MITO-NUCLEAR SEQUENCING IS PARAMOUNT TO CORRECTLY IDENTIFY SYMPATRIC HYBRIDIZING FISHES

Mito-nuclear sequencing is paramount to correctly identify sympatric hybridizing combination of mito-nuclear markers. Materials and methods. Specimens of Luciobarbus from six Guadiana River sub-basins were collected and sequenced for the cytochrome b and beta-actin genes. For comparative purposes, samples of Luciobarbus from other 12 river basins were also used. Four levels of taxonomical identification were conducted based on: identification made in the field ( in loco identification), cyt b gene only, beta-actin gene only, and mito-nuclear combined genomes. Results. Results showed that interspecific hybridization seems to be high (around 41%) and likely favoured by non-random mating and the loss of fluvial connectivity. About 34% of the hybrids showed mito-nuclear discordance. Misidentifications were frequent when only phenotypic characters are considered, and the use of a single mitochondrial gene is not sufficient: the use of two mito-nuclear markers showed that around 82% of the in loco identifications based on the phenotype were not correct. Conclusion. Incorrect species assignment likely generated biased results in previous studies on the biology and ecology of Guadiana barbels and in the assignment of conservation status and, consequently, on the establishment of conservation management measures.


INTRODUCTION
Since the 1930s, when the process of natural hybridization began to receive more attention from researchers, two divergent approaches emerged: botanists highlighted its potential for generating diversity (hybrids could occupy new habitats and originate new clades) and zoologists tended to see it as a reproductive mistake that limits diversification and retards evolution (reviewed by Barton 2001, Mallet 2005. According to the latter view, hybridization is the converse of reproductive isolation and challenges the biological species concept, thus the study of hybrids would only be relevant as a tool to understand the development of reproductive isolation (Mallet 2005).
The emergence of molecular biology techniques exposed the important role of hybridization as a source of genetic variation, a way to generate functional novelty and adaptability, and an important mechanism in the formation of new species (Mallet 2005, Bohling 2016.
According to the literature, the production of viable and fertile hybrids may lead to at least three possible evolutionary scenarios: If hybridization occurs repeatedly, the extensive gene flow may lead to the extinction of one of the hybridizing species through genetic assimilation (Costedoat et al. 2007) or to the merging of the hybridizing species (Taylor et al. 2006) If hybrids show reduced fitness a hybrid zone may be established, where gene exchange may occur but merging of the parental taxa is prevented (Barton et al. 1985) If reticulate evolution, when hybrids are at least partially reproductively isolated from the hybridizing species, and the formation of a new, allopolyploid or homoploid, hybrid species occurs (Schumer et al. 2014).
Once formed, fertile F1 hybrids may backcross with one or both the parental species, allowing the occurrence of gene flow and eventually leading to the incorporation of the genes of one species into the genome of the other species (introgressive hybridization) or to the complete merging of the previously isolated hybridizing species ("hybrid swarm") (Scribner et al. 2000). If two hybridizing species are common in their habitats, even low rates of hybridization may have relevant evolutionary consequences. The introgression of genes through hybridization may contribute to adaptive evolution and diversification by enhancing genetic variability (Mallet 2005), but may also raise important concerns regarding species integrity and conservation (Costedoat et al. 2007, Bohling 2016. Evidence for the occurrence of introgression of genes has increased in the last decades, as more molecular techniques became available to the researchers (Arnold 2006). These techniques inclusively detect unsuspicious cases of introgression, namely when parental species and their hybrids are morphologically similar (Gerlach et al. 2016, Paterson et al. 2016. The high incidence of hybridization in fish taxa seems to be a result of several contributing factors: external fertilization, weak behavioural isolating mechanisms, unequal abundance of the two parental species, competition for limited spawning habitat, decreasing habitat complexity, loss of fluvial connectivity, and susceptibility to secondary contact between recently evolved forms (reviewed by Scribner et al. 2000). The anthropogenic modification of river systems, namely through damming and habitat destruction, may thus be linked with higher levels of hybridization in fish (Hasselman et al. 2014).
Among European cyprinid fish, several cases of hybridization and introgression between native species belonging to the same (Machordom et al. 1990, Congiu et al. 2001, Almodóvar et al. 2008, Lajbner et al. 2009, Geiger et al. 2016 or to different genera (Bianco 1982, Hänfling et al. 2005, Ünver et al. 2005, Pereira et al. 2009, Aboim et al. 2010, Matondo et al. 2010, Kuparinen et al. 2014, Witkowski et al. 2015 were reported in the last three decades. The hybridization between sympatric species of Luciobarbus inhabiting the Iberian Peninsula was suggested by the finding of morphological intermediates (Almaça 1967, 1972, Machordom et al. 1990, Geiger et al. 2016, Gante et al. 2015 and later corroborated by molecular data (Machordom et al. 1990, Callejas and Ochando 2002, Gante et al. 2015. Recently, a wide study on Iberian Barbus and Luciobarbus pointed to the existence of incomplete reproductive isolation between sympatric species with semi-permeable barriers to gene flow (Gante et al. 2015). This work also showed that, despite the homogenizing effects of hybridization, Iberian barbels might still be discriminated by the combined used of morphological and molecular tools (Gante et al. 2015).
There are a number of morphometric and meristic measures considered to be diagnostic features of the Luciobarbus species inhabiting the Guadiana River (Almaça 1967, Doadrio et al. 2011 (Fig. 1). Nevertheless, routine biological surveys and technical reports (e.g., for environmental impact studies), and also scientific papers on the biology and ecology of these barbels, rely only on the morphological identification of captured fish using user friendly and in loco traits such as the width of the head, the relative size of the eye, the length of the barbels or the profile of the head (Lobón-Cervia et al. 1984, Encina and Granado-Lorencio 1990, Pires et al. 2001, Morán-López et al. 2005. Conservation efforts are also dependent on an adequate knowledge of barbel populations, and the location of priority areas for protection. Assuming that i) the use of morphological characters in loco may be subjective and observer-dependent, ii) there is previous evidence of interspecific hybridization between the sympatric Luciobarbus species in the Guadiana River, and iii) technical surveys conducted in this river basin (for monitoring the ichthyofauna of the Guadiana National Park or for environmental impact studies, for instance) usually do not include taxonomical identification based on molecular tools, our main goal is to assess the reliability of in loco species identification using morphometric indices, mtDNA only, and a combination of mito-nuclear markers. Considerations about conservation and management studies on Portuguese Luciobarbus and implications for future taxonomical assignments of hybrid species will be drawn.

MATERIALS AND METHODS
A total of 378 Luciobarbus specimens were collected between 2011 and 2012 in six Guadiana River sub-basins (Fig. 2), using electrofishing (SAMUS725G portable device): Ardila (n = 75), Caia (n = 11), Chança (n = 66), Cobres (n = 35), Degebe (n = 33), Odeleite (n = 54), Oeiras (n = 39), and Vascão (n = 65). Fish were identified in the field by two observers, according to practical guidelines followed by the technicians of the local Natural Park "Parque Natural do Vale do Guadiana" for more than two decades, based on the head dorsal profile and on the length of the second pair of barbels relative to the eye (Almaça 1967, Doadrio et al. 2011. Phenotypic traits used to identify Luciobarbus in the Guadiana River are summarized in Table 1. Juveniles smaller than 10 cm (n = 72) were also sampled but were not identified using the above mentioned diagnostic characters (Table 1) nor used for morphometric analyses, as their identification in the field is considered to be unreliable (Godinho et al. 1997). They were, nevertheless, included in genetic analyses. Dorsal fin clips were taken from all the specimens and preserved in 96% ethanol as vouchers for the tissue collection of For comparative purposes and to allow for the validation of nuclear-specific tags (see below), 157 samples of additional Luciobarbus populations (Fig. 2), available from the same tissue collection, were used for genetic analyses: 20 L. comizo from the Tagus River, 26 L. sclateri from rivers Seixe and Arade, and 111 Luciobarbus bocagei (Steindachner, 1864) from nine river basins located throughout the species distribution area (Table 2 and Fig. 2). These samples, except for the Tagus population (where L. bocagei is sympatric with L. comizo), were from populations where only one Luciobarbus species occurs, avoiding the potential noise of current interspecific hybridizations on the nuclear signal. The complete dataset of samples is presented in Table 2. Molecular data analyses. Total genomic DNA was extracted from fin clips using REDExtract-N-Amp Tissue PCR kits (Sigma-Aldrich) following the manufacturer's instructions. One mitochondrial (cytochrome b -cytb) and one nuclear gene (beta-actin) were amplified using the primers LCB1new-ACTTGAAGAACCAC-CGTTG (newly designed, based on the LCB1 primer described by Brito et al. 1997) and HA-CAACGATCTC-CGGTTTACAAGAC (Schmidt and Gold 1993) for cytb, and BACTFOR-ATGGATGATGAAATTGCCGC and BACTREV-AGGATCTTCATGAGGTAGTC (Robalo et al. 2007) for beta-actin. PCR conditions followed the ones described in Sousa-Santos et al. (2014). Primers LCB1new and BACTFOR were used for forward sequencing reactions and the PCR products were purified and sequenced at the GATC company (Germany). Obtained sequences were aligned and manually edited using CodonCode Aligner v4.0.4 (CodonCode Corp., USA). All beta-actin and some cytb (n = 60) sequences were newly obtained for this work whilst the remaining cytb sequences (n = 194) were previously obtained by the research team under the scope of the FISHATLAS project and were already available in GenBank (Table 3). GenBank accession number of all cytb sequences are presented in Table 3. The obtained beta-actin sequences were not genotyped (and, thus, not deposited in GenBank) since it was our goal to detect specific tags among the superimposed double peaks (see below). Chromatograms are available to be sent by the authors upon request.
As Luciobarbus species are tetraploid (Ráb and Collares-Pereira 1995), the amplification of the nuclear beta-actin gene generates mixed PCR products and the consequent production of traces with multiple peaks for the majority of the loci. Indeed, if the four alleles exhibit the same nucleotide in a particular locus, a single peak will be read by the sequencer. Contrastingly, it is possible that four nucleotides, one from each allele, may appear superimposed at a given locus of the sequenced gene  L. sclateri Barbels extending beyond the posterior edge of the eye L. steindachneri Barbels reaching the middle of the eye L. microcephalus Short barbels (not reaching the anterior edge of the eye) and short head with concave dorsal profile L. comizo Short barbels (not reaching the anterior edge of the eye) and long head with concave dorsal profile and duck-like snout Hybrids Intermediate characteristics from the above and automated/statistical approaches (Stephens et al. 2001, Bhangale et al. 2006, Scheet and Stephens 2006, Chen et al. 2007, Dmitriev and Rakitov 2008 have been developed in the past decade to disentangle both gene complements of diploid heterozygotes and hybrids. In some cases, these approaches benefit from the presence of diagnostic heterozygous insertions-deletions mutations (indels) which result in a phase shift in the trace, from the mutation point onwards (Sousa-Santos et al. 2005). For tetraploid individuals which are heterozygous for a given nuclear gene, however, the haplotype determination is not an easy task, especially if indels are present. Iberian barbels exhibit four copies of the betaactin gene, identical two by two, with several indels in conserved regions (Sousa-Santos unpublished data). These haplotypes may be recovered by using paralog-specific primers, as already described for the direct sequencing of the S7 and growth hormone genes of Barbus and Luciobarbus (see Gante et al. 2011). However, in the presently reported study, as our goal was to use the nuclear beta-actin gene to validate the direct sequencing of the mitochondrial cytb gene (or, on the contrary, to refute the mtDNA-based classification when hybridization signals are detected), we opted for a less expensive and expedite methodology which may be easily replicated in future studies. Thus, traces obtained from the beta-actin gene sequencing of Luciobarbus from the Guadiana River were aligned with CodonCode Aligner v4.0.4 (CodonCode Corp., USA). Automated base calling using the nucleotide ambiguity code (IUPAC) was made by CodonCode Aligner ("calling secondary peaks" function) and each locus was posteriorly manually inspected to search for point mutations which can be used as diagnostic of the species sampled. Samples of L. bocagei, L. sclateri, and L. comizo from other river basins (Table 2) were also manually inspected for support to the former identification of diagnostic loci. Tags identified for L. comizo and L. sclateri from Guadiana specimens were validated with the sequencing results obtained for individuals of the same species from other river basins (respectively, Tagus and Arade/Seixe). Identical procedure was not possible to conduct for the validation of L. microcephalus and L. steindachneri tags since these species only occurs in the Guadiana River.
After the identification of the diagnostic loci for the beta-actin gene, each individual was assigned to one of four categories: L. sclateri, L. microcephalus, L. comizo, and hybrid (when the beta-actin sequence shows tags which are specific of two or more Luciobarbus species).
Finally, we built a matrix summarizing the four levels of taxonomical identification for all sampled individuals: identification made in the field (in loco identification) (four categories, corresponding to the four Luciobarbus species), cytb genome (three categories, corresponding to L. sclateri, L. comizo and L. microcephalus), beta-actin genome (four categories, above mentioned); and mitonuclear combined genome (four categories: L. sclateri, L. microcephalus, L. comizo, and hybrid). Morphometric analyses. For each of the 276 collected adult specimens, individual images were taken with a PENTAX Optio E85 camera (available to be sent by the authors upon request). Each image was processed in Adobe Photoshop CS5 Extended with an X-Ray Filter to enhance general contrast, adjust brightness and contrast, cropped to reduce file size, and saved as a .tiff file with a traceable unknown ID label (Spec_nnn). Afterwards, each image was analysed using FIJI ImageJ v. 1.49 (Schindelin et al. 2012) with a multipoint tool to assign each point to up to 15 morphometric landmarks (Fig. 3). Image calibration was not possible in the field but instead, XY pixel coordinates were obtained for each point. The resulting matrix was transformed to .tps format and an EDMA (all distances between landmarks) matrix was calculated in PAST v. 2.16 (Hammer et al. 2001), from which 19 morphometric measures adapted from Armbruster (2012) were extracted (Table 4). These measures were selected since they were related with the main features used for the identification of Luciobarbus species, namely the morphology of the head and the relative Table 2 Luciobarbus specimens sampled in the Guadiana River sub-basins (n = 378) and in nine river basins located outside the study area (n = 157) In the river basins where more than one Luciobarbus species occur (Tagus and Guadiana), the field identification of the species was made according to the criteria described in the Materials and methods section. As occurred in the Guadiana, the L. comizo specimens from the Tagus were distinguishable from the sympatric L. bocagei due to the presence of short barbels (not reaching the anterior edge of the eye) and long head with concave dorsal profile and duck-like snout.

Table 3 Identification of the individuals sampled in the lower Guadiana River sub-basins (n = 194), based on different criteria
Individual ID Sub-basin        position of the fins (Almaça 1967, Doadrio et al. 2011. The use of geometric morphometrics, which is proven to be reliable for the distinction of Barbus species (Geiger et al. 2016), was not possible due to the impossibility of adequately collect landmarks in the field, since sampling was directed to collect fin clips for population genetic studies under the scope of the FISHATLAS project (Sousa-Santos et al. 2016). Thus, and although the main goal of the present paper was to assess the reliability of in loco species identification (commonly used in technical reports and environmental impact studies), we decided to use photographs made in the field to take traditional morphometric measures and further test if a simple and expedite methodological procedure as such, which could be conducted by less experienced technicians, would be reliable for species differentiation.

Statistical analyses.
To explore the relation between the genetic identity of the Guadiana barbels and the morphological variables studied we used raw data residuals of log-log regressions of the morphometric variables, using the head size (distance between the anterior limit of the head and the posterior limit of the operculum) as an independent variable. The variables SLLD (a meristic variable) and ANG (an angular measure) were not transformed. To perform a principal component analysis (PCA) and to maximize the number of individuals included (due to the high number of missing values), twelve variables were selected (the residuals concerning LDE, TDE, PEPO, MHD, TSAD, TSAP, TSAV, DP, DV, PV, DPO, and ANG). This later variable (ANG) was transformed by a logarithmic function to improve normality. Five extreme outliers were extracted from the analysis, which was conducted with the remaining 152 individuals. Both graphic visual inspections and values of kurtosis and skewness (between -2 and 2) suggest that these variables have approximately a normal distribution.
Kaiser-Meyer-Olkin measure of sampling adequacy and Bartlett's test of Sphericity were used to evaluate the suitability of the PCA analysis to this dataset. A PCA analysis was performed using the correlation matrix. The relation between the mito-nuclear genetic identity of individuals and the morphological analysis was examined using scatterplot graphic concerning the PCA components, here treated as new variables. Additionally, a discriminant analysis was performed using the same log-log residuals of the genetically pure L. sclateri and L. comizo (for L. microcephalus the number of individuals with morphological measures was very restricted). The discriminant function was then used to classify the genetically hybrid individuals. All statistical analyses were conducted using IBM SPSS Statistics, version 22. The presently reported study was carried out in accordance with the Portuguese state regulations and necessary permits to conduct fieldwork were required to the National Institute for the Conservation of Nature and Forests, Portugal.

RESULTS
Mitochondrial and nuclear genotyping. A total of 320 and 226 Luciobarbus adult specimens from the Guadiana River were sequenced for the cytb and beta-actin genes, respectively. Provisional species identifications based only on mtDNA results retrieved 141 L. sclateri, 135 L. comizo, and 44 L. microcephalus while that based on the beta-actin gene sequences retrieved 81 L. sclateri, 53 L. comizo, 20 L. microcephalus, and 72 hybrids (Table 5). Among nDNA sequences, 42 diagnostic single nucleotide polymorphisms (SNPs) were identified (Table  6). For the majority of the cases, instead of single peaks, two or three bases were found superimposed in each locus (represented in Table 6 by the respective ambiguity codes).
The analysis of the SNPs patterns (Table 6), corroborated with the mtDNA results and the sequencing of specimens from other river basins (for L. comizo and L. sclateri), revealed that at least six loci of the studied beta-actin gene fragment may be used as species-specific tags ( Table 7). All the L. sclateri specimens from Guadiana and from the rivers Arade and Seixe (where the species occurs allopatrically) show the exclusive R-C-G-Y-R-S/B combination of nucleotides (Tables 6 and 7). Luciobarbus comizo from Guadiana differ from L sclateri by a single diagnostic locus: a double peak of A and G nucleotides (R code of ambiguity) is found in locus 117, instead of the single G peak found in L. sclateri specimens ( Table 7). The L. comizo specific tag (117 R) was validated by using specimens from the Tagus River (n = 51): although no pure L. comizo individuals were detected, it was possible to detect the L. comizo specific tag in 11 L. bocagei × L. comizo hybrids (these two species occur in sympatry in the Tagus). Moreover, these hybrids showed a triple peak (B ambiguity code) at locus 279, corresponding to the mixture of the typical GC (S) of L. comizo with the typical GT (K) of L. bocagei (Table  6), reinforcing their identification as interspecific hybrids.
The SNP pattern of L. bocagei (R-C-R-Y-A-K) identified in specimens from the Tagus was validated by using 80 specimens of L. bocagei captured in six river basins where this species occurs allopatrically (Table 6). The beta-actin gene sequences of L. bocagei are extremely conserved: all these individuals, from the whole distribution area of the species, showed the same SNP pattern for the 42 analysed loci (Table 6). Interestingly, the typical L. bocagei SNP pattern was detected in one individual from the Guadiana bearing a L. microcephalus mtDNA (Table 6).
Regarding the target specimens from Guadiana, the majority of the individuals (n = 134, 59.3%) belonged to a given Luciobarbus species (61 L. sclateri, 49 L. comizo, and 24 L. microcephalus), confirmed by the analysis of their mtDNA and nDNA profiles, while the remaining individuals (n = 92; 40.7%) were interspecific hybrids. Among the hybrids, 51.1% (n = 47) had a mixture of species-specific tags of more than one Luciobarbus species in their nuclear sequences, and 48.9% (n = 45) exhibited mito-nuclear incongruence: the nDNA profile was typical of a Luciobarbus species which was distinct from that identified at the mtDNA level (Table 6). This latter group of interspecific hybrids (representing 33.6% of the total number of individuals sequenced) included 28 L. sclateri with L. comizo mtDNA, 15 L. comizo with L. sclateri mtDNA, one L. sclateri with L. microcephalus mtDNA and one individual with L. comizo and L. bocagei nuclear species-specific tags bearing mtDNA of L. microcephalus (Table 6).
When considering the percentage of hybrids in each sub-basin, we found that in the majority of the subbasins the percentage of hybrids was below 35% (25.0% in Degebe,27.3% in Cobres,29.6% in Oeiras,34.2% in Vascão,and 34.6% in Odeleite), while in the other subbasins these individuals were prevalent (50.0% in Ardila, 54.8% in Chança, and 60.0% in Caia). Taxonomical identification based on phenotypic and genetic data. Individual results of the phenotypic-, mtDNA-, and mito-nuclear-based assignments are presented in Table 6. From the 194 individuals identified in loco based on phenotypic traits (see Materials and Methods), 74 (38.1%) were identified as L. sclateri (Table 8). The remaining three species, L. steindachneri, L. comizo, and L. microcephalus, accounted for 17.0%-23.2% of the total number of individuals (Table 8). No hybrids were phenotypically assigned (Table 8).
If only mtDNA data is considered for taxonomical identification, the most common species would still be L. sclateri, although its relative frequency was higher than the one obtained when phenotypic characters were used for in loco identification (44.3% vs. 38.1%) ( Table 8). The relative frequency of L. comizo will also be higher (38.7% vs. 21.7%, respectively, for mtDNA-based and in loco identifications) but, contrastingly, the relative frequencies of L. microcephalus and L. steindachneri will be lower when using the identification based on mtDNA (17.0% vs. 23.2% and 0% vs. 17.0%, respectively) ( Table 8).
If only nDNA is considered, a fourth taxonomical category emerged ("hybrids", with a relative frequency of Nucleotides or ambiguity codes (in the case of intra-and interspecific hybrids) are indicated for each loci. The first six loci listed were considered to be species-specific tags and are highlighted in different colours (blue for L. microcephalus, green for L. sclateri, orange for L. comizo    the Luciobarbus species occurring in the Guadiana basin, considered to configure an extremely interesting model of speciation-with-gene-flow due to weak constraints to hybridization in breeding grounds (Gante et al. 2015).
In practical terms, however, this scenario results in the possibility of incorrect species identifications in the field and, consequently, calls into question all the surveys and published data made without molecular validation of the studied individuals. Indeed, our results clearly demonstrate that the in loco species identification based on phenotypic characters, which has been used for the last 30 years, is not reliable to distinguish Luciobarbus species in the lower Guadiana River and, consequently, their relative abundances were overestimated and hybrids, which represented approximately 41% of the individuals sequenced, were not detected nor considered. Simple morphological analyses using traditional morphometric measures, which could be an alternative to in loco phenotypic identification, are also not reliable since, according to our results, no clear separation between the genetic entities was found in the Guadiana River. More elaborate analyses or the use of geometric morphometrics might improve discrimination and clarify their eventual morphological distinctiveness, however, even if that was the case, the presence of a high percentage of hybrids justifies the use of molecular tools in all the studies conducted with Luciobarbus from the Guadiana River.
The implications of misidentifications might, in some cases, raise high levels of concern since important previously published data on the ecology and biology of the species (e.g., distribution areas, reproductive seasons, local abundance estimates, spawning behaviour, feeding ecology, or growth rates) may be questionable, as well as the conservation statuses assigned. Additionally, the establishment of conservation and management measures is usually made using data on population declines and fragmentation of populations, which may both be compromised by erroneous species identification in the field.
Along with species misidentification, our study also highlights the need to clarify the taxonomy of L. steindachneri since there is no evidence of a significant genetic divergence from the remaining Luciobarbus species that could support its specific status (Gante et al. 2009(Gante et al. , 2015. Indeed, our results show that the individuals identified in loco as L. steindachneri were instead interspecific hybrids (with mtDNA of one of the other three Luciobarbus species, indicating that mothers of all the species are involved in hybrid crosses) or showed pure genotypes (mostly of L. sclateri). The validity of this species, described almost 50 years ago based on morphological and meristic data (Almaça 1967), has been questioned (Doadrio 1988, Doadrio et al. 2002. Recently, in line with our findings, Gante et al. (2015) referred to this species as being the local product, in the Guadiana River, of the introgressive hybridization between L. comizo and L. microcephalus or L. sclateri. These authors suggest that L. steindachneri is an ecotype of hybrid origin, with intermediate molecular, morphological, trophic, and ecological characteristics. However, in our view, the maintenance of hybrids as an independent taxonomical entity with a conservation status is questionable and may result in more disadvantages than advantages, so we suggest that L. steindachneri species name should be considered no longer valid. Also regarding taxonomy, the detection of an individual with L. microcephalus mtDNA and a mixture of nuclear species-specific tags of L. comizo and L. bocagei is worth further investigation. The presence of L. bocagei genes outside of the species distribution area may be due to a human introduction. Another hypothesis which cannot be discarded yet is that the specific-tags of L. bocagei at the beta-actin level may be identical to the ones of Luciobarbus guiraonis, a species endemic to the Mediterranean slope of the Iberian Peninsula but which also occurs in some rivers of the upper Guadiana River basin (Doadrio et al. 2011).
Concerning the presence of intrageneric hybrids, previous studies had already reported their occurrence among barbels from Guadiana based on the detection of intermediate phenotypes (Gante et al. 2015). However, our study ads an extra worrisome result by showing that even individuals undoubtedly assigned to a certain species were indeed hybrids when genotyped (cryptic hybrids).
The existence of phenotypically unidentifiable hybrids could also explain the failure to clearly discriminate all the Luciobarbus species occurring in the Guadiana using morphological indices, despite the obtained significant discrimination between L. sclateri and L. comizo. These results corroborate the view highlighted by Gante et al. (2015) that the genomes of Iberian sympatric barbels remain porous and allow for gene exchange, despite being sufficiently divergent species. Indeed, previous phylogenies based on two mitochondrial genes calibrated using fossil evidence, showed that the lineage which originated L. microcephalus diverged around 7 Mya, and the lineage which originated the other three species was split around 4 Mya, given rise to L. sclateri on one hand, and to L. bocagei and L. comizo on the other hand (these two species were differentiated from each other more recently, around 1.9 Mya) (Gante et al. 2011).
Our study also revealed the occurrence of mito-nuclear discordance in a considerable number of individuals (around 34% of the total number of individuals sequenced) suggesting the presence of, at least, second-generation hybrids. The presence of cryptic hybrids and mito-nuclear discordances were already reported for a wide variety of animals (Toews and Brelsford 2012), including freshwater fish (Gante et al. 2009, Choleva et al. 2014, Sousa-Santos et al. 2014, Geiger et al. 2016. Mito-nuclear discordance, in particular, likely stem from the loss of a species-specific signal due to lineage sorting and/or non-assortative mating, as already proposed for other cyprinids (Freyhof et al. 2005, Broughton et al. 2011, Sousa-Santos et al. 2014.
Although females of all species of sympatric Luciobarbus (L. comizo, L. sclateri, and L. microcephalus) were involved in interspecific crosses, the prevalence of L. comizo-mtDNA in cryptic hybrids points to an eventual sexually biased direction of hybridization, as already suggested for hybrids between sympatric species of European Barbus (see Lajbner et al. 2009, Meraner et al. 2013, Buonerba et al. 2015. Indeed, as these authors suggest, females for the largest species (in our case, L. comizo) are more likely to attract smaller males from other species than vice versa, resulting in a higher percentage of interspecific hybrids carrying the mtDNA of the largest species that would be expected if mating was random. The local relative abundances of sympatric Luciobarbus species might also explain the detected differential contributions of females for interspecific crosses: females of the less common species will most likely produce hybrid progeny since finding a mate among conspecifics will be less probable than mating with congeners or with hybrids. Thus, as discussed by Wirtz (1999) and Rosenthal (2013), the absence of behavioural barriers to interspecific mating may promote hybridization and, furthermore, mate preferences and the scarcity of conspecific mates results in unidirectional hybridization processes. Massive mtDNA unidirectional introgressions attributed to demographical and/or behavioural reasons were already reported for a wide variety of taxa (Wirtz 1999, Ritz et al. 2008, Nevado et al. 2009, Sequeira et al. 2011. Future studies should be designed to allow the establishment of correlations between the type of mtDNA found in hybrid barbels from the Guadiana River and the local relative abundances of each parental species. Alongside distinct mtDNAs, cryptic hybrids also showed distinct frequencies according to the sub-basin considered, with a tendency to be more frequent in the ones with more dams (Ardila, Chança, and Caia). On the other hand, the higher percentages of concordance between in loco species identifications and mito-nuclear pure genotypes were detected in well preserved and dam free sub-basins (Vascão and Cobres). Thus, we suggest that hybridization may have been potentiated by the loss of river connectivity, which compromises the upstream migration of these potamodromous species and prevents the use of preferred spawning grounds, and by the lower availability of adequate habitats in more artificialized river systems. Positive correlations between damming and the occurrence of hybrids were already reported (Hasselman et al. 2014). This will undoubtedly lead to genetic homogenization, culminating in a loss of biodiversity.
The proven inadequate phenotypic diagnostic characters, the occurrence of cryptic hybrids in such an expressive percentage and the suggestion to fail to consider L. steindachneri as an independent taxonomical entity (and instead consider these individuals as interspecific hybrids) highlight the need to a careful review of the previously published data on biological and ecological features of Luciobarbus species. Furthermore, on-going conservation measures for threatened barbel populations should be reviewed in view of this hybrid puzzle scenario.
Several cases of hybridization between barbels were also reported elsewhere in Europe, based on the occurrence of morphologically intermediate hybrids and, less frequently, on inconsistencies between phenotypes and mitochondrial genotypes (reviewed by Geiger et al. 2016). Thus, the herein proposed use of a combination of mitochondrial and nuclear markers as a reliable method to non-erroneously identify barbels in the Guadiana River should become widely used in those river systems where different intrageneric sympatric species with soft mechanisms of reproductive isolation that might potentially interbreed. Reliable taxonomical assignments are crucial for species preservation since successful conservation plans need to consider the genetic integrity of their conservation units. We thus suggest that mito-nuclear sequencing becomes a standard practice to correctly identify fish where sympatric hybridizing species occur.