Introduction

The minor class of ribosomal DNA (rDNA) comprises the 5S rRNA gene family, which is arranged in higher eukaryotes in several thousands of copies of tandemly repeated units. Each unit consists of a highly conserved coding sequence of 120 base pairs (bp) encoding for the 5S rRNA, and a flanking region of variable length that is not transcribed (nontranscribed spacer [NTS]) and contains some regulatory elements for the transcription of the coding sequence (Sajdak et al. 1998; Wasko et al. 2001). Because it is nontranscribed, the NTS is neutral and it is expected to freely mutate. However, the 5S rDNA fits in a concerted evolution model (Drouin and Moniz de Sà 1995), which allows for the homogenization of the repeated sequences, thus decreasing intraindividual and intrapopulation heterogeneity (Dover 1982). For this reason, the NTS has been widely employed as a molecular marker for species identification and phylogenetic studies, although its application for this last purpose in closely related species is being currently discussed (Pasolini et al. 2006).

Many studies have been published dealing with the structure, chromosomal location, and sequence variation of the 5S rRNA genes in fungi (Kramer et al. 1978; Cihlar and Sypherd 1980; Tabata 1980; Cassidy and Pukkila 1987; Duchesne and Anderson 1990; Amici and Rollo 1991), plants (Ganal et al. 1988; Nedi et al. 2002), and animals (Brown et al. 1977; Bogenhagen et al. 1980; Bogenhagen and Brown 1981; Komiya et al. 1986), including freshwater and marine fishes (Pendás et al. 1994; Morán et al. 1996; Sajdak et al. 1998; Martins and Galetti 2001; Wasko et al. 2001; Sola et al. 2003; Robles et al. 2005; Pasolini et al. 2006; Gornung et al. 2007).

Two different types of 5S rDNA have been found in Xenopus laevis, one expressed in somatic cells and the other in oocytes, derived from the somatic type by gene duplication (Komiya et al. 1986). This dual system of paralogous 5S rRNA genes has been documented in other animal taxa, including many fish species (Sajdak et al. 1998; Martins and Galetti 2001; Wasko et al. 2001; Sola et al. 2003; Robles et al. 2005; Pasolini et al. 2006). The main difference between these two types of 5S rDNA relies on the length of the NTS, although in some cases nucleotide substitutions in the 120 bp-long coding sequence of the two 5S rDNA types have been reported (Pendás et al. 1994; Martins and Galetti 2001; Wasko et al. 2001; Pasolini et al. 2006; Gornung et al. 2007).

The genus Merluccius is included in the family Merlucciidae, which is considered the most basal group within the Gadoidei (Teletchea et al. 2006). The 12 species of the genus Merluccius are distributed in the Atlantic (European–African coasts: M. merluccius, M. senegalensis, M. polli, M. capensis, and M. paradoxus; American coasts: M. albidus, M. bilinearis, and M. hubbsi) and the Pacific (M. productus, M. angustimanus, M. gayi, and M. australis) oceans. Hake fisheries are a priority for many regions (Pitcher and Alheit 1995); thus, many works published about hake genetics are mainly focused on the population structure of Merluccius species (Lundy et al. 1999; Castillo et al. 2004; Cimmaruta et al. 2005; von der Heyden et al. 2007) for application in fisheries management. The phylogeny of the genus has been scarcely studied, based only on allozyme variation (Roldán et al. 1999; Grant and Leslie 2001) and mitochondrial loci (Quinteiro et al. 2000; Campo et al. 2007).

Nuclear ribosomal RNA genes have been studied in the genus Merluccius for application as species-specific markers to identify commercial seafood based on the length of polymerase chain reaction (PCR) products (Pérez and García-Vázquez 2004; Pérez et al. 2005). Simultaneous occurrence of ≥2 amplification products of different lengths for the 5S rDNA locus in species, such as Merluccius paradoxus, M. gayi, and M. bilinearis (Pérez and García-Vázquez 2004), suggests the existence of >1 locus at this gene in the genus. However, the type or origin of these duplicated loci has not been investigated until now.

Two main objectives were achieved in this study. First, a comparative analysis of the 5S rDNA organization among the 12 species of Merluccius was carried out to contribute to deciphering the pattern of evolution of this multigene family. Second, phylogenetic relations were inferred for all the species in the genus based on the 5S rDNA sequences here obtained and for some American hake species based on genetic distances estimated with 5 microsatellite loci. They were compared with the molecular phylogeny and speciation patterns previously proposed for Merluccius (Stepien and Rosenblatt 1996; Roldán et al. 1999; Quinteiro et al. 2000; Grant and Leslie 2001; Campo et al. 2007).

Material and Methods

Sampling, DNA Extraction, and PCR Amplification

The Merluccius samples analyzed in this work (2 to 10 for each of the 12 Merluccius species yielding a total of 51 individuals) belong to the collection already analyzed by Campo et al. (2007). Total DNA was extracted from muscle tissue according to Chelex resin protocol (Estoup et al. 1996).

Amplification of the NTS and partial-coding sequence of the 5S rDNA was done using the universal primers 5SA and 5SBR (Pendás et al. 1994). The total coding sequences were amplified according to the protocol described by Pérez and García-Vázquez (2004) by employing the primers 5S C (5′-AAGCTTACAGCACCTGGTATT-3′) and 5S MD (5′-TTCAACATGGGCTCCGACGGA-3′) described therein. PCR reactions were carried out in a total volume of 40 μl containing Promega buffer 1× (Promega, Madison, WI), 2.5 mM MgCl2, 250 μM each dNTP, 40 pmol each primer, 0.2 μl Promega GoTaq polymerase (Promega, Madison, WI), and 2-μl sample of DNA. PCR was performed in a GeneAmp PCR System 9700 (Applied Biosystems) with the following conditions: initial denaturing step at 95°C for 5 minutes, followed by 30 cycles of denaturing at 95°C for 30 seconds, annealing (for 30 seconds) at 65°C for both pair of primers, and extension at 72°C for 30 seconds, ending with a final extension at 72°C for 15 to minutes. In addition, for M. bilinearis, M. albidus, M. hubbsi, M. australis and M. gayi, 5 dinucleotide microsatellite loci were analyzed (in 24, 24, 25, 50, and 25 individuals, respectively): Maus7, Maus30, and Maus32 (Machado-Schiaffino and Garcia-Vazquez 2009); Mmer-UEAW01 (Rico et al. 1997), and Mmer-Hk20 (Morán et al. 1999). PCR amplifications were performed on reaction mixtures containing approximately 50 ng extracted hake DNA template, 10 mM Tris-HCl (pH 8.8), 2.5 mM MgCl2, 50 mM KCl, 0.1% Triton x-100, 0.35 μM fluorescently labelled primers, 0.5 U Promega Taq polymerase, and 250 μM each dNTP in a final volume of 20 μL.

DNA Purification and Sequencing

PCR products were loaded in 50-ml 2.5 % agarose gels and stained with 2 μl 10 mg/ml ethidium bromide. Bands corresponding to the 5S rDNA fragments amplified were removed from the gel, and DNA was purified using the Wizard SV Gel and PCR Clean-Up System and then sequenced. Automated fluorescence sequencing was performed with both primers in every case on an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems) with BigDye 3.1 Terminator system, in the Unit of Genetic Analysis of the University of Oviedo (Spain). For the microsatellite markers, size of the labelled PCR products were determined employing the same genetic analyzer, and the results were visualized employing GENESCAN V. 3.7 software (Applied Biosystems).

Phylogenetic Analyses

Sequences were edited with BioEdit (Hall 1999) and aligned with ClustalW (Thompson et al. 1994) with a penalty of 6 for gap opening and 4 for gap extension. However, alignments had to be edited manually a posteriori because due to the enormous differences in length, some regions were not properly aligned by the program. To perform phylogenetic analyses, gaps were coded according to the methods proposed by Simmons and Ochoterna (2000) (i.e., simple and complex indel coding methods) as implemented in SeqState (Müller 2005). We constructed two phylogenetic trees: (1) one with a mixed data set composed of the sequence alignment (without gaps) plus the codification of the gaps according to the simple indel coding method and (2) another one using only the gaps coded with the complex indel coding procedure. The phylogenetic analysis of the first data set was done in MrBayes 3.1.2 (Huelsenbeck and Ronquist 2001) with default settings to establish the initial heating values for four Markov chains, which ran simultaneously and were sampled every 100 cycles. MrModelTest software version 2.2 (Nylander 2004) was employed to determine the model of sequence evolution that best fitted the DNA data (according to Akaike criterion), and this information was implemented in the Bayesian analysis. In contrast, for the data set containing only gap information, maximum parsimony (MP) analysis was done with the program PAUP (ver. 4.0b10; Swofford 2003) using an heuristic search with 10 random-addition sequence replicates and the Tree-Bisection-Reconnection (TBR) algorithm for branch-swapping. The statistical robustness of MP tree nodes was tested with 100 bootstrap replicates (Felsenstein 1985).

We also constructed a neighbor-joining tree with a distance matrix calculated from frequency data for the five microsatellite loci previously mentioned in computer package PHYLIP (Felsenstein 1989). Statistical support of nodes was calculated in this case with 1000 bootstrap replicates. Finally, the program FigTree 1.1.2 (Rambaut 2008) was employed to visualize the trees.

Results

Molecular Organization of 5S rDNA in Merluccius species

The electrophoretic banding pattern of the fragments amplified with the primers 5SA and 5SBR (Pendás et al. 1994) was very heterogeneous among the 12 species of Merluccius (Table 1). Some species exhibited only 1 band (M. angustimanus, M. australis, M. capensis, M. gayi, M. merluccius, M. productus, and M. senegalensis), whereas others yielded 2 (M. albidus, M. hubbsi, M. paradoxus, and M. polli) or 3 bands (M. bilinearis). After sequencing the fragments obtained with the 2 pairs of primers A-BR and C-MD for the 51 individuals analyzed, a consensus sequence of the 5S rDNA repeat unit (coding sequence plus NTS) was obtained for each band. These sequences were deposited in GenBank under the accession numbers FJ196623 to FJ196640 (Table 1).

Table 1 Size (bp), type, and GenBank accesion number for each of the 5S rDNA repeat units (coding sequence plus nontranscribed spacer) found for all Merluccius species
Full size table

As reported by Campo et al. (2007), only two types of coding sequences were found (Fig. 1), differentiated by two nucleotide substitutions at positions 3 and 25. One of the sequences (sequence A) was obtained for M. merluccius, M. senegalensis, and M. capensis. The other (sequence B) was obtained for all the bands of the rest of hake species. Neither heterozygotes nor intraspecific variation were found. All of the internal control regions (ICRs) were identified in the coding sequence of all species (box A, internal element, and box C in Fig. 1). No nucleotide variation was found within these ICRs.

Fig. 1
figure 1

Alignment of the two haplotypes of the 5S rDNA coding region found in all of the individuals sequenced from the Merluccius species. In bold letters (positions 3 and 25) are shown the two nucleotide substitutions found. The internal control regions (box A = position 50 to 64; internal element = positions 67 to 72; box C = nucleotides 80 to 97) are shaded

Full size image

All 5S rDNA sequences contained the TATA box control element within the NTS at position −30 bp (base pairs) upstream from the next array (Sajdak et al. 1998; Wasko et al. 2001). In M. merluccius, M. senegalensis, and M. capensis it has been modified to AATA. All of the sequences analyzed exhibited an additional TATA-like region at exactly 16 residues upstream from the TATA box. They also presented the 5 thymidine residues required for transcription termination (Bogenhagen and Brown 1981) at positions 119 to 123 and a second T-cluster 2 bases downstream from the primary one.

Comparative Analyses of the NTS Sequences

In M. hubbsi, M. polli, and M. paradoxus, the NTS of the longer band contained of a simple sequence repeat (SSR), which also appeared in the single band of M. productus, M. angustimanus, M. gayi, and M. australis; in the two bands of M. albidus; and in the two longer bands of M. bilinearis. The SSR consisted of a variable number of repeats of the CA motif (between 4 and 14) preceded by a variable number of Cs (between 1 and 13). Its location was always similar, starting at position 118, 119, or 121 of the NTS.

Based on the absence or presence of this microsatellite sequence within the NTS, we classified the 5S rDNAs of hakes in type-A (absent) and type-M (microsatellite present). In addition, two variants of different length because of insertions and/or deletions were found within type-M sequences of M. albidus and M. bilinearis. These two variants were called “S” and “L” (short and long, respectively). In contrast, some intraspecific variation at the size of the NTS type-M caused by the number of Cs and CA motif repeats was found for M. productus, M. bilinearis, M. albidus, and M. hubbsi but not for M. polli and M. paradoxus. However, intraindividual variation cannot be ruled out for the two latter species because the methodology employed in this study (direct sequencing of PCR products without cloning them) does not allow unambiguously identification of minor differences in the chromatogram. Because there is some variation, only the most frequent size found for each type-M band is listed in Table 1.

After introducing long gaps for solving the alignment, four groups of sequences were inferred based on nucleotide similarity and position of gaps. Group A comprised all type-A sequences (M. merluccius-A, M. senegalensis-A, M. capensis-A, M. paradoxus-A, and M. polli-A), except the sequences of M. bilinearis-A and M. hubbsi-A, which were 223 bp– and 241 bp–long, respectively. These two sequences could be considered a group apart based on their short length. Group M-I was composed of type M sequences of M. productus, M. angustimanus, M. gayi, M. australis, M. bilinearis-M-L, and M. albidus-M-S. Group M-II comprised the type-M sequences of M. polli and M. paradoxus. Finally, group M-III included the sequences M. bilinearis-M-S, M. albidus-M-L, and M. hubbsi-M.

Two conserved “blocks” were identified in the NTS alignment as evidenced by high similarity between all sequences (Fig. 2). The first block (block-1) corresponded to the 1 to 66 nucleotides within the NTS, and the second block (block-2) comprised the last 105 residues of each sequence, with the exceptions being the short M. hubbsi and M. bilinearis type-A sequences, which only matched partially to these blocks because of long deletions. In addition, alignment regions with nucleotide homology between ≥ 2 sequences (i.e., with no gaps) showed little variation. The three groups of sequences carrying microsatellites (M-I, M-II, and M-III) were clearly different from each other, although group M-II exhibited more fragments of the alignment in common with group M-III than with group M-I.

Fig. 2
figure 2

Schematic representation of the alignment of the eighteen type-A and type-M sequences found for all the species within the genus Merluccius. Dark grey represents fragments with nucleotide alignment, whereas light grey indicate gaps. The numbers below the diagram point: 1 = first conserved block (positions 1 to 66); 2 = zone with the SSR; 3 = zone of nucleotide homology between type-A and type-M sequences; and 4 = second conserved block (positions 570 to 675)

Full size image

Phylogenetic Relations Among NTS Sequences

The model of evolution obtained from MrModelTest (Nylander 2004) was the Hasegawa–Kishino–Yano 85 (Hasegawa et al. 1985), with a proportion of invariable sites of 0.6076 and equal rate of substitution for all sites. We did an analysis with all NTS sequences but later decided to remove M. hubbsi and M. bilinearis type-A sequences in the final phylogenetic reconstruction because due to their much shorter length they introduced noise into the phylogenetic inference. The final phylogenetic tree is shown in Fig. 3a. Almost identical topologies were recovered in Bayesian (DNA + ”simple indel” coded gaps) and MP (“complex indel” coded gaps) analyses, with only two minor differences between them (see later text). Three main clades can be depicted from the tree, and the sequences belonging to the same group inferred from the alignment (A, M-I, M-II, and M-III) clustered together in all cases. From up to down, the first clade consisted of all the M-I sequences divided into two subgroups: (1) M. productus-M + M. angustimanus-M + M. gayi-M + M. australis-M] and (2) M. bilinearis-M-L + M. albidus-M-S. For this clade, the Bayesian tree exhibited a multifurcated pattern involving the first subgroup taxa, whereas in the MP tree they were clustered in a separate branch with 61% of bootstrap support (tree not shown). The next branch clustered M-II and M-III sequences. The Bayesian tree placed M. bilinearis-M-S as a sister taxon of group M-II (M. paradoxus-M + M. polli-M), whereas the MP tree split both groups in separate branches. Finally, all type-A sequences (group A) clustered in a well-separated branch with two subclades: one comprising M. paradoxus-A + M. polli-A and the other comprising M. merluccius-A + M. senegalensis-A + M. capensis-A.

Fig. 3
figure 3

a Tree topology obtained in the bayesian phylogenetic analyses of the type-A and type-M sequences of the 12 Merluccius species plus the alignment gaps coded according to the simple indel coding method of Simmons and Ochoterna (2000). Values above the branches indicate posterior probability support, whereas numbers below the branches indicate bootstrap support after 100 replicates for MP analyses of only the alignment gaps coded according to the complex indel coding method of Simmons and Ochoterna (2000) as implemented in SeqState (Müller 2005). Type-A sequences of M. bilinearis and M. hubbsi were excluded in both analyses (see reasons given in the text). b Neighbor-joining tree estimated from frequency data of 5 microsatellite loci for 5 American hake species. c ML tree topology for Merluccius species based on the combination of 4r mitochondrial partial gene sequences (12S, 16S, control region, and cytochrome b) as adapted from Campo et al. (2007). Outgroup = Gadus morhua

Full size image

The four groups of sequences seem to constitute well-differentiated evolutionary clades. This was supported by high values of bootstrap and posterior probability for almost all nodes supporting these groups. Figure 3b shows the midpoint-rooted neighbor-joining tree obtained from frequency data of five microsatellite loci (Maus7, Maus30, Maus32, Mmer-UEAW01, and Mmer-Hk20) for five American hake species. M. bilinearis was separated as the most divergent taxon, whereas M. australis and M.gayi were clustered as sister species in the most derived branch.

Discussion

Molecular Organization and Evolution of 5S rDNA in Merluccius species

All sequences analyzed here likely correspond to functional genes because they exhibit all the necessary features for the correct gene expression: the three ICRs (box A, internal element, and box C in Fig. 1), the TATA box, and the poly T region. The second T cluster that was found two bases downstream from the primary one could be a “backup” cluster, a feature already described for Xenopus 5S RNA genes (Bogenhagen and Brown 1981). This has also been reported in other fishes (Gornung et al. 2007). Similarly, the second TATA-like region, found 16 residues upstream from the TATA box, could be a “backup” TATA box.

For the genus Merluccius, we found at least 2 types of 5S rDNA, of different length, in 5 species (M. bilinearis, M. albidus, M. hubbsi, M. paradoxus, and M. polli) of 12. The existence of 2 classes of 5S rDNA differing mainly in the size of the NTS, and sometimes also in the nucleotide sequence of the coding region, has been described for many animal species, including fish (Komiya et al. 1986; Pendás et al. 1994; Martins and Galetti 2001; Wasko et al. 2001; Sola et al. 2003; Pasolini et al. 2006; Gornung et al. 2007). In addition, the existence of conserved blocks within NTS sequences and the low number of nucleotide substitutions found in the homologous sequence alignment regions indicate that the differences between the NTS sequences found in Merluccius species are mainly caused by insertions and deletions (more than nucleotide substitutions), such as in Characiformes (Wasko et al. 2001) and other taxa as separate as sturgeons (Robles et al. 2005). However, the organization of Merluccius 5S rRNA genes may be somewhat different from that of other fish taxa. First, 4 types of NTS sequences (groups A, M-I, M-II, and M-III), instead of 2, could be considered for this genus, clustering the type-M sequences into 3 well-separated groups in the reconstructed phylogenetic tree. Second, differences in the coding sequence between the 2 types (long type-II and short type-I NTS) of 5S rDNA, which have been reported for other fish (Komiya et al. 1986; Pendás et al. 1994; Martins and Galetti 2001; Wasko et al. 2001; Sola et al. 2003; Pasolini et al. 2006; Gornung et al. 2007), did not occur in Merluccius, where nucleotide substitutions in the coding region were found only for the clades M. merluccius, M. senegalensis, and M. capensis, the 3 most recently diverged species within the genus (Campo et al. 2007).

Distinct families of 5S rRNA genes, often characterized by variants of spacers, have been described associated with differential expression in somatic and oocyte cells (Komiya et al. 1986; Martins and Galetti 2001; Wasko et al. 2001; Pasolini et al. 2006). Such kinds of tissue specialization can not be generalized for the genus Merluccius because only one type of NTS (type-A or type-M) exists for seven species, such as in European M. merluccius (type-A), Pacific M. australis (type-M), and others.

Pasolini et al. (2006) suggested that the dual 5S rRNA gene system corresponds to the ancestral condition of the Piscine genome and that the loss of a 5S rRNA gene cluster might have occurred secondarily in fish taxa that bear only one type of 5S rDNA. M. bilinearis, supposed to be the most ancient species of the genus (e.g., Quinteiro et al. 2000; Campo et al. 2007) exhibited the two types of 5S rDNA. In the most recent M. merluccius–M. senegalensis–M. capensis lineage, the loss of type-M 5S rDNA in the ancestral species could have led to the current presence of only type-A 5S rDNA. Deletion or loss of the type-A locus in all species within the Pacific Ocean lineage (M. productus, M. angustimanus, and M. gayi) explains their 5S rDNA organization. However, the evolution of this gene family in the remaining clade can not be explained by simple loss of one type of 5S rDNA. The north Atlantic American M. albidus and M. bilinearis exhibit two different types of NTS containing microsatellites; M. bilinearis possesses one additional NTS without SSR, which absent in M. albidus. This could be explained by a duplication of type-M locus in the M. albidus–M. bilinearis lineage plus a loss of type-A in M. albidus. Additional deletions in M. bilinearis type-A could explain its short feature. A complex combination of duplications, insertions, and deletions, in general genome rearrangements, has likely been involved in the evolution of this gene family in the genus Merluccius.

With respect to M. hubbsi-A and M. bilinearis-A, in addition to not having a microsatellite, they do not present any of the other features shared by the rest of type-A sequences, being just short sequences that match only the common blocks of the general alignment. They may have been originated through deletions from a longer sequence, but whether this ancestral state was type-A or type-M cannot be determined using the present data.

Phylogenetic Inference

According to the phylogenetic trees constructed from NTS sequences and gaps alignment (Fig. 3), the species of the genus Merluccius generally fit a vicariant model of distribution (sequences of Pacific species cluster in the same branches as do Atlantic American and Atlantic Euro-African species), with a couple of exceptions. Type-M sequences of two African species (M. paradoxus and M. polli) cluster within the clade formed by three North American Atlantic hakes (M. bilinearis, M. albidus, and M. hubbsi). In contrast, one of the two type-M sequences of the Northwestern Atlantic species, M. albidus and M. bilinearis, MS and ML, respectively, cluster together as a sister clade of the group formed by the four Pacific hakes (M. productus, M. angustimanus, M. gayi, and M. australis).

When comparing the phylogenetic relations here obtained from the analysis of NTS sequences of the 12 Merluccius species with the phylogeny constructed from mitochondrial genes (Campo et al. 2007), the main scheme is maintained (Fig. 3c). Species cluster together by geographic proximity, with M. bilinearis being likely the most ancient species. However, there is an important difference between the two trees. The South American hake M. australis, clustered as a sister species of the Argentine M. hubbsi from mitochondrial genes (in geographic concordance; their distributions overlap in Southwest Atlantic waters), was grouped with the Pacific hakes M. productus, M. angustimanus, and M. gayi in NTS sequences analysis. In addition, the phylogenetic tree constructed from five microsatellite frequency data in five American hakes (Fig. 3b) also supports this pattern, placing M. australis as the sister species of the Pacific M. gayi. Moreover, phylogenetic relations among Merluccius species inferred from other nuclear markers, such as allozyme loci (Stepien and Rosenblatt 1996; Roldán et al. 1999; Grant and Leslie 2001) also place M. australis more related to the Pacific lineage than to the Atlantic one.

Although not uncommon, it is not easy to explain large incongruence of nuclear and mitochondrial phylogenies. In other cases—from fish (e.g., Egger et al. 2007; Koblmüller et al. 2007) to lizards (e.g., Leaché and McGuire 2006) to mammals (e.g., Ting et al. 2008)—it has been interpreted as a signal of repeated hybridization and introgression, leading to the hybrid origin of some species and/or reticulate phylogeny. In the present case, phylogenetic incongruence between nuclear and mitochondrial markers for only one species could be explained by ancient asymmetric hybridization leading to the formation of a species of ancestral hybrid origin, M. australis. This process would have involved two ancestral populations, one from the M. productus–M. angustimanus–M. gayi branch (Pacific Ocean) and other from M. albidus–M. hubbsi lineage (Atlantic Ocean). If male individuals from the former population successfully reproduced at first with female individuals from the latter, and then also with the new hybrid female individuals with higher fitness than male individuals coming from the latter population and from the new hybrid pool, after a considerable number of generations the nuclear genome of the M. albidus–M. hubbsi lineage could have been lost in the new hybrid population pool, being replaced by that of the M. productus–M. angustimanus–M. gayi ancestor, whereas they would have kept the mitochondrial lineage of the latter. Then these two ancestral populations could have split and migrated to the North Pacific and the South Atlantic oceans, respectively (Campo et al. 2007). Therefore, the analysis of the nuclear sequences would place M. australis together with M. productus–M. angustimanus–M. gayi lineage, whereas mitochondrial DNA phylogeny would cluster it closely related to the Atlantic species M. albidus and M. hubbsi.

Thus, we propose hybridization as a third mechanism of speciation (together with vicariance and geographic dispersion; Campo et al. 2007) to explain the evolutionary history of the genus Merluccius. Further work, such as extensive genome and karyotype analysis, should be done to confirm this hypothesis because hybridization can result in genomic changes, including alterations of gene expression, chromosomal structure, and genome size (Baack and Rieseberg 2007).