Phylogeography and taxonomic status of trout and salmon from the Ponto‐Caspian drainages, with inferences on European Brown Trout evolution and taxonomy

Abstract Current taxonomy of western Eurasian trout leaves a number of questions open; it is not clear to what extent some species are distinct genetically and morphologically. The purpose of this paper was to explore phylogeography and species boundaries in freshwater and anadromous trout from the drainages of the Black and the Caspian Seas (Ponto‐Caspian). We studied morphology and mitochondrial phylogeny, combining samples from the western Caucasus within the potential range of five nominal species of trout that are thought to inhabit this region, and using the sequences available from GenBank. Our results suggest that the genetic diversity of trout in the Ponto‐Caspian region is best explained with the fragmentation of catchments. (1) All trout species from Ponto‐Caspian belong to the same mitochondrial clade, separated from the other trout since the Pleistocene; (2) the southeastern Black Sea area is the most likely place of diversification of this clade, which is closely related to the clades from Anatolia; (3) The species from the Black Sea and the Caspian Sea drainages are monophyletic; (4) except for the basal lineage of the Ponto‐Caspian clade, Salmo rizeensis, all the lineages produce anadromous forms; (5) genetic diversification within the Ponto‐Caspian clade is related to Pleistocene glacial waves; (6) the described morphological differences between the species are not fully diagnostic, and some earlier described differences depend on body size; the differences between freshwater and marine forms exceed those between the different lineages. We suggest a conservative taxonomic approach, using the names S. rizeensis and Salmo labrax for trout from the Black Sea basin and Salmo caspius and Salmo ciscaucasicus for the fish from the Caspian basin.

S. t. lacustris and S. t. fario, and the anadromous S. t. trutta. Anadromy may not be a heritable character for brown trout (Berg, 1959(Berg, , 1962; hence, these names probably do not have any taxonomic meaning. Fishbase (www.fishbase.org), considering recent changes in taxonomy, lists more than 20 species that formerly were qualified as geographic populations or subspecies of S. trutta. Eight of those are potentially present in the Caucasus Ecoregion (as defined in Zazanashvili, Sanadiradze, Bukhnikashvili, Kandaurov, & Tarkhnishvili, 2004) and, broader, in the basins of the Black and the Caspian Seas (from here onwards-Ponto-Caspian Basin: the drainages of the Black and the Caspian Seas formed a contiguous body of water separated from the Mediterranean in the geological past- Popov et al., 2004). Salmo ciscaucasicus (Dorofeeva, 1967;syn. S. trutta ciscaucasicus) is native for the northwestern drainages of Caspian Sea, including the basin of the Terek (Tergi) River (Kottelat & Freyhof, 2007). Salmo caspius (Kessler, 1877; syn. S. trutta caspius) is native to the southern Caspian and the rivers of the northern Iran (Turan, Kottelat, & Engin, 2009). Salmo coruhensis (Turan et al., 2009) is an anadromous form from the southeastern and, possibly, eastern Black Sea basin in Turkey and Georgia.
Genetic and/or morphological and/or geographic distinctiveness was not demonstrated sufficiently well for some of these species. Turan et al. (2009) showed nearly fixed morphological and genetic differences between S. coruhensis and S. rizeensis from the southeastern Black Sea drainage. The individuals of S. labrax from the northern Black Sea drainage, described in the same paper, differ morphologically from another anadromous form, nominal S. coruhensis, although the differences are not fully diagnostic and the individuals were not studied genetically. It is not clear whether there are fixed differences between S. labrax and S. trutta from the rivers draining into the Black Sea from the north and the west, although some differences are mentioned (Kottelat & Freyhof, 2007). Fixed differences are suggested between S. caspius and S. labrax in the number of gill rakers (Turan et al., 2009); however, the studied individuals of S. caspius were smaller than those of the latter, and the influence of size on morphology cannot be excluded a priori. Moreover, the described S. caspius individuals were not studied genetically.
One could expect that speciation in brown trout should follow geological patterns, that is, catchments separated later should have more closely related trout lineages. This can be used as a null hypothesis challenged by the observed taxonomic diversity, which assumes the presence of more than one species in the same catchment: S. coruhensis, S. rizeenzis, but also potentially S. labrax in the western drainage of the Black Sea; S. ischchan and S. caspius in the Kura-Aras catchment (Lake Sevan is connected with the Aras River through the River Razdan); S. labrax and S. trutta in central and eastern European rivers.
Aiming to clear up these taxonomic and evolutionary puzzles and to infer which nominal species of brown trout are present throughout the Caucasus Ecoregion, the authors collected samples from six river drainages in Georgia, including four flowing into the Black Sea and two into the Caspian, we characterized external morphology of these fish, and analyzed their mitochondrial haplotypes, along with the haplotypes of brown trout from different sea and river drainages available from GenBank.

| Sampling
The sampling locations (and exact or approximate locations of fish for sequences that were downloaded from Genbank) are shown in Figure 1. The total number of samples used in genetic and morphological study is shown in

| Sequence analysis
We analyzed an 842-bp-long region of the mitochondrial cytochrome b (48 samples and 27 sequences downloaded from GenBank) and a 505-bp-long control region fragment (26 samples and 37 sequences downloaded from GenBank, Table 1 and Table S1). Unfortunately, the cytochrome b and control region sequences downloaded from GenBank were from different publications and described different individuals and populations; hence, we did not concatenate the studied fragments while inferring phylogenies and conducted the analyses separately for these two fragments of mitochondrial DNA.
We used three methods for the analysis of cytochrome b sequences: (1) Minimum spanning network was constructed only for the trout from the rivers drawing into the Black and the Caspian Seas (Ponto-Caspian Basin) using NETWORK 5.0 (Bandelt, Forster, & Rohl, 1999).
(2) Maximum likelihood (ML) tree was built and bootstrap support was estimated using software MEGA 7.0.21 (Kumar, Stecher, & Tamura, 2016 Carlo with chain length set at 100,000,000 to provide sufficient sample size for each parameter (i.e., effective sample size ≫ 100). The same substitution model was used as in the ML analysis.
We inferred the best-fit model and conducted ML analysis for the obtained and downloaded sequences of control region (see Table 1 for the list of the individuals used in this analysis). These sequences did not show much informative variability among the studied taxa, and we did not apply BI for this dataset. The software used was MEGA 7.0.21.
We used BEAST v. 1.8.4 (the uncorrelated log-normal relaxed molecular clock model) to account for variable rates of evolution among the lineages and to infer the 95% HPD intervals for the node ages,

| Morphometry
Fish caught in Georgia were photographed from the lateral side. The images were used for scoring the 28 conventional distances among

| Phylogeny based on cytochrome-b sequences
The optimal substitution model for the ingroups included in the analysis was TN93+G (Kumar et al., 2016).

| Control region
The optimal substitution model for the ingroups included in the analysis was T92 (Kumar et al., 2016). Both novel and downloaded sequences showed little variation within the studied fragment. However, all sequences from the drainages of Black and Caspian Seas were clustered in a monophyletic clade, albeit with modest bootstrap support, and related to trout from the Mediterranean (it was not possible to exactly identify locations for the sequences from Turkey). Remarkably, two sequences of S. trutta oxianus from the Aral Sea watershed belonged to the same clade. However, the sequences could not resolve F I G U R E 5 Timescale for the divergence of the clades shown in Figure 2, according to the consensus calibration of www.timetree.org. The numbers at the nodes-estimated divergence time (millions of years, mya); gray horizontal bars-95% HPD intervals phylogenetic relations of fish from the Ponto-Caspian basin, and failed to separate Black Sea, Caspian, and Aral Sea lineages ( Figure 6).

| Morphometry
Principal component analysis based on the standardized residuals of 27 size-removed body measurements extracted seven principal components with eigenvalues exceeding one. Cumulative weight of the components 1-4 exceeded 68% (Table 2a) Of the 32 studied individuals, the majority had 10 rays in dorsal fins, and three individuals, two from the Caspian, and one from the Black Sea catchment, had nine rays. Finally, the number of rays in pectoral fin varied between 12 (12 individuals) and 13 (20 individuals), without respect to body size, genetics, or river catchment.
There are more or less stable differences in color pattern between the marine and freshwater forms of trout (Figure 8). The background F I G U R E 6 Maximum likelihood tree based on the control region sequences of Salmo. Bootstrap support shown above the nodes. Identical haplotypes shown separately if found at more than one location. Clades with bootstrap support below 49 shown as polytomies. Red branches-samples from the drainage of the Black Sea; blue branches-samples from the drainage of the Caspian Sea; green branches-samples from the drainage of the Aral Sea was silver in marine form but yellowish in the freshwater trout. The individuals from Terek (Tergi) River were gray rather than yellowish.
Red spots on the lateral side are paler in marine forms.
In conclusion, the studied qualitative characters did not help much to distinguish between individuals of different genetic or geographic groups, in contrast with body shape dimensions.
T A B L E 2 The outcome of principal component analysis based on 28 size-removed measurements of fish body (only trout from Georgia included). (a) Eigenvalues and % of variance for the PC exceeding one; (b) loadings of individual distances (see Figure 2 for details) on the first seven PC axes Most of the subclades within this clade produce both riverine and anadromous forms, or may spontaneously switch between these two life modes. There are some differences in body proportions between different lineages of brown trout, although morphology should be used very carefully while differentiating among the nominal species: Our analyses suggest that body proportions and number of gill rakers, commonly used in taxonomy, may be associated with highly variable body size of the fish.
The populations from the Black Sea were captured during the consequential glacial cycles in small isolated streams flowing into the southeastern part of the sea, and in interglacial periods, probably used the sea as a transit area, dispersing through different rivers of the same drainage. One of these lineages might gain some morphological peculiarities and lose ability to switch to anadromy, the form recently described as S. rizeensis (Turan et al., 2009 (Figure 7), the morphological overlap is high.
Morphological differences between the fish from the basin of Terek and Mtkvari (Kura) are stronger and fixed.
The split between the four Black Sea clades, including the Danube clade, occurred shortly after the initial split and could be associated with increasing time of glacial cycles and decreasing temperature during the glacial maxima in mid-Pleistocene (Imbrie et al., 1993).

| Riverine and lacustrine versus anadromous mode of life-obligatory or facultative?
It has never been established to what extent anadromy is an inheritable feature of individual evolutionary lineages of brown trout. Even in better studied American rainbow trout, there are multiple gaps of knowledge concerning this question (Kendall et al., 2014 of anadromy of these forms may be related to a high salinity of the seas connecting to the rivers populated by these species (Thunell & Williams, 1989). We hypothesize that the reappearance of anadromy in brown trot from the Black Sea and Caspian catchments is related with low salinity of these seas.

| Taxonomic inference
The trout from the Caspian Sea basin ( We should emphasize the findings reported here are based on sequences from a mitochondrial cytochrome b fragment, and might be refined if nuclear sequence data were added to the analysis. However, based on a limited number of existing studies that combine nuclear and mitochondrial sequence data from salmonids, mitochondrial and nuclear sequences converge on the same clades in genus Salmo (Crête-Lafrenière et al., 2012). Simultaneously, one can expect patterns of incomplete lineage sorting and gene flow between closely related lineages, similar to that revealed in Central Europe as a result of microsatellite genotyping (Schenekar et al., 2014), and obviously further analysis including improved sampling and more DNA markers will substantially improve our knowledge of western Eurasian trout and salmon phylogeny and population structure.

ACKNOWLEDGMENTS
The project has been supported by the Ministry of Environment and Natural Resources Protection of Georgia (2014/ #148) and internal grant of Ilia State University. We appreciate Mariam Gabelaia who assisted with morphometric measurements, Marine Murtskhvaladze for assisting in the laboratory work. Cort Anderson corrected English of the manuscript and made valuable comments on the text and the analyses. We appreciate anonymous referee for comments on the first draft of the paper.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
LN conducted the field work, processed the samples for molecular genetic analysis, and together with DT analyzed phylogenies. DT and LN designed the study and prepared the manuscript. EG conducted morphometric analysis and assisted in DNA processing.