Mediterranean Sea. It is a marine biodiversity hotspot with 713 fish species inventoried (FishBase; www.fishbase.org). Samples were obtained from fish markets in the Languedoc-Roussillon region (Gulf of Lion, France), in the north-western Mediterranean coast.

Cantabric Sea. Less diverse than the Mediterranean Sea, it contains 148 fish species inventoried. Catch from commercial fisheries was sampled from fish markets in Asturias (North of Spain).

Amazon River. It is the main freshwater biodiversity hotspot of the world (1218 inventoried fish species). We have sampled catches obtained in different fish markets of Manaus (Brazil). This is the area where the two main Amazonian drainages (the rivers Negro and Solimões) join.

Narcea River (North of Spain). As other North Iberian rivers, it exhibits reduced biodiversity with only 17 fish species inventoried. Fisheries are strongly targeted and focused on sport angling of salmonids. Samples were obtained in situ from fishermen in the lower reach of the river.

The two most exploited species (those that yield more tonnes in official catch statistics) from each site were chosen for this study. They were: mackerel Scomber scombrus (Goode, 1884) and anchovy Engraulis encrasicolus (Linnaeus, 1758) from the Mediterranean Sea; mackerel and albacore tuna Thunnus alalunga (Bonnaterre, 1778) from the Cantabric Sea; Curimatá Prochilodus nigricans (Spix & Agassiz, 1829) and jaraquí Semaprochilodus insignis (Jardine & Schomburgk, 1841) from the Amazon River; Atlantic salmon Salmo salar (Linnaeus, 1758) and brown trout Salmo trutta (Linnaeus, 1758) from the Narcea River. These species do not exhibit population sub-division in the fishing areas considered. The West Mediterranean and the Eastern Atlantic Ocean populations of mackerel seem to form a panmictic unit (Zardoya et al. 2004). The highly migratory albacore tuna exhibits only inter-oceanic population differentiation or between the Atlantic and the Mediterranean, not within the same ocean (Chow and Ushiama 1995, Viñas et al. 2004). For anchovy, the whole north-western Mediterranean likely harbors a single population (Tudela et al. 1999). Curimatá and jaraquí, the main catch in the Brazilian Amazon state, have a shallow genetic structuring in the Amazon basin and can be considered homogeneous populations around Manaus (Ardura et al. 2013). Finally, Atlantic salmon and brown trout populations are not subdivided within rivers in North Spain unless there is strong habitat fragmentation (e.g. Horreo et al. 2011a, b), yet this is not the case for the lower accessible zone of River Narcea.

Ten samples were analyzed per species.

DNA extraction was automatized with QIAxtractor robot following the manufacturer’s protocol (QIAGEN DX Universal DNA Extraction Tissue Sample CorProtocol), which yields high quality DNA suitable for a wide variety of downstream applications. The procedure is divided into two sections: digestion and extraction. The digestion process favors tissue dissociation and liquid suspension, and is ready for extraction.

Briefly, a 96 well round well lysis block (Sample Block) is loaded with 420 µl DX Tissue Digest (containing 1% v/v DX Digest Enzyme) manually or using the Tissue Digest Preload run file. Once the DX Tissue Digest is loaded with the sample, the sample block is sealed and incubated at 55 °C with agitation for at least 3 h. 220 µl of supernatant is transferred from the sample block in position C1 to the lysis plate in position B1. 440 µl of DX Binding with DX Binding Additive is added to the lysis plate. The lysate is then mixed 8 × and incubated at room temperature for 5 min. 600 µl of the lysate is added into the capture plate (Pre-mixed 8 ×). A vacuum of 35 kPa is applied for 5 min. 200 µl of DX Binding with DX Binding Additive is loaded into the capture plate. A vacuum of 35 kPa is applied for 5 min. 600 µl of DX Wash is loaded into the capture plate. A vacuum of 25 kPa applied for 1 min, repeated (2 iterations). 600 µl of DX Final Wash is loaded into the capture plate. A vacuum of 35 kPa is applied for 1 minute. A vacuum of 25 kPa is applied for 5 min to dry the plate. The carriage is moved to elution chamber. 200 µl of Elution buffer is loaded into the capture plate. The sample is then incubated for 5 min. A vacuum of 35 kPa is applied for 1 min.

We employed the QIAxtractor Software application. The tube was frozen at -20 °C for long-time preservation.

Fragments of four different mitochondrial genes were amplified by polymerase chain reaction (PCR): 12S rDNA, COI, cyt b and D-Loop (Table 1). We employed primers commonly used for fish published by Palumbi (1996), Ward et al. (2005), Kocher et al. (1989) and Lee et al. (1995) respectively. Amplification reactions were performed in a total volume of 23 µl, including 5 PRIME Buffer 1 × (Gaithersburg, MD, USA), 1.5 mM MgCl₂, 0.25 mM dNTPs, 1 µM of each primer, 20 ng of template DNA, and 1.5U of DNA Taq polymerase (5 PRIME).

The PCR conditions were the following:

12S rDNA: an initial denaturation at 95 °C for 10 min, then 35 cycles of denaturation at 94 °C for 1 min, annealing at 57 °C for 1 min and extension at 72 °C for 1.5 min, followed by a final extension at 72 °C for 7 min.

COI: an initial denaturation at 94 °C for 5 min, then 10 cycles of denaturation at 94 °C for 1 min, annealing at 64–54 °C for 1 min and extension at 72 °C for 1.5 min, followed by 25 cycles of denaturation at 94 °C for 1 min, annealing at 54 °C for 1 min and extension at 72 °C for 1.5 min, finally a final extension at 72 °C for 5 min.

cyt b: an initial denaturation at 94 °C for 5 min, then 10 cycles of denaturation at 94 °C for 1 min, annealing at 60–50 °C for 1 min and extension at 72 °C for 1.5 min, followed by 25 cycles of denaturation at 94 °C for 1 min, annealing at 54 °C for 1 min and extension at 72 °C for 1.5 min, finally a final extension at 72 °C for 5 min.

D-Loop: an initial denaturation at 94 °C for 5 min, then 10 cycles of denaturation at 94 °C for 1 min, annealing at 57 °C for 1 min and extension at 72 °C for 1.5 min, followed by 25 cycles of denaturation at 94 °C for 1 min, annealing at 54 °C for 1 min and extension at 72 °C for 1.5 min, finally a final extension at 72 °C for 5 min.

Sequencing was carried out by the DNA sequencing service GATC Biotech (Germany).

Sequences were visualized and edited employing the BioEdit Sequence Alignment Editor software (Hall 1999). Sequences were aligned with the MEGA v4.0 software (Tamura et al. 2007).

Putative proteins (amino acid sequences) from the COI and cyt b sequences were inferred with the software MEGA v4.0 (Tamura et al. 2007).

The sequences obtained were compared with those existing in the public database GenBank using the BLAST tool (http://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=blastn&BLAST_PROGRAMS=megaBlast&PAGE_TYPE=BlastSearch). Species were identified based on maximum BLAST scores with matching sequences, corresponding to 100% coverage and 100% identity. When the haplotype was new (i.e. not present in GenBank and BOLD), a 100% coverage with 99% identity, or in a few cases 98% identity, was found for the matching sequence. COI barcodes were also compared against the BOLD database, uploading them in the BOLD identification system in FASTA format at http://www.boldsystems.org/index.php/IDS_OpenIdEngine. The system retrieves matching sequences with the corresponding % similarity (matching nucleotides) and gives the most likely species for the query sequence. If matching sequences from more than one species are retrieved with a similar probability, then the system displays all the possible putative species the query can be assigned to.

The two databases were accessed for species identification in September 2013.

Three well-known diversity indices were employed: number of haplotypes, haplotype diversity and nucleotide diversity. They were calculated with the DnaSP software (Librado and Rozas 2009). The same program was employed to generate concatenated data files with the different markers analyzed and re-estimate genetic diversity parameters.

Haplotype diversity is a measure of population variation, as the probability of two randomly chosen haplotypes in the sample being different. It is calculated with the formula described by Nei and Tajima (1981).

Nucleotide diversity indicates how different sequences are to each other. Its value is higher when sequences belong to distant taxa. It is defined as the average number of nucleotide differences per site between any two DNA sequences chosen randomly from the sample population, and is symbolised as π (Nei and Li 1979).

We have also used the simplest diversity measure Nh/n (number of haplotypes divided by the number of samples analysed).

Comparison between genes for their polymorphic content was made based on means and variances of diversity parameters. It was performed using the software SPSS 13.0 software (SPSS Inc., Chicago, IL, USA).

For three study areas, the two most harvested species belonged to the same family (Table 1), viz. Curimatidae, Salmonidae and Scombridae in the Amazon River, Narcea River and Cantabric Sea, respectively. In the Mediterranean Sea, the two most harvested species were respectively anchovy Engraulis encrasicolus (Engraulidae) and mackerel Scomber scombrus (Scombridae).

PCR yielded positive amplifications in all cases, and sequences of different length were obtained for each marker and species analyzed: 380–439, 605–635, 293–322, 412–462 base pairs (bp) for 12S rDNA, COI, cyt b and D-Loop respectively (Table 1). The concatenated sequences were thus 1692–1856 bp long. The sequences obtained were submitted to the GenBank where they are available with the accession numbers reported in Table 1.

Clear and unambiguous species identification from significant matches with the databases was not always possible (Table 2). All the 12S rDNA sequences yielded a 100% identity score with at least one GenBank reference sequence (other than those generated in the present study) belonging to only one species, and were hence considered as being unambiguously identified. However, the results were less clear for the other genes and also varied among species. All mackerel samples were well-identified by the four genes and the two databases, whereas tuna retrieved more than one species with identical scores or match probabilities (Thunnus alalunga, Thunnus thynnus and Thunnus orientalis) for all cyt b and many COI and D-Loop sequences (Table 3). One D-Loop sequence retrieved Thunnus albacares as the closest match (Table 3). Ambiguous results (more than one putative species) were obtained from BOLD also for anchovy (COI sequences assigned to any of Engraulis encrasicolus, Engraulis eurystole, Engraulis australis and Engraulis japonicus species), brown trout (assigned indistinctly to Salmo trutta and Salmo ohridanus by BOLD), curimatá (Prochilodus nigricans, Prochilodus rubrotaeniatus, Prochilodus lineatus, Prochilodus costatus) and jaraquí (Semiprochilodus insignis, Semiprochilodus taeniurus, Curimata inornata). In GenBank ambiguous COI species identifications occurred for five tuna haplotypes that yielded identical and maximum matching scores with Thunnus alalunga and Thunnus orientalis sequences, and for jaraquí (Semaprochilodus insignis and Semaprochilodus taeniurus sequences yielded identical and maximum matching scores with our haplotypes). For cyt b of jaraquí (Table 3) the problem was not ambiguity but lack of external reference sequences in GenBank, viz. all the sequences yielding > 91% matching scores with ours were from the present study, and the closest identity with an external sequence (91%, unlikely the same species for a conserved coding gene) occurred with the sequence AY791437 of Prochilodus nigricans.

As expected, the four DNA regions exhibited different degrees of variability (Table 4). The non-coding D-Loop (58 haplotypes in total) was more variable than the two protein coding loci (31 and 27 haplotypes for cyt b and COI respectively) and the ribosomal 12S rDNA gene (15 haplotypes). The four marine species, the Amazonian jaraquí (Semiprochilodus insignis) and the north Spanish brown trout (Salmo trutta) exhibited ten different haplotypes in total considering the concatenated mitochondrial sequences analyzed. Fewer haplotypes were obtained for the Amazonian Prochilodus nigricans (6 haplotypes) and the Spanish Salmo salar (two haplotypes). In this latter species polymorphism occurred in the 12S rDNA gene, but not in the D-Loop, which was the most variable region in the other species. Overall nucleotide diversity was higher for marine than for freshwater settings for all markers as well as the concatenated sequence (Table 4). The highest Hd for both 12S rDNA and COI genes corresponded to the Amazonian samples, whereas marine catches were most variable at the less conserved cyt b and especially at the D-Loop. The least diverse Narcea River exhibited higher Hd at the highly conserved 12S rDNA than the two marine catches, due to Atlantic salmon polymorphism (likely due to a mixture of lineages remaining from past stocks transfers from North European populations; e.g. Horreo et al. 2011b).

The trade-off between using the same genetic analysis for simultaneously authenticating specimens and rapidly evaluating population diversity is that conserved species-specific sequences may not exhibit enough polymorphism. This is exemplified in Figure 1 and in the total number of variants of each marker found in this study, with 58 D-Loop versus only 15 12S rDNA haplotypes. Comparison between DNA regions for polymorphic information – measured as mean variation for each gene as in Figure 1 – yielded, despite small sample sizes, highly significant differences for all parameters when the six sequences were considered at the same time (p = 0.011, p = 0.006 and p = 0.000, for Hd, π and Nh/n, respectively). Most polymorphisms were provided by the non-coding D-Loop (Figure 1), and adding more nucleotides (concatenated sequence of all loci) did not increase significantly the level of polymorphism (p = 0.639, p = 0.109 and p = 0.428, for Hd, π and Nh/n, respectively). As expected, in relation with its length, the D-Loop was the most informative gene for quantifying diversity.