Samples were obtained from targeted collecting trips and specimens deposited at different museums or collections: the Museum of the “Charles Darwin” Department of Biology and Biotechnologies, La Sapienza University, Rome, Italy (BAU); the British Museum of Natural History, London, United Kingdom (BMNH); the Invertebrate Zoology collection at the California Academy of Sciences, San Francisco, United States (CASIZ); the Colección Nacional de Moluscos, Universidad Nacional Autónoma de México, Mexico City, México (CNMO); the California State Polytechnic University Invertebrate Collection, Pomona, United States (CPIC); the Museo Nacional de Ciencias Naturales, Madrid, Spain (MNCN); and the Museu Nacional de História Natural e da Ciência, Lisbon, Portugal (MUHNAC, formerly Museu Bocage MB).

DNA extractions were performed using the Qiagen DNeasy Blood and Tissue Kit following the manufacturer’s instructions, with some minor changes (100 μL final extraction instead of 200 μL). Partial sequences of COI, 16S and H3 were amplified by PCR using LCO1490 and HCO2198 universal primers for COI (Folmer et al. 1994), 16S ar-L and 16S br-H for 16S (Palumbi et al. 1991), and H3AD5’3’ and H3BD5’3’ for H3 (Colgan et al. 1998). The master mix for the PCR was carried out in 25 μL volume reactions. PCR contained 2.5 μL of Qiagen buffer, 2 μL of DNA, 2.5 μL of dNTP (2 mM), 5 μL of Q-solution (Qiagen), 1.5-3.5 μM magnesium chloride, 1 μL of each forward and reverse primer (10 µM), 0.25 μL of DNA polymerase (250 units µ^–1), 2-3 μL of DNA template, and nuclease-free water. Successful PCR products were purified and sequenced by Macrogen, Inc. All new sequences obtained were deposited in GenBank. COI amplification was performed with an initial denaturation for 3 min at 94-95°C, followed by 39-40 cycles of 30-45 s at 94°C, 30-45 s at 46°C (annealing temperature) and 1-2 min at 72°C, with a final extension of 5 min at 72°C. 16S amplification was performed with an initial denaturation for 3 min at 94-95°C, followed by 39 cycles of 39-45 s at 94°C, 30-50 s at 45-51.5°C (annealing temperature) and 2 min at 72°C, with a final extension of 5-10 min at 72°C. H3 amplification was performed with an initial denaturation for 3 min at 95°C, followed by 40 cycles of 45-60 s at 94-95°C, 45 s at 50°C (annealing temperature) and 2 min at 72°C, with a final extension of 10 min at 72°C.

Molecular analysis included a total of 21 specimens including eight species of Felimare, four specimens of three other genera of Chromodorididae Bergh, 1891 and one specimen of Prodoris clavigera (Thiele, 1912), originally ascribed to Bathydoris Bergh, 1884, as an outgroup (Table 2).

DNA sequences were assembled, edited and aligned using Geneious 8.1.2 (Kearse et al. 2012). The alignments were checked by eye. All the sequences were checked for contamination with BLAST (Altschul et al. 1990) implemented in the GenBank database. Protein-coding sequences were translated into amino acids for confirmation of alignment. Pairwise uncorrected p-distance values between each taxon, uncorrected p-distances between all taxa, and level of saturation for first, second, and third codon positions (p-distances against transitions plus transversions) were calculated in MEGA 5.0 (Tamura et al. 2011) for the COI and H3 genes. The most variable regions from the 16S rRNA alignment were removed in the first analyses, using both the default settings and the standard options for stringent and less stringent selection in Gblocks (Talavera and Castresana 2007). When these regions were excluded from the analyses, the combined phylogenetic tree was poorly resolved and with low node support. Therefore, final analyses were performed including all bases. The best-fit models of evolution for each gene were determined using the Akaike information criterion (Akaike 1974) implemented in MrModeltest v. 2.3 (Nylander 2004), resulting in the GTR+I+G model for COI, 16S and H3. Maximum likelihood (ML) analyses were performed using the RA×ML software v7.0.4 (Stamatakis 2006) and node support was assessed with non-parametric bootstrap (BS) with 5000 replicates, random starting trees, and parameters estimated from each dataset under the model selected for the original dataset. Bayesian Inference (BI) analyses were conducted using MrBayes version 3.1.2b (Ronquist and Huelsenbeck 2003) for five million generations with two independent runs and a sampling frequency of 1000. The models implemented were those estimated with MrModeltest v. 2.3. Convergence was diagnosed graphically by plotting for each run the likelihood against the number of generations using the software Tracer version 1.4.1 (Drummond and Rambaut 2007). For each analysis, the first 1250 trees were discarded (‘burn-in’ period). Node support was assessed with posterior probability (PP). BI and RA×ML phylogenetic analyses were performed on the 280-core “PhyloCluster” hosted at the Center for Comparative Genomics, California Academy of Sciences. Only nodes supported by PP ≥ 0.95 and BS≥75 were considered as resolved. The combined tree provided better resolution than COI (658 pb), H3 (328 pb) or 16S (up to 476 pb) separately. The combined dataset yielded a sequence alignment of 1462 positions. The ABGD method (Puillandre et al. 2012) was performed for the COI alignment using the online version of the software (available at http://wwwabi.snv.jussieu.fr/public/abgd/) (18 Nov 2016). ABGD was run by selecting Jukes-Cantor and Kimura parameters distance, Pmin=0.001, Pmax=0.1, Steps=10, and relative gap width=1.

Two specimens of Felimare were obtained in two different surveys in May 2009 and July 2011 through scuba diving at Tarrafal, Cape Verde, and preserved in 96% ethanol. One specimen was dissected by dorsal incision. The internal features were examined using a dissecting microscope and drawn with a camera lucida. The buccal mass was removed and dissolved in 10% sodium hydroxide until the radula and the labial cuticle were isolated from the surrounding tissue. Both were then rinsed in water, dried and mounted for examination under a Quanta 200 scanning electron microscope.

Morphological and anatomical comparison between the new species and congeners from the eastern Atlantic blue chromodorid chromatic group was based on published information and personal observations.

We successfully obtained 64 new sequences; 12 additional sequences were obtained from GenBank (Table 2). The combined tree of COI, 16S and H3 provided better resolution than individual genes trees (Fig. 1). The individual genes trees can be seen in the supplementary material Figure S1. Figure 1 shows the phylogenetic hypothesis based on the combined dataset constructed by BI. The topology of the ML tree was almost identical (not shown).

The genus Felimare sensu Johnson and Gosliner (2012) was recovered by the BI (ML=not recovered), and their species were arranged into two major clades: one clade including Felimare porterae (Cockerell, 1901) (PP=1, BS=100) and one containing the remaining species of Felimare included in this study (PP=0.99). This last clade is subdivided into two subclades: one (PP=1, BS=79) comprised by Felimare bilineata (Pruvot-Fol, 1953), Felimare n. sp., Felimare pinna (Ortea, 1988), and Felimare tricolor (Cantraine, 1835); and one (PP=1, BS=99) comprised by Felimare picta (Schuitz in Philippi, 1836) and Felimare tema.

The maximum intraspecific pairwise uncorrected p-distance for COI between specimens of F. bilineata was 4.23%. The minimum interspecific p-distance within Felimare was between Felimare n. sp. and F. bilineata and was 8.63% (Table 3). Both ABGD species delimitation method analyses recovered eight partitions with prior maximal distance P=0.001 for 1 group and P=0.035 for 8 groups. (Fig. 1).

Family CHROMODORIDIDAE
Genus Felimare Ev. Marcus and Er. Marcus, 1967

Felimare aurantimaculata n. sp.
(Figs 2, 3 and 4)

Hypselodoris sp., Wirtz, 2009: 54, Fig. 2.

Holotype: MB28-004390, 30 mm in length alive, Peter Wirtz, Tarrafal, Santiago Island, Cape Verde, 2009. Paratype: MB28-004391, dissected, 50 mm in length alive, Emanuel de Oliveira, Tarrafal, Santiago Island, Cape Verde, 23 m deep, 08 June 2011.

External morphology. Body high and elongate, with dark blue colour and smooth surface. Small round orange spots all over the body, including the foot (Fig. 2). Mantle edge narrow. Dorsum with a series of large and unclustered MDFs on the edge of the mantle, totally absent in the anterior area facing the rhinophores, 15 on the 50 mm specimen (MB28-004391), and 9 on the 30 mm specimen (MB28-004390). MDFs easily seen through the mantle. Posterior end of the foot not covered by the notum (Fig. 2). Rhinophores and branchial leaves dark blue, slightly lighter in colour. Rhinophores of two specimens with 22 lamellae, arranged nearly horizontal. Ten unipinnate branchial leaves in the 50 mm specimen (MB28-004391).

Anatomy (Figs 3, 4). Jaw composed of two pieces covered by unicuspid rodlets (Fig. 3A, B). The radular formula of the 50-mm specimen (MB28-004391) is 69×114.0.114. (Fig. 3C). Without rachidian radular tooth. Innermost lateral teeth sharp bifid, inner lateral teeth with two small denticles on their outer edge (Fig. 3D). Middle lateral teeth curved and bicuspid (Fig. 3E). Outermost teeth slightly bifid, broad tips, with up to four denticles on outer edge (Fig. 3F).

Reproductive system (Fig. 4A, B) with vestibular gland elongated. Vagina very wide and muscular. Pyriform seminal receptacle with a short and thin conduct that joins along the middle length of the vagina. Bursa copulatrix spherical. Uterine duct short and narrow, entering the female gland near the entrance of the oviduct. Ampulla elongated and wide. Penis unarmed. Prostate elongated, located anteriorly to female gland, narrowing slightly to a convoluted deferent duct, all over the reproductive system. Female gland half the size of the entire reproductive system.

Distribution. To date only known from Tarrafal, Santiago Island, Cape Verde.

Etymology. The name refers to the Latin words aurantiacus, meaning orange, and maculatus, meaning spot, which refers to the chromatic pattern of the body, scattered with orange polka dots.

Comparative diagnosis. Though all the Felimare species distributed along the Atlantic coast of Africa and belonging to the blue Atlantic chromatic group have some chromatic yellow pattern over light to dark blue, Felimare aurantimaculata n. sp. is the only one having small orange dots instead of a white middle line with a distinct and unique pattern as in Felimare pinna; parallel white line or yellow stripes as in Felimare bilineata, Felimare ciminoi (Ortea, Valdés and García-Gómez, 1996), Felimare francoisae (Bouchet in Bouchet and Ortea, 1980), Felimare garciagomezi (Ortea and Valdés, 1996), Felimare tema, Felimare tricolor and Felimare xicoi (Ortea and Valdés, 1996); or irregular patterns as in Felimare gofasi (Ortea, Valdés and García-Gómez, 1996). The morphotype of Felimare picta from the Azores Islands has small yellow spots in a lower density all over the body that make it resemble Felimare aurantimaculata n. sp., but our results show that they are different species nested in different subclades (Fig. 1). The size of the MDFs of Felimare aurantimaculata sp. nov. is larger than that of the remaining known species of the genus. Moreover, the arrangement of the MDFs in Felimare aurantimaculata n. sp. is also unique among the above blue chromatic group, since the remaining species lack MDFs in the middle region of their mantle edge or they are limited to the posterior end (Table 4). The species F. francoisae and F. garciagomezi (previously named as Mexichromis) differ from Felimare aurantimaculata n. sp., besides the chromatic pattern, in having jaw rodlets with 3-4 cuspids, instead of unicuspid rodlets as in F. ciminoi and F. gofasi (Table 4). The reproductive system is like that of other species of Felimare, formerly ascribed to Hypselodoris, such as the size and shape of the vestibular gland; the muscularized wide vagina; the spherical shape of the bursa copulatrix; and the pyriform seminal receptacle, generally attached in some part of the vagina, with the exception of Felimare molloi, that joins just at the entrance of the bursa copulatrix (Ortea et al. 1996, Gosliner and Johnson 1999).

Recently, using molecular techniques, Almada et al. (2016) and Furfaro et al. (2016) ended the controversy related to the Felimare picta complex. They both stated that Felimare tema (Edmunds, 1981) is a valid species distributed from Cape Verde and Senegal to Angola, in the southern hemisphere, while Felimare picta is restricted to the northern hemisphere, but with a broader distribution, from the type locality in the Mediterranean sea up to the Gulf of Mexico. Although there is now no doubt that F. picta is not distributed in the southeastern Atlantic, we decided to include F. picta in our phylogenetic analyses to discard the possibility that Felimare aurantimaculata n. sp. could be a chromatic morphotype of this species. IB, ML, p-distance and ABGD strongly support the hypothesis that Felimare aurantimaculata n. sp. is a new valid species.