Molecular validation of the specific status of Parablennius sanguinolentus and Parablennius parvicornis ( Pisces : Blenniidae ) *

There has been considerable controversy over the taxonomic status of the blenniids Parablennius sanguinolentus (Pallas, 1811) and Parablennius parvicornis (Valenciennes, 1836). Several authors have used different criteria to classify P. sanguinolentus and P. parvicornis as two different species (see Bath, 1977; Bath, 1990; Santos et al., 1997), or as two sub-species of P. sanguinolentus (see Arruda, 1979; Zander, 1980; Almeida and Harmelin-Vivien, 1983). The only notable characters that distinguish these taxa from each other are: an additional dorsal spine in P. sanguinolentus, small teeth anteriorly to the canines in the upper jaw in P. parvicornis, which

The only notable characters that distinguish these taxa from each other are: an additional dorsal spine in P. sanguinolentus, small teeth anteriorly to the canines in the upper jaw in P. parvicornis, which are absent in P. sanguinolentus (Bauchot, 1966), and slight differences in pigmentation (Zander, 1979).
According to Zander (1986), Bath (1990), Oliveira et al. (1992), Gonçalves et al. (1993), andSantos et al. (1997), the distribution area of P. sanguinolentus includes the Mediterranean and the Atlantic coast between France (Gulf of Biscay) and Morocco, although Bath (1990), did not list P. sanguinolentus as inhabiting African shorelines.The distribution of P. parvicornis includes the West African coast, from Morocco or Mauritania to the Congo River including the NE Atlantic archipelagos of Azores, Madeira and the Canaries and Cape Verde islands.
In this paper we discuss the taxonomic status of P. sanguinolentus and P. parvicornis using mitochondrial rDNA data.

MATERIALS AND METHODS
The species and outgroup taxa included in the phylogenetic analysis, their sample localities, number of specimens and corresponding Genbank accession numbers are listed in Table 1.
Samples were collected underwater and in tide pools and fixed in 96% ethanol.Total genomic DNA was extracted either from muscle tissue or from finrays using a proteinase K/SDS based extraction buffer and phenol/chloroform with ethanol precipitation (Maniatis et al., 1982).
For both fragments, PCR mixtures were prepared with a total volume of 20 µl, with: 1.5 mM MgCl 2 , 200 mM of each dNTP, 0.5 mM of each primer, 0.5 units of Taq polymerase (Gibco BRL, Life Technologies Inc., Gaithersburg, MD), 1x buffer supplied by the manufacturer and approximately 20 ng of genomic DNA.The amplification process was performed in a Biorad Gene-Cycler TM as follows: 4 minutes at 94ºC and 30 cycles of 1 minute at 94ºC, 1 minute at 55ºC and 1 minute at 72ºC.After this sequence, these products where kept at 72ºC for 10 minutes.PCR products were purified with a GFX 520 F. ALMADA et al.PCR DNA purification kit (Amersham-Pharmacia), following the manufacturer's recommendations.Automatic sequencing of PCR purified products was performed with a CEQ 2000 XL, Beckman Coulter, with the same primers.Alignments were made using Clustal X 1.81 (Thompson et al., 1997), with default settings.Character congruence between the two fragments was tested using the incongruence-length difference test (ILD) (Farris et al., 1995), available in PAUP 4.0b10 Win (Swofford, 2002).The null hypothesis of congruence between the two data sets was not rejected (P=1).Therefore, we analysed the 12S and 16S rDNA sequences combined in one single fragment (but see Dolphin et al., 2000).Regions where the alignment was ambiguous were removed from the analysis, which resulted in a fragment with 847 bp.The degree of saturation was assessed by plotting transitions and transversions against uncorrected p-distances.
Sequences were analysed with three methods of phylogenetic inference: maximum-parsimony (MP), maximum-likelihood (ML) and minimum evolution (neighbour-joining -NJ) (Saitou and Nei, 1987).The phylogenetic analysis was performed with PAUP 4.0b10 Win (Swofford, 2002).Bootstrapping (Felsenstein, 1985), was used to assess robustness of the nodes in the trees with 1000 replicates for MP and NJ and 100 replicates for ML.The heuristic search option "random addition of taxa" and tree bisection reconnection (TBR), with the MULPARS option in effect, was used with the three inference methods.MP analysis was conducted with the ACC-TRAN option.
In order to choose the model of molecular evolution that best fitted our data we used the program Modeltest 3.06 (Posada and Crandall, 1998).For the combined 12S-16S rDNA dataset the ML settings selected, according to the results of the Modeltest, corresponded to the general time reversible model (GTR+G) with rate heterogeneity.The distribution of rates at the variable sites was assumed to follow a gamma distribution with a shape parameter equal to 0.2668.NJ was based on the distance estimator derived from the ML settings selected for the combined fragment.

RESULTS AND DISCUSSION
We analysed a total of 367 bp of the mitochondrial 12S rDNA and 480 bp of the mitochondrial 16S rDNA, which makes a combined sequence of 847 bp.Of these, 267 sites were variable and 137 sites were parsimony informative.The TS/TV ratio was 1.53.The base frequencies were: A=0,2950; C=0,2501; G=0,2289 and T=0,2260.There was no evidence of saturation either for transitions or transversions.
The three methods of phylogenetic inference converged into the same topology represented in Figure 1.
All samples of P. sanguinolentus and P. parvicornis formed a well supported monophyletic group that was well differentiated from the remaining species of Parablennius.Within this group, all the samples of P. sanguinolentus formed a well supported clade as did those of P. parvicornis.It is interesting to note that all samples of P. sanguinolentus from Mainland Portugal to Lebanon are represented by the same haplotype.The genetic distance (p-distance) between P. sanguinolentus and P. parvicornis was 2.43% (SD = 0.07; Min = 2.40; Max = 2.64).The p-distance in relation to the variation within the P. parvicornis clade was 0.17% (SD = 0.14; Min = 0; Max = 0.60).The highest values were found between fish from Cape Verde and some Azorean samples, which are the more peripheral and distant localities within our study area.
One of the authors (A.Brito) conducted three expeditions along the African coast (in 1986, 1988 and 1998), in order to examine the distribution of P. sanguinolentus and P. parvicornis by sampling the intertidal fish fauna.P. sanguinolentus was not found south of the Casablanca area (34ºN), a conclusion supported by Brownell (1978).P. parvicornis was never collected north of Cape Blanc (21º N), a finding also supported by Bath and Wirtz (1992).South of this point, on the coast of Senegal, it is a common fish.These results show that P. sanguinolentus and P. parvicornis are separated by a gap of at least 13º of latitude.
From a biogeographical point of view it is interesting to note that the northern limit of P. parvicornis is much more to the south on the West African coast than it is near the Atlantic islands.Indeed, it reaches its northern limit at the Azores well to the north of the entrance to the Mediterranean.This distributional pattern compares well with the results of previous studies (e.g.Santos et al., 1995 and the references therein), that have demonstrated that there are significant affinities between the ichthyofauna of the Azores and those of Madeira and the Canaries.These authors discussed the possible oceanographic processes fish of tropical African origin may use to reach the Azores via the Canaries and Madeira.
Further genetic studies using larger samples for each geographical location and rapidly evolving markers, such as the control region of the mitochondrial DNA, could help to test this biogeographical hypothesis.

TABLE 1 .
-List of Blenniidae species and outgroup taxa included in the phylogenetic analysis, sample localities, number of specimens and corresponding Genbank accession numbers.