Phylogenetic relationships of freshwater fishes of the genus Capoeta (Actinopterygii, Cyprinidae) in Iran

2.1 Sample collection

Three hundred and five specimens of the genus Capoeta and one specimen of Barbus lacerta were collected by electrofishing at 47 sites in 13 river basins, covering most of its distribution in the country (Table 1; Fig. 1) with local authority permission. A fragment of pelvic fin was cut and stored in microtubes in 96% ethanol and deposited in the Tissue and DNA Collection of the National Museum of Natural Sciences of Madrid (MNCN-CSIC), Spain. Few fish from each site were killed with overdoses of MS222, fixed in 8% formalin, and later preserved in 70% ethanol in the Ichthyology collection of MNCN-CSIC, Spain.

2.2 DNA extraction, amplification, and sequencing

DNA was extracted using the DNeasy^® Blood & Tissue Kit (QIAGEN, Hilden, Germany). The entire cytochrome b (cytb) gene (1140 bp) was amplified by polymerase chain reaction (PCR) according to the protocol described in Perdices and Doadrio (2001) using GluDG.L and H16460 primers. Briefly, DNA was amplified in 25 μl reactions [1× buffer, 1.5 μM MgCl2, 0.5 mM of each primer, 0.2 μM dNTP of each nucleotide, 17.55 μl ddH₂O, 1 μl template DNA, and 1U Taq polymerase (Biotools, Madrid, Spain)]. PCR was performed at 95°C (2 min) followed by 30 cycles of 95°C for 30 s, 54°C or 1 min 20 s, 72°C for 2 min 20 s, and a final extension of 10 min at 72°C. PCR products were visualized on 1.5% agarose gels and later purified by ethanol. Both strands were sequenced using the service of Macrogen Inc. (Seoul, Korea).

Alignments of nucleotide sequences were constructed with CLUSTAL W using default parameters (Larkin et al., 2007), or with Geneious software (Geneious v. 8.0.3; Biomatters, http://www.geneious.com/), and visually verified to maximize positional homology. Codification of amino acids was used to confirm the alignment and to avoid the inclusion of stop codons. Sequences of the complete cytb gene were trimmed to the size of the smallest fragment, and alignments produced a dataset of 1040 base pairs (bp).

The dataset used for phylogenetic inference included 439 cytb sequences, 306 of which we amplified and other 133 correspond to the homologous region available for Capoeta in GenBank. Sequences of Luciobarbus subquincunciatus (Günther 1868), Luciobarbus brachycephalus (Kessler 1872), and Barbus lacerta Heckel 1843, obtained from GenBank, were used as outgroup (Berrebi & Tsigenopoulos, 2003; Levin et al., 2012; Machordom & Doadrio, 2001a). All sequences and GenBank accession numbers are listed in Tables S1–S3.

2.3 Data analysis

The sequences were collapsed to haplotypes using the program Alter (Glez-Peña, Gómez-Blanco, Reboiro-Jato, Fdez-Riverola, & Posada, 2010). Uncorrected-p pairwise distances between and within species (Table S4) were calculated with Mega 6 (Tamura, Stecher, Peterson, Filipski, & Kumar, 2013). A bootstrapping process was implemented with 1000 repetitions. As multiple tests, p-values were further adjusted by Bonferroni's correction (Rice, 1989).

The Akaike information criterion implemented in PartitionFinder v. 1.1.1 (Lanfear, Calcott, Ho, & Guindon, 2012) selected K80+I+G, F81+I, and GTR+I+G evolutionary models, considering each codon position as an independent partition. jModelTest 2.1.4 (Darriba, Taboada, Doallo, & Posada, 2012) selected GTR+I+G as the best evolutionary model for nonpartitioned sequence alignment. RAxML (Stamatakis, 2006) implemented in raxmlGUI 1.3 (Silvestro & Michalak, 2012) was used to estimate the maximum-likelihood (ML) tree using the selected evolutionary model for the sequence alignment. Bayesian inference was conducted with MrBAYES v. 3.2.2 (Ronquist et al., 2012). Two simultaneous analyses were run on 10⁷ generations, each with four MCMC chains sampling every 100 generations. Convergence was checked on Tracer 1.6 (Rambaut & Drummond, 2013). After discarding the first 10% of generations as burn-in, we obtained the 50% majority rule consensus tree and the posterior probabilities.

We reconstructed haplotype network sequences to resolve relationships among closely related haplotypes (Crandall, 1994). Haplotype genealogies in these clades were obtained by HaploView v. 4.2 (Barrett, Fry, Maller, & Daly, 2005).

To delimit species, two approaches were used. The Generalized Mixed Yule-Coalescent (GMYC) model, which uses a time-calibrated tree and delimits species on a divergence time basis (Fujisawa & Barraclough, 2013), and a Poisson tree process (PTP) model, using a distance-based tree to delimit species, implemented in Bayesian PTP (bPTP) (Zhang, Kapli, Pavlidis, & Stamatakis, 2013). Both GMYC and bPTP were accessed at Exelixis Labs (http://sco.h-its.org/exelixis/web/software/PTP/index.html).

Divergence times within Capoeta were estimated using a Bayesian relaxed molecular clock approach, implemented in BEAST v. 1.8 (Drummond, Suchard, Xie, & Rambaut, 2012). Dating analyses were performed using the evolutionary models described above. To calibrate the tree, an expanded sequence matrix was considered that included sequences of Barbus and Luciobarbus species (Table S3). For the molecular clock, multiple calibration points were based on fossil evidence of Barbus (18–19 Mya) and Luciobarbus genus (16.9–17.7 Mya) (Böhme & Ilg, 2003). The third calibration point considered was the Iberian Peninsula Luciobarbus species dated at 5.33–7.05 Mya (Doadrio & Casado, 1989; García-Alix, Minwer-Barakat, Martín Suárez, Freudenthal, & Martín, 2008). The branch rates were derived following an uncorrelated lognormal distribution and a Yule speciation prior (Drummond, Ho, Phillips, & Rambaut, 2006). Each final MCMC chain was run for 10⁸ generations (10% burn-in), with parameters sampled every 1000 steps. Output from BEAST was examined in Tracer 1.6 (Rambaut & Drummond, 2013), and the tree results were summarized using TreeAnnotator 1.7 (Drummond et al., 2012).

3.1 Mesopotamian clade

The Mesopotamian clade was the basal clade within Capoeta with high support values, primarily comprising species from the Tigris and Euphrates Rivers together with several river basins in southwestern Iran (Figs 2 and 3). Results suggested three well-supported subclades that corresponded to Capoeta trutta (Heckel 1843), Capoeta mandica Bianco and Banarescu 1982, and Capoeta barroisi Lortet 1894. The taxonomic status of the recently described Capoeta turani Özulug & Freyhof, 2008; was not well supported by our data. Genetic distances among these subclades ranged from 1.1% to 1.7% (Table S4). Species delimitation methods recognized C. mandica, C. barroisi, C. trutta, and C. turani as valid, even given the relatively low genetic distances separating them.

3.2 Aralo-Caspian clade

The Aralo-Caspian clade comprised populations of rivers that flow to the Aral, Orumieh, and Caspian seas, and several rivers in central Iran. This was a well-supported clade separated into three subclades (Fig. 4) with genetic distances among species ranging from 1.1% to 3.7% (Table S4).

The most basal subclade (Capoeta sp7) comprised specimens from the Tejan River (Caspian basin), which could not be attributed to any described species.

The second subclade included Capoeta capoeta (Güldenstädt 1773), Capoeta sevangi De Filippi, 1862, and Capoeta ekmekciae Turan, Kottelat, Kirankaya, and Engin, 2006, from Turkey, northwestern Iran, and Armenia (Fig. 5). Phylogenetic relationships within this subclade were unresolved, and genetic distances between species were low (0.7%). Results of the GMYC and bPTP differed for this subclade, suggesting the use of a genetic marker more sensitive to population differences. Sequences of C. ekmekciae were clustered together, and both species delimitation methods recognized it as a distinct evolutionary unit, which, along with the low genetic differences within this subclade (0.7%), could possibly be attributed to population differences within the same species. Interestingly, a specimen of this subclade was captured in the Caspian basin, which is geographically distant from the locations in which the other specimens were obtained, that is, chiefly the Orumieh basin and other areas in Turkey. This sample was checked twice to be sure that the result is not due to contamination (we repeated the DNA extraction, amplification and sequencing). The lack of more specimens presenting the same condition prevented us to reach any conclusion on it, so we prefer to interpret the data literally until we find more specimens and we study more in depth this question.

The third subclade included populations from north and northeastern river basins of Iran and river basins of Turkmenistan. We found two groups separated by genetic distances varying from 1.1% to 2.9%. The first consisted of two well-supported subgroups, Capoeta aculeata (Valenciennes 1844) occurring in the Karun and Kor basins, and a second occurring in the Karkheh River of the Tigris basin (C. aff aculeata). Both GMYC and bPTP identified the subgroups as distinct taxonomic units. The second group comprised three well-supported subgroups occurring in the Caspian, Tedzhen, and Namak basins. Mean genetic distances between sequence pairs within this group ranged from 1.4% to 2.5%. Relationships within this group were unresolved. One subgroup was found in southeastern regions of the Caspian basin and in the Tedzhen basin, which is located mainly in Turkmenistan and Afghanistan. For this subgroup, Capoeta heratensis (Keyserling 1861) is suggested as the valid name, as the species was described from specimens caught in a river near Herat in Afghanistan. The other two subgroups, belonging to Iranian populations with southern Caspian distribution, could not to be assigned to any described species. One was recognized by Levin et al. (2012) and designated Capoeta sp1. We retain this nomenclature for populations of the southern Caspian basin. The third subgroup occurred in Namak endorheic basin from two geographically close but separated rivers, the Jajrud and Namrud Rivers, and is herein referred to as Capoeta sp6. Both GMYC and bPTP recognized C. sp6, C. sp1, and C. heratensis as distinct species.

The haplotype network of the populations from the northern and northeastern river basins demonstrates clear structuring among basins (Fig. 6). No haplotypes were shared among populations of different basins. The group designated Capoeta sp1 was the most diverse taxon in this network, showing 19 haplotypes. Capoeta aculeata presented the lowest diversity, with two detected haplotypes. It is possible that the bigger number of samples for C. sp1 is biasing a little our results in some grade, but mostly, we believe that C. aculeata is living in a very arid region with very strong fluctuations on the water level (specially the population in Kor basin), so we suppose in their history they suffered many bottleneck events, which is not the case of C. sp1 which is present in a very humid region with high levels of precipitation. Also, we suppose that it have something to do with the fact that rivers in the Caspian basin, where C. sp1 is present, are all independent without freshwater connections what can help to keep rare alleles established in smaller independent habitats, which will show a higher diversity.

3.3 Anatolian-Iranian clade

The Anatolian-Iranian clade, the sister clade of the Aralo-Caspian clade, includes species widespread throughout the Anatolian peninsula and river basins of western and central Iran. This well-supported clade was the most diverse among the Capoeta, comprising six subclades, with genetic distances ranging from 1.5% to 5.4% (Figs 2 and 7).

The first subclade consisted of Capoeta sieboldii (Steindachner 1864) from the Kelkit River in Turkey, which drains into the Black Sea.

A second subclade split into two well-supported groups separated by a genetic distance of 2.7%. The first included populations from Turkey and was attributed to Capoeta bergamae Karaman 1969, and the second, also from Turkey, was described by Levin et al. (2012) and designated Capoeta sp2. Both groups were recovered as valid species by both methods used.

The third subclade included populations of Capoeta mauricii Küçük, Turan, Sahin, and Gülle 2009, inhabiting the Sarioz Stream and Eflatum Spring in the Beysehir Lake basin in southwestern Turkey.

The fourth subclade included Capoeta antalyensis (Battalgil 1943) from the Boga Cayi River in Turkey, near the type locality of the species. Again, both methods of species delimitation supported subclades three and four as valid species.

A fifth subclade was divided into two groups. The first group was formed by two well-supported subgroups. The first subgroup included samples identified as C. antalyensis. The genetic distance between sequence pairs within this subgroup was 0.2%, and both species delimitation methods considered the subgroup as a valid species. Hence, we tentatively interpret this as misidentification of these specimens, as samples of type locality of C. antalyensis were present in the fourth subclade. A second subgroup consisted of Capoeta baliki Turan, Kottelat, Ekmekçi, and Imamoglu 2006, and Capoeta tinca (Heckel 1843). However, the species delimitation methods used did not recover them as separate species. The second group of this subclade consisted of samples identified as Capoeta banarescui Turan, Kottelat, Ekmekçi, and Imamoglu 2006 and as C. cf banarescui by Levin et al. (2012). Both GMYC analysis and bPTP recognized two evolutionary units, one of which corresponded to C. banarescui.

The final subclade comprised two well-supported groups. One consisted of specimens from the Karkheh basin in western Iran. In a previous phylogenetic study, these fish were assigned to Capoeta sp3 (Levin et al., 2012). Both GMYC analysis and bPTP recognized it as a valid species. The second group of this subclade separated into two subgroups, the first found mainly in Turkey and the other primarily in Iran. The Turkish subgroup comprised two sets: Capoeta caelestis from the Ilica Stream in central Turkey and the Kargi Cayi River in southwestern Turkey, recognized as a single species by both species delimitation approaches, and a second set including Capoeta damascina (Valenciennes 1842); Capoeta kosswigi, Karaman 1969; Capoeta angorae (Hankó 1925); and two Iranian specimens from the Dez basin (Tigris tributary) not previously described (Capoeta sp4). Neither delimitation method recognized species differences, with the exception of Capoeta sp4, which both recognized as a valid species.

The second subgroup occurring mainly in Iran also split into two mitochondrial lineages. One included Capoeta buhsei Kessler 1877, of the Namak basin; Capoeta coadi Alwan, Zareian, and Esmaeili 2016, from the Tigris and Zayandeh Rud basins; and Capoeta sp5 not described previously, from the Zohreh basin. The methods used showed C. buhsei and Capoeta sp5 as valid species, but divided the C. coadi into two species, one occurring in the endorheic Zayandeh Rud basin and the other in the Karun basin of the Tigris drainage. However, the genetic distances were the lowest obtained in this analysis. The second mitochondrial lineage included Capoeta saadii (Heckel 1847), presenting strong geographic structuring. The GMYC method recognized three species: one in the Dalaki and Mand basins both draining into the Persian Gulf, one in the endorheic Kor basin, and one in the Rodan basin flowing into the Gulf of Oman. bPTP identified the four haplotypes of the Kor basin as separate evolutionary units, which we consider to be an artifact of the program. As the developers of both species delimitation methods recommend, their results have to be corroborated with other information sources. Here, in these four different “species” recognized by bPTP, all the information suggest that they are all the same species: GMYC does not recognize different species, genetic distances between them are very low, lower than some other within species distances in the genus, all samples come from the same sampling point, and they form all together a well-supported clade very close to all other samples from this species. So rather than different species, we interpret the results on this group as highly structured populations within the same species.

The network analyses showed high haplotype diversity and strong geographic structuring in C. saadii in comparison with the remaining haplogroups, with 14 haplotypes present in three basins and no haplotype shared among basins (Fig. 8). Generally, all Capoeta Species occurring in Karun and Mand basins presented high haplotype diversity, with seven haplotypes of C. coadi in Karun and seven haplotypes of C. saadii in Mand.

Separation of Capoeta from Luciobarbus was estimated to take place during the Middle Miocene, ca. 17.5 Mya (17–17.9 Mya) (Fig. 9). The divergence of the three main clades of Capoeta, the Mesopotamian, the Aralo-Caspian, and the Anatolian–Iranian, was estimated to have occurred in the Mio-Pleistocene (15.6–12.4 Mya), with the Mesopotamian clade diverging from the two other clades ca. 15.6 Mya (13.8–17.2 Mya) and separation of the Aralo-Caspian and Anatolian-Iranian ca. 12.4 Mya (10.5–14.4 Mya).

4.1 Taxonomy and species relationships of Capoeta in Iran

The phylogenetic analyses and the GMYC/bPTP clustering methods agreed in most cases and suggested novel taxonomic findings within Capoeta. With multiple species-level taxa within its nominal species, C. capoeta forms a species complex. We propose a new taxonomic status for C. mandica and recognize new candidate species C. sp4, C. sp5, C. sp6, C. sp7, and C. aff aculeata (Table 2). The species C. sp1, C. sp2, and C. sp3 previously proposed by Levin et al. (2012) were supported by our analyses. Twenty-six Capoeta species, including the eight species (C. sp1, C. sp2, C. sp3, C. sp4, C. sp5, C. sp6, C. sp7, and C. aff aculeata) proposed as candidate species, covering mostly of the genus distribution in Iran were recognized (Table 2).

4.2 Hypothesis for the origin of Capoeta in Iran

We found clear correspondence between geographical and genetic origin of the clades, seeming to follow a south–north pattern of distribution. The most highly diverged Mesopotamian clade occupied the southern regions with the Aralo-Caspian clade in the northern regions and the Anatolian-Iranian clade in the middle regions being the most widespread. The Anatolian-Iranian clade overlapped with the other clades at the borders of their distribution. This is probably due to recent dispersion events of the Anatolian-Iranian clade.

The divergence of the three main Capoeta lineages was estimated to have occurred around 15.6–12.4 Mya. The uplift of the Zagros Mountains ca. 35–20 Mya in southern Iran and their stabilization ca.12.4–10 Mya (Mouthereau, 2011) and the uplift of the Alborz Mountains ca. 20–17 Mya (Ballato et al., 2008, 2010) in northern Iran correlate with the divergence times of the three main Capoeta clades. These major mountainous systems likely acted as barriers to dispersion of the fauna and flora of the region (Ansell et al., 2011; Feldman & Parham, 2004; Hrbek & Meyer, 2003; Kapli et al., 2015; Parsa, Oraie, Khosravani, & Rastegar-Pouyani, 2009; Rastegar-Pouyani, Rastegar-Pouyani, Kazemi Noureini, Joger, & Wink, 2010; Šmíd & Frynta, 2012; Vamberger et al., 2013). Secondary dispersions, especially noticeable in the Anatolian-Iranian lineage and less obvious in Aralo-Caspian lineage, shaped the current picture of the distribution of the clades.

To decipher evolutionary and biogeographical patterns in Capoeta, previous authors have calibrated a molecular clock for the cytochrome b gene using fossil records (Levin et al., 2012). According to the molecular evolutionary rate based on a relaxed molecular clock, Capoeta arose in the middle Miocene 17.5 Mya when the Gomphotherium Landbridge was an important route of terrestrial fauna exchange between Africa and Asia (Harzhauser et al., 2007; Rögl, 1999). This period of middle-to-late Miocene was marked by the alternating periods of closure of the Tethys Sea and probably explains the split of the three main Capoeta clades, which is concurrent with the early development of Zagros Mountains, influencing the separation of basins and populations, especially in Iran. The Zagros Mountain uplift began in the mid-Miocene as a result of tectonic activity primarily resulting from contact of the Iranian and Arabian plates (Molnar, 2006). During the formation of Zagros Mountains, new freshwater bodies, along with new barriers within existing basins, shaped the generation of the main lineages, as supported by our results. This suggests major tectonic processes as the main speciation force within Capoeta, which may also be applicable to other freshwater groups.

4.3 Conservation

As mentioned before, there is an urgent need to assess the conservation status of these species and freshwater fauna in general in Iran. In general, main threats affecting Capoeta genus seem to be the water abstraction for irrigation projects and other human needs and habitat loss, especially in a mainly arid region which become dryer and warmer every year with less precipitations. In addition, this already fragile environment is also affected by the industrialization and constructions and pollutions related to it, which will have certainly a major role as a threat for all freshwater fauna in a developing country and show the need of more conservational controls and policies. Finally, there is also a very important impact on the local fauna caused by the invasive species, mainly commercially interesting species for human use as food. More regulations and a better environmental management are necessary in the country to preserve the rich and unique fauna living in the region.