A review of the Tripterygion tripteronotus (Risso, 1810) complex, with a description of a new species from the Mediterranean Sea (Teleostei: Tripterygiidae)

Molecular data provide a complementary approach to discriminate species separated by subtle morphological characters (Knowlton, 1993; Avise, 1994; Held and Wagele, 2005). In the last few years, numerous authors have used molecular methods to detect cryptic species, either in fishes (Gilles et al., 2000; Gysels et al., 2004; Almada et al., 2005a) or in other marine organisms (Tarjuelo et al., 2001). The family Tripterygiidae contains species of bottom-living blennioid fishes, usually associated with rocky habitats and inhabiting cold, temperate, subtropical and tropical areas (Fricke, 2002). The SCIENTIA MARINA 71(1) March 2007, 75-86, Barcelona (Spain) ISSN: 0214-8358


INTRODUCTION
Molecular data provide a complementary approach to discriminate species separated by subtle morphological characters (Knowlton, 1993;Avise, 1994;Held and Wagele, 2005).In the last few years, numerous authors have used molecular methods to detect cryptic species, either in fishes (Gilles et al., 2000;Gysels et al., 2004;Almada et al., 2005a) or in other marine organisms (Tarjuelo et al., 2001).
The family Tripterygiidae contains species of bottom-living blennioid fishes, usually associated with rocky habitats and inhabiting cold, temperate, subtropical and tropical areas (Fricke, 2002).The genus Tripterygion Risso, 1826, is the only genus of the family Tripterygiidae in the Mediterranean Sea and on the northeastern Atlantic coast (Zander, 1986).Three species have been described: T. tripteronotus, Risso, 1810, andT. melanurus, Guichenot, 1845, are endemic to the Mediterranean, and T. delaisi Cadenat and Blache, 1971, is found in both areas (Wirtz, 1980).Individuals of the three species are common in shallow coastal waters, always living in rocky areas.T. tripteronotus inhabits light-exposed and shady biotopes preferably between 0 and 3 m, whereas T. delaisi uses similar biotopes but at greater depth (between 0 and 40 m) and also biotopes with reduced light such as under overhanging rocks or entrances to caves.Finally, T. melanurus inhabits walls or ceilings of marine caves and other dimly lit biotopes (Wirtz, 1978;Macpherson, 1994;Zander, 2004).
The species of the genus Tripterygion form a monophyletic group and each previously described species is well differentiated genetically (Carreras-Carbonell et al., 2005).However, this recent phylogeographic study, using molecular data, indicated that: (1) the two morphotypes of T. melanurus, traditionally considered as two different subspecies by Zander (1986), were not genetically different with the markers used, although there may be differences on other parts of the DNA sequence, (2) the two currently accepted subspecies for T. delaisi (Zander, 1986) were molecularly validated, and (3) T. tripteronotus, considered at present as a single species, showed two well-defined and highly supported clades with greater divergence than that shown between the two T. delaisi subspecies, revealing the existence of two cryptic species within T. tripteronotus (Carreras-Carbonell et al., 2005).Zander and Heymer (1970) had already described two different pattern bands in the caudal region for T. tripteronotus individuals from Banyuls-sur-Mer (France) and Mdiq (Morocco).Later, Zander and Heymer (1976) showed slight morphological differences in the dorsal fins between T. tripteronotus specimens from Israel and Lebanon in comparison with specimens from the northwestern Mediterranean.Although no taxonomic status was assigned, these morphological differences could be related to the two T. tripteronotus clades found by Carreras-Carbonell et al. (2005).
The aim of the present work was to describe the new species and search for morphological characters that allow the two species to be differentiated using specimens from 52 localities of the Mediterranean Sea and adjacent waters.

Sampling and repositories
Specimens of the two species of Tripterygion were collected at different localities of the Mediterranean Sea and the Gulf of Cadiz; specimens from the Staatliches Museum fuer Naturkunde (Stuttgart, SMNS) were also used, with the result that the individuals came from a total of 52 localities (Fig. 1).The number of individuals used for morphological and molecular analyses, as well as supplementary details about each sampling locality, are shown in Table 1.
The type series of the new species are deposited in the collections of the Instituto de Ciencias del Mar (Barcelona, IIPB), the Museo Nacional de Ciencias Naturales (Madrid, MNCN) and the Staatliches Museum fuer Naturkunde (Stuttgart, SMNS) (see Table 1).

Morphological analysis
In the description of the new species, the data of the paratypes follow those of the holotype, in parentheses.Lengths given and the terminology and other measurements used mainly follow Zander and Heymer (1970), Wheeler and Dunne (1975) and Fricke (1997).Lengths are explained below: Predorsal length (PD) distance between middle of upper lip and base of the 1 st spine of the first dorsal fin.
Head length (HL) distance between middle of upper lip and upper insertion of operculum.
Orbital diameter (OD) maximum eye diameter.Preorbital length (PO) distance between middle of upper lip and anterior margin of eye.
The middle of the upper lip is used as the starting point for several lengths rather than the tip of the upper jaw, as the latter may be protractile.
Mandibular pore formula.This formula gives the number of pores under left dentary + number of median pore(s) + number of pores under right dentary.
Individuals were photographed alive in order to check their colour pattern; one or two right gills were removed and kept in absolute ethanol at room temperature.Specimens were individually fixed using buffered formol with 2% borax to maintain the colour pattern for further morphological analyses.

Key
The morphological taxonomic key only works for both sexes when morphometric measurements are used.Sometimes, males can also be distinguished by discrete morphological characters, while females are identifiable only by their geographical distribution and accompanying males.
The homogeneity of base composition across taxa was assessed using the goodness-of-fit (χ 2 ) test and the incongruence length difference test (ILD) (Farris et al., 1994) was computed to assess analytical differences between genes; both tests are implemented in PAUP* ver.4.0b10 (Swofford, 2001).In the latter test only parsimony informative characters were included and heuristic searches were performed with 10 random stepwise additions with TBR branch swapping and 1000 randomisations.Furthermore, trees were considered significantly incongruent whenever different gene trees conflicted at nodes that were supported by BI posterior probabilities >95% (Moyer et al., 2004).
Phylogenetic trees were inferred by Bayesian inference (BI) using Mr Bayes 3.0b4 (Huelsenbeck and Ronquist, 2001) because it seems to be the best methodology for inferring phylogenetic relationships between species (Alfaro et al., 2003), and its reconstruction does not seem to be affected by saturated positions (Carreras-Carbonell et al., 2005).The computer program MODELTEST ver.3.06 (Posada and Crandall, 1998) 1 for locality abbreviations and further details.resulting in 15000 trees.The first 1500 trees were eliminated since they did not reach the stationarity of the likelihood values and the rest were used to construct the consensus tree and obtain the posterior probabilities of the branches.
(Figs. 2 and 3a) Etymology.The name tartessicum referred to the old Spanish culture (Tartessos, at least dating from 1000 BC) located on the south coast of the Iberian peninsula (in modern Andalusia, Spain), where the new species is partially distributed.
Morphological description.Body elongate and compressed, greatest height at base of anal fin, being about one-sixth total length.Scales ctenoid, covering entire body except base of pectoral fin and ventral abdominal region back to vent.Lateral line having two sections: anterior section with 20 (19-22) pored scales, posterior section with 22 (21-24) notched scales, having 42 (40-46) in total.Upper, anterior, section commencing at upper angle of opercular opening, slightly curving up over pectoral fin base and running parallel to dorsal profile to point below last 1-3 rays of second dorsal fin; canal running across exposed width of each scale.Lower, posterior, section commencing below, and in front of last scale or two of upper section, running along the mid-line of tail to caudal fin base; each scale with shallow notch in free-edge tip.
Three dorsal fins with III + XVI + 13 (III + XVI-XVIII + 12-13) rays.First dorsal fin lower than second and second higher than third.First just above pre-operculum, rays being of equal height.Second separated by short interspaces, origin slightly behind  base of the pectoral fin; first ray longest, in mature males being nearly twice as long as rays in middle region, with distal half not united by membrane with following ray.Base of third fin about 0.6 length of second dorsal fin base.Caudal fin truncate, with X (IX-X) principal branched rays, and 2 (2-3) procurrent lower and upper.
Anal fin elongate and of uniform height, with II + 23 (II + 22-24) rays.Anteriorly, 2 weak, slender, unsegmented rays, first shorter than second, which is slightly longer than the first segmented ray; succeeding rays united by membrane and decreasing in length posteriorly.
Pectorals long and broad, slightly overreaching mid-length of second dorsal fin and base of anal fin; with 16 (15-16) rays, upper three rays short and simple, remainder branched; ninth ray, counted from upper edge, longer than others.
Pelvic fins with one short spine and two slender and segmented rays; longest ray reaching midlength of pectoral fin.
Head broad, scale less, profile acute, lips prominent.Head length 0.19 (0.16-0.22) times total length (TL).Orbit large, almost circular, diameter 0.32 (0.28-0.51) times head length, upper edge forming ridge along upper head profile.Pre-orbital and predorsal lengths 0.05 and 0.14 times total length, respectively (0.06 and 0.18).Interorbital region concave.Mouth nearly horizontal, maxilla extending to level of front of pupil.Gill membrane continuous across throat.Teeth conical, in band in upper and lower jaws.Anterior nostril tubular, posterior nostril close to orbit edge.Cephalic canal pores as illustrated in Figure 3a, with preopercular-dentary series complete.The mandibular pore formula (Fricke, 1997) was 3+2+3 (3-4+2+3-4), basically depending on the fish TL, suggesting that an increase in length could be associated with the appearance of a new pore in both dentaries.However, no significant relationship was found between this formula and TL, or between the two species.The interorbital series 2 (2-4) opened singly from the upper interorbital region to the upper lip.The preopercular series opened singly along the lower side of the preopercular canal, opening in pairs on the posterior pre-opercular edge.The nasal and suborbital canals usually opened in pairs, running along the lower and the posterior margins of the orbit; nasal pores 3 (1 to 3) placed in front of the anterior border of the eye; Habitat.The new species inhabits similar habitats to T. tripteronotus: shallow rocky shores to 6 m, preferably between 0 and 3 m; in light-exposed and shady biotopes dominated by algal communities (e.g.Corallina elongata, Cladophora spp., Litophyllum spp., Enteromorpha spp.).Nests are usually situated in sciaphyl habitats dominated by steep rocky zones, without arborescent algae.

Morphological data
The morphological comparison of the present material of T. tartessicum with specimens of T. tripteronotus from different localities of the Mediterranean and adjacent waters showed that there are only slight differences between the two species.They can be differentiated by a morphometric measurement: the orbital diameter (OD) is significantly longer in the new species (mean ratio head length/orbital diameter = 2.69±0.36)than in T. tripteronotus (3.16±0.29;Mann-Whitney U-test, p<0.05).When HL/OD was represented in front of TL, two well-differentiated and almost non-overlapping groups were found, corresponding to both species (Fig. 4).In order to assure this differentiation, a multivariate analysis of covariance (MAN-COVA) was implemented using TL as the covariate and HL/OD as the dependent variable.The results showed a highly significant differentiation between the two groups (F = 415.72,p<0.001).
The first ray of the second dorsal fin of the mature males has the distal half not united by a membrane with the following ray in T. tripteronotus, whereas the first two rays can be united by a membrane from their respective tips in T. tartessicum.Additionally, the caudal fin usually has four red or brownish bars (black in preserved specimens) in the new species, whereas these bars are usually not distinct in T. tripteronotus.These two differences are similar to the ones described by Zander and Heymer (1976), although they should be considered with caution since they were not always observable in all the individuals collected.
We also observed that the mating season seems to start slightly later in the new species.In fact, all mature males of T. tripteronotus are active on the Catalan coast (NE Spain) in early May, whereas most mature males of the new species are not active at this time on the coasts of Murcia and Almeria (SE Spain).

Molecular data
We analysed a total of 1732 bp for all genes combined in 55 individuals (18 T. tartessicum and 37 T. tripteronontus).A total of 10 haplotypes were found for T. tartessicum, whereas 17 were shown for T. tripteronotus (Fig. 5).Generally, all individuals from one locality shared the same haplotype, and TK4-TK6 and SP6-SP10 also shared the same haplotype.However, at some localities (SP2, FR1, IT1, IT2, GR2, SP5 and SP11) more than one haplotype was found.For each of the four mitochondrial genes the sequence obtained was of 419 bp for 12S rRNA, 699 bp for 16S rRNA, 73 bp for tRNA-valine and 541 bp for Cytochrome Oxidase I.All genes used showed a similar percentage of parsimony informative sites (chi-square = 7.57 p = 0.36) ranging from 2.74 to 10.35%, but only the RNA genes had similar variable sites (chi-square = 7.52 p = 0.18) ranging from 10.50 to 14.59%, the percentage being higher for COI (19.41%).For the COI protein coding gene, third codon positions were 54.19% variable, second codon positions were invariant and first codon positions were 4.47% variable.The Ts/Tv ratio ranged between 2.61 (16S) and 6.00 (COI) with 4.13 for 12S, and 2.63 for tRNA-valine.There was no evidence of sequence saturation in the analysed genes.For each gene sequence the goodness-of-fit test showed homogeneous base composition across taxa (P = 1.00) and the partition homogeneity test showed no significant heterogeneity between genes (P ILD range from 0.15 to 1.00), and although there is no generally accepted p-value for significant results, most authors agree to combine data when p-values are greater than 0.05 (Cristescu and Hebert, 2002;Russello and Amato, 2004).
As assessed in Carreras-Carbonell et al. (2005), two well-supported clades for T. tripteronotus (northern and southern) were found with posterior probabilities of 100%.The southern clade belonging to T. tartessicum showed no well-supported structure pattern between different localities.However, the northern clade (T.tripteronotus) showed several well-supported subclades that could be related to defined geographical areas (e.g.Cyprus and Turkey), indicating some degree of isolation between different populations (Fig. 5).
Molecular divergence between T. tripteronotus and T. tartessicum ranges between 9.14% (COI) and 2.79% (tRNA-valine), with a mean value combining all genes of 6.89% (Table 2).No genetically and/or morphologically hybrid populations or individuals were found.

DISCUSSION
The new species is geographically distributed along the southern coast of Spain, from Cape La Nao (SP7) to the Gulf of Cadiz (SP12), the Balearic Islands (SP5 and SP6), and northern Africa, from Plage David (MC; Morocco, Atlantic Ocean) to Tunisia (TU1) (see Fig. 1).The eastern boundary in the distribution of the new species is unfortunately unknown.Some morphological characteristics (e.g.rays of the second dorsal fin and caudal bands) of the specimens collected in Israel by Zander andHeymer (1970, 1976) are closely related to those observed in the new species, suggesting the presence of T. tartessicum in that area.
However, as we have mentioned above, these morphological characters are not constant, and unfortunately we could not analyse specimens from this locality.Future studies are recommended to confirm the taxonomic position of this material.
The individuals from Nice (FR2) and Messina (IT2) were grouped within the northern clade of T. tripteronotus, suggesting that all specimens from these localities belonged to the species described by Risso (1810).Therefore, T. melaenocephalus, described by Cocco (1829), can be considered as a junior synonym of T. tripteronotus, in agreement with previous studies (e.g.Zander, 1986).The specimens from the Black Sea, originally identified as T. nikolskii (Maksimov, 1909) and synonymised with T. tripteronotus, could not be analysed.However, the presence of T. tripteronotus on the Aegean coasts of Greece and Turkey, as well as in the Marmara Sea, suggests that the specimens from the Black Sea may belong to T. tripteronotus or T. nikolskii, but not to the new species.
Our results confirm the validity of subtle morphological characters for distinguishing species of the genus Tripterygion, and the existence of a cryptic species, as occurs in other fish taxa (Gleeson et al., 1999;Henriques et al., 2002;Yamazaki et al., 2003).Nevertheless, the criteria used to designate distinct species based on molecular data are always controversial (Cracraft, 1989;Avise, 1994).The genetic divergence between T. tripteronotus and T. tartessicum is 9.14% for COI, 5.32% for 12S and 6.72% for 16S, similar to the divergence observed between other fish taxa.Yamazaki et al. (2003), using COI, found a sequence difference of 9.10±0.36%between two cryptic species of brook lamprey.For 16S, genetic distances between congeneric species of the families Soleidae, Mullidae and Apogonidae range between 4.6 and 11.70% (Tinti et al., 2000;Apostolidis et al., 2001;Mabuchi et al., 2003).Finally, for 12S the mean genetic distance between congeneric species of the genus Coryphaenoides was 3.31% (Morita, 1999), 4% within the genus Macullochella (Jerry et al., 2001) and a mean of 6.5% within different blenniidae genera (Stepien et al., 1997).Henriques et al. (2002), in a revision of the genus Lepadogaster (Teleostei: Gobiesocidae), observed that the minimum distance between valid species was 3% at 12S rRNA.Furthermore, Almada et al. (2005b), using 12S and 16S genes, showed that the genetic differences between clearly morphologically differentiated European blenniid species of the genus Parablennius and Lipophrys were even smaller (1.3-1.6%).Within the genus Tripterygion, T. tripteronotus and T. tartessicum showed the smallest divergence, indicating a more recent speciation event (Carreras-Carbonell et al., 2005).
The estimated divergence time found between the two species was approximately 3.17 Myr when the evolutionary rates of 0.81±0.23%/Myrfor 12S and 1.10±0.23%/Myrfor 16S inferred for the genus Tripterygion (Carreras-Carbonell et al., 2005) were applied.This divergence could be caused by the marine regressions during the Pliocene glaciations (2.7-3.6 Mya), when the sea level fell several meters.During the glaciations, a barrier could be formed between Cape La Nao (SP7) and the Balearic Islands (SP5 and SP6), acting as a separation between the two basins and allowing diversification between the two clades.However, we cannot discard the existence of a barrier elsewhere (e.g. the Gibraltar Strait) and a later expansion, the boundaries being the results of secondary contacts.Today, the low larval and adult dispersal capabilities of Tripterygion species (Heymer, 1977;Wirtz, 1978;Sabatés et al., 2003;Carreras-Carbonell et al., 2006) and the circulation regime that separates the northern from the southern basins (Send et al., 1999) could be maintaining the distribution areas of the two species non-overlapping.

Key to the Mediterranean tripterygiids
Modified from Zander (1986).
FIG. 1. -Sampling localities for T. tripteronotus (•) and T. tartessicum (+).Dashed line shows the break zone between the two species along the Spanish Mediterranean coast.(*): Holotype locality.Localities which individuals were molecularly analysed are underlined.See Table1for locality abbreviations and further details.
FIG. 5. -Haplotype tree inferred from Bayesian Inference for T. tartessicum and T. tripteronotus species using all genes together.Only probabilities above 95% are shown; ( + and ++ ): different haplotypes found in the same locality; ( * ): the same haplotype found in different localities.See Table1for locality abbreviations and further details or Figure1for a quick geographical location.

TABLE 1 .
-Specimens of the two Tripterygion species collected at different localities of the Mediterranean and Atlantic adjacent waters.The number of individuals used for morphological (Nm) and molecular (Ng) analyses, for each locality, are detailed.

TABLE 1 (
cont.).-Specimens of the two Tripterygion species collected at different localities of the Mediterranean and Atlantic adjacent waters.The number of individuals used for morphological (Nm) and molecular (Ng) analyses, for each locality, are detailed.

TABLE 1 (
cont.).-Specimens of the two Tripterygion species collected at different localities of the Mediterranean and Atlantic adjacent waters.The number of individuals used for morphological (Nm) and molecular (Ng) analyses, for each locality, are detailed.
(0): no amplifications were done, (-): amplifications were done but they did not succeed, (n.a.): no available data.The holotype and paratypes are labelled; the catalogue number for each individual is shown.(IIPB): Instituto de Ciencias del Mar de Barcelona, (SMNS): Staatliches Museum fuer Naturkunde Stuttgart.The first two letters in the map code identify each country, the number identifies the locality and the lower case letter identifies different collection dates.

TABLE 2 .
-Polymorphism and divergence within and between species, for each gene separately and all genes together (mean ± SD percentage).