Atlantic. These nodes (i.e. 6 and 7) are predicted to have two possible ancestral ranges: Caribbean (61%) and Caribbean + NW Atlantic (39%), and NW Atlantic (72%) and Caribbean + NW Atlantic (28%), respectively. The ancestral distributions obtained using a DEC model were consistent with those obtained by S-DIVA (see Supplementary Fig. 5).

Molecular dating the Mediterranean Paramuricea endemics

Using either a conservative or a speculative molecular rate (see Fig. 2 and Supplementary Fig. 6), molecular dating resulted in a divergence time of 4.6 Ma (95% HPD 3–7 Ma) or 2.6 Ma (95% HPD 1–4 Ma), respectively, for P. clavata. The speciation event leading to the diversification of P. macrospina from a NW Atlantic + Mediterranean ancestor was dated at 5.8 Ma (95% HPD 3–8 Ma) or 3.2 Ma (95% HPD 2–5 Ma). Similar divergence times were recovered regardless of the clock model and molecular rates used, indicating that the dates obtained are robust to changes in the methods used to infer them. Details on the age estimates for the Mediterranean species are reported in the Supplementary Fig. 7.

Discussion

Mitogenomics, phylogeny and evolution of Paramuricea

The genus Paramuricea is restricted in the Mediterranean Sea to two endemic species that according to our phylogenetic analyses are not closely related. The level of mitochondrial sequence divergence between these two species was high (~2.6%), well above the 1% threshold proposed for mtMutS-based species delimitation in octocorals (McFadden et al. 2011). Intrageneric nucleotide diversity in octocorals changes according to the taxonomic group considered and the molecular marker used. In the Mediterranean Paramuricea the Nad5 and Nad4 showed the highest mutation rate among mitochondrial protein-coding genes, and the intergenic region between these two genes was remarkably different in terms of size (100–500 bp). For members of Paramuricea, this particular mitogenomic region (Nad5-igr-Nad4) may be of interest as a complementary barcode to the mtMutS and Cox1 regions and should be further investigated. Differences in the length of IGRs in non-bilaterian metazoans (i.e. Cnidaria, Ctenophora, Placozoa and Porifera) contribute wide range of size variation observed in the mitochondrial genomes of these animals and is mainly caused by the presence of repetitive elements such as hairpin-forming elements, palindromes and duplicated segmental DNA (Erpenbeck et al. 2009; Lavrov 2010; Park et al. 2011). Octocorals display a wide range of mitochondrial genome sizes—between 18,668 and 20,246 bp among sequenced members of Holaxonia reported in GenBank and the two genomes obtained in this study. Interestingly, neither direct variant repeats (DVRs) nor duplicated segmental DNA, as it has been reported in Calicogorgia granulosa (for details, see Park et al. 2011), were found in any of the other published octocoral mitochondrial genomes or those sequenced here leaving open questions about the molecular mechanisms responsible for mitochondrial genome expansion/contraction in octocorals. A broader sampling effort across the subclass may help to better understand mitochondrial genome evolution among octocorals.

Concerning our phylogenetic results, they showed the shallow-water P. clavata sister to NE Atlantic P. grayi and P. macrospina closely related to deep-sea congeners from the Gulf of Mexico and NW Atlantic. Our analyses also revealed a close genetic relation between the genera Paramuricea, Echinomuricea and Placogorgia, in agreement with previous studies (see Wirshing et al. 2005; McFadden et al. 2011). The genus Echinomuricea includes approximately 30 species with a broad geographic distribution, but generally is regarded as limited to the Indo-West Pacific (Williams and Chen 2014). Grasshoff (1977) reassigned the Mediterranean endemic Echinomuricea klavereni to Spinimuricea and the status of the Echinomuricea species from the West Indies and the Gulf of Mexico is not clear and needs revision. Similarly, Placogorgia represents another poorly known group in need of taxonomic and systematic treatment. It is interesting to note that these three genera as well as others with thornscale or thornspindle sclerites (i.e. Muriceides, Villogorgia) were grouped into the same family (Paramuriceidae) by Bayer (1956), now disregarded. The phylogenetic affinity of paramuriceids (sensu Bayer 1956) with similar sclerites highlights the importance of morphological characters in evolutionary studies. For instance, the thornscale sclerites, reported in the family Plexauridae, represent a synapomorphic character that was likely present in the most recent common ancestor of this family.

Historical biogeography, vicariance and speciation

Paramuricea and most members of the family Plexauridae occur in different biogeographic areas such as the Atlantic and Pacific oceans and the Mediterranean Sea. Due to their broad geographic distribution, a Tethyan common ancestor has been proposed for this family (Bayer 1961; Vargas et al. 2010). The Mediterranean is considered to be the evolutionary centre of Tethys fauna remnants (Dov Por 2009) and putative Tethyan relicts have been found in different groups of organisms in this region (Pérès and Picard 1964; Boury-Esnault et al. 1992; Dov Por 2009). These relicts are thought to have survived the Messinian and Gelasian crises (5.2 and 2.6 Ma), during which a dramatic decrease in the number of marine species occurred due to the extreme geological and climatic changes driven by the closing of the strait of Gibraltar and the following desiccation of the Mediterranean Sea (Hsü et al. 1973).

Among octocorals, the existence of early Miocene (~20 Ma) fossils belonging the family Corallidae (Vertino et al. 2014) suggests that a Mediterranean ancestor of this group survived the Messinian crisis 5.2 Ma. In the case of octocorals, where fossils are lacking, it is hard to establish whether Mediterranean species are derived from an Atlantic ancestor (i.e. they speciated after an ancestral range expansion into the Mediterranean Pérès 1985; Sarà 1985) or speciated in situ from a wide spread Atlanto-Mediterranean ancestor. In historical biogeography, however, the distribution of extant species can be used to reconstruct ancestral distributions and clarify the speciation process of taxa when other information sources are lacking. Our biogeographic analysis suggests that the two Mediterranean endemic species originated from two independent vicariance events and we pose that these events are related to the Messinian and Gelasian crises. Interestingly, our molecular clock estimates are in line with this hypothesis. Based on conservative or speculative mutation rates, the divergence estimates of 4.6 and 2.6 Ma for P. clavata and ~5.9 ± 0.2 and 3.3 ± 0.1 Ma for P. macrospina match the geological age of these crises. Although these estimates are based on mutation rates calculated for the genus Leptogorgia, they seem to be robust to variations in the analytical setting used. Hence, the origin of the Mediterranean Paramuricea seems to be in agreement with the canonical hypothesis about the role played by the Messinian (and Gelasian) crisis in shaping the evolutionary history of the marine fauna of the Mediterranean. Indeed, after the Messinian crisis the Mediterranean Sea was ‘refilled’ with Atlantic water (Garcia-Castellanos et al. 2009). According to our results, this event could have provided the conditions for the ancestral vicariance event we detected leading to the diversification of the ‘new’ Mediterranean biota due to the new environmental conditions offered by this basin. Dispersal has been proposed as the main mechanism responsible for biogeographic patterns in the sea (see, e.g. Lessios and Robertson 2006), yet our results highlight the importance of vicariance as a driving-force for speciation in the ocean. In this respect, recent studies on the evolution and speciation of Mediterranean organisms (e.g. Domingues et al. 2005; Tougard et al. 2014) reveal that the endemic species are either the result of inter-basin speciation after the Messinian Salinity Crisis or derived from Atlantic ancestor populations which became extinct after the ‘colonization’ of the Mediterranean.

Despite our phylogenetic trees included some branches with uncertainty, the majority of the deep nodes are well supported and the ancestral reconstruction analysis seems to be robust to topological uncertainty. In this regard, the high occurrence frequency values obtained by S-DIVA indicate that the underlying biogeographic signal is strong enough to overcome the topological uncertainty present in the phylogeny of Paramuricea at shallow levels. Finally, we would like to highlight the somewhat unexpected high phylodiversity of the Mediterranean Paramuricea, which represent two highly divergent lineages resulting from independent speciation events. Interestingly, this diversity is also reflected in the ecology of the endemic Paramuricea species with P. clavata occurring along the littoral and P. macrospina being restricted to deeper habitats.

We hope our study provides further information on the genus Paramuricea in the Mediterranean that can be used to guide conservation planning and decision making in the area by taking into account the evolutionary and biogeographic history of these organisms.