The phylogenetic relationships and species richness of host-specific Dactylogyrus parasites shaped by the biogeography of Balkan cyprinids

Parasites exhibiting a high degree of host specificity are expected to be intimately associated with their hosts. Therefore, the evolution of host-specific parasites is at least partially shaped by the evolutionary history and distribution of such hosts. Gill ectoparasites of Dactylogyrus (Monogenea) are specific to cyprinid fish. In the present study, we investigated the evolutionary history of 47 Dactylogyrus species from the Balkan Peninsula, the Mediteranean region exhibiting the highest cyprinid diversity in Europe, and from central European cyprinids. Phylogenetic analyses revealed four well-supported clades of endemic and non-endemic Dactylogyrus spp. with four basal taxa. Endemic cyprinids with a limited distribution range were parasitized by endemic Dactylogyrus species, but some of them shared several Dactylogyrus species with central European cyprinids. Species delimitation analyses based on molecular data suggest that Dactylogyrus diversity is higher than that defined from morphology. Some endemic cyprinid species harboured Dactylogyrus species of different origins, this probably resulting from multiple host switching. Our results support the view that the evolution of Dactylogyrus in the Balkans has been influenced not only by the historical dispersion and distribution of their cyprinid hosts, but also by recent contacts of non-native cyprinid species with endemic cyprinid fauna in this region.

Phylogenetic analyses and genetic distances. The concatenated sequence alignment of partial 18S and partial 28S rDNA from representatives of 54 Dactylogyrus species from the Balkan Peninsula and central Europe contained 1158 unambiguous nucleotide positions. The data were treated as partitioned and GTR+I was selected as the most optimal evolutionary model for the 446 bp-long partial 18S rDNA sequences, and GTR+I+G for the 712 bp-long partial 28S rDNA sequences. BI (Bayesian inference) and ML (Maximum Likelihood) analyses produced trees with identical topologies which varied in node support values (Fig. 1). The resulting phylogram divided most of the species into 4 strongly-to-moderately supported clades. Four Dactylogyrus species SciEntiFic REPoRTs | (2018) 8 Dactylogyrus zandti appeared to be a sister species to clades 3 and 4, but its position was not supported.
To resolve the phylogenetic relationships among groups within the second clade, we used a concatenated alignment of partial 18S, 28S rDNA, and the highly variable ITS1 (Internal Transcribe Spacer 1) region. The alignment of 86 sequences comprised 1503 unambiguously aligned nucleotide positions.The most optimal evolutionary models were TrNef+I for the alignment of 446 bp-long partial 18S rDNA sequences, SYM+G for the alignment of 344 bp-long ITS1 sequences, and TVMef+I+G for the alignment of 713 bp-long partial 28S rDNA sequences. BI and ML analyses generated trees with the same topologies (Fig. 2). The resulting trees were rooted using clade 1 from the first phylogenetic reconstruction (Fig. 1).
The phylogenetic analyses divided clade 2 into several strongly-to-moderately supported groups. Group A included species parasitizing Pachychilon, these sharing the same type of haptoral ventral bar with five radii, similar to the 'cornu' type 45 . This monophyletic group of Dactylogyrus spp. from Pachychilon was highly supported by both BI and ML analyses. All Dactylogyrus species of Scardinius (D. difformis, D. difformoides and D. izjumovae) formed a highly supported monophyletic group (group C). The group of two Dactylogyrus species from Alburnus (group B) formed a sister clade to the abovementioned species from Scardinius. Dactylogyrus prostae, D. nanoides, and D. folkmanovae from Squalius formed three very strongly supported monophyletic groups (groups D, E, and F, respectively). Group E also clustered with D. rysavyi from A. thessalicus, Dactylogyrus sp. 7 from C. knerii, and Dactylogyrus sp. 1 from S. tenellus, with strong support from both analyses. All three species exhibit a similarly shaped MCO and parasitize phylogenetically closely related cyprinid lineages 26,45 .
The phylogenetic relationships between Dactylogyrus spp. of Barbus and those of Luciobarbus were unresolved. However, Dactylogyrus spp. of these cyprinids formed three well supported groups (G, H and I). All While Dactylogyrus sp. 2 and Dactylogyrus sp. 3 were found to be almost identical on the basis of morphological characters, they differed at the molecular level (concatenated partial 18S rDNA and ITS1 region, p-distance = 0.041). Our results did not support the monophyly of D. petenyi, as this species clustered with D. malleus and D. prespensis (group G). Dactylogyrus omenti from Aulopyge huegelii appears also to be phylogenetically closely related to the species parasitizing Barbus and Luciobarbus, but its position was only moderately supported by BI analysis. The position of D. rosickyi of P. pictum was also uncertain; however, BI analysis strongly supported its position within the clade including groups C-I. Dactylogyrus rutili from Rutilus formed a well-supported group (group J) and, according to our results, appears to be phylogenetically closely related to D. suecicus (whose monophyly was not supported) and Dactylogyrus sp. 8 from T. karsticus. Surprisingly, D. ergensi collected from three host species formed a paraphyletic group. Dactylogyrus ergensi from C. ohridana was phylogenetically related to  The computation of genetic distances between specimens of generalist Dactylogyrus species revealed moderate-to-high interpopulation genetic variability. Pairwise genetic distances were calculated for D. vistulae, D. rarissimus, and D. folkmanovae after eliminating all positions containing gaps and missing data. The selected species are representatives of Dactylogyrus with a wide distribution range in Europe. While D. folkmanovae is a parasite only of Squalius spp., D. vistulae and D. rarissimus are real generalists parasitizing on species of different cyprinid genera. An alignment of 994 nucleotide positions was used for D. vistulae collected from 24 cyprinid species of six genera at 20 localities across the Balkan Peninsula and the Czech Republic. Pairwise sequence diversities varied from 0.000 to 0.020 (Table 2). Generally, geographically adjacent populations were more similar at the molecular level, a finding supported by the Mantel test (P = 0.015). Dactylogyrus vistulae from S. tenellus, S. svallize, S. illyricus, Phoxinellus pseudalepidotus, P. alepidotus, and T. metohiensis were genetically identical and all their host species were from the Dalmatian ichthyogeographical district. The same pattern was observed for D. vistulae specimens from C. nasus and Leuciscus idus, both from central Europe: they were similar at the molecular level. One of the few exceptions was D. vistulae from S. cephalus in the Czech Republic, which was genetically more similar to Balkan populations collected from S. squalus and S. vardarensis than to central European populations. Dactylogyrus rarissimus was collected from 11 species including four cyprinid genera -Alburnus, Pelasgus, Rutilus and Telestes. After removing gaps and missing data, the final alignment contained a total of 978 nucleotide positions. The interpopulation genetic variability ranged from 0.001 to 0.030 (Table 3). The pairwise distances revealed that D. rarissimus from R. rutilus and R. lacustris were the most similar (p-distance = 0.003). Specimens of D. rarissimus from T. alfiensis were the most genetically dissimilar to all other specimens collected from other host species (p-distance > 0.021). Regarding D. rarissimus, the Mantel test did not reveal any significant spatial genetic structure (P > 0.05). Dactylogyrus folkmanovae specimens were collected from seven Squalius species at nine localities from the Balkans and central Europe. The final alignment contained 977 positions and genetic distances varied from 0.002 to 0.037 (Table 4). Interpopulation genetic variability was found even between specimens collected from two populations of one host species, namely S. prespensis (p-distance = 0.002), where both populations were in the same ichthyogeographical district. Surprisingly, the same genetic distance was observed between D. folkmanovae specimens collected from S. cephalus in Bosnia and Herzegovina and from S. cephalus in the Czech Republic. The Mantel test indicated a positive correlation between genetic and geographical distance for D. folkmanovae populations (P = 0.001).  Species delimitation. The species status of Dactylogyrus parasites exhibiting high interpopulation molecular diversity was investigated on the basis of a statistical analysis of our sequence data using PTP. We examined all specimens from clade 2 (Fig. 2). Results of the maximum likelihood analysis (Fig. 3)

Discussion
The present study suggests that the diversity of Dactylogyrus species parasitizing endemic cyprinids in the Balkans is poorer when compared to the diversity of Dactylogyrus from central European cyprinids and from cyprinids with a large distribution range (e.g. Šimková et al. 11 documented up to 9 different Dactylogyrus species from widely distributed Rutilus rutilus in the Czech Republic). High numbers of Dactylogyrus species were also observed on African cyprinids from the genus Labeo, such as L. coubie with 9 Dactylogyrus species 46 . In contrast, we observed a maximum of 5 Dactylogyrus species on a single cyprinid species. These numbers are consistent with previous observations of southern European Dactylogyrus fauna, where no more than 5 species were collected from one cyprinid host species 7,44,45 . Such low Dactylogyrus species diversity probably has several causes. The distribution range of host species highly influences parasite diversity 47 . Our observations support Gregory's hypothesis 37 , i.e. fish species with a wide distribution range are exposed to more parasite species; therefore, they exhibit high parasite diversity. Another potential explanation could be the following: host species with a wide distribution range include a much higher number of populations in comparison to endemic species, which favours parasite speciation. This is illustrated in the present study by R. rutilus and R. aula. While R. rutilus, referred to above as a species with a high Dactylogyrus species richness, is the cyprinid species with the widest distribution range in Europe, the distribution area of R. aula is limited to the Adriatic basin in Italy and the northwestern Balkans (the Northern Adriatic ichthyogeographical district 15 8 ) and include several sampling periods from different river basins, while the present study is focused on a single sampling period in a specific region. The investigated cyprinid hosts endemic to the Balkans are generally distributed in a restricted region where the number of populations potentially harbouring different parasites is expected to be rather lower than in central Europe. Therefore, also following Gregory's hypothesis, we expected lower parasite diversity in endemic cyprinids with a restricted distribution range. Only a few host species, such as S. squalus, were collected from several distinct localities; however, the different host populations did not differ in their numbers of Dactylogyrus species. It was also shown that the composition of monogenean communities is influenced by environmental factors, especially water temperature. In such cases, shifts in the species compositions of monogenean communities within host species were observed throughout the year [49][50][51][52][53] . The present phylogenetic analyses revealed four well-to-moderately supported clades including both endemic and non-endemic Dactylogyrus species, while four species -namely, D. erhardovae, D. crucifer, D. caballeroi, and D. rarissimus (all parasites of Rutilus spp.) -had external positions to these clades. Dactylogyrus erhardovae is considered to be a genus specific parasite of Rutilus, the first description of this species originating from R. rubilio 54 , an endemic species of the Apennine Peninsula 55,56 . In the Balkans, Dactylogyrus erhardovae was also found on R. aula and R. basak, phylogenetically closely related species 26,57 distributed in the rivers of the Adriatic Sea basin, which is the proximal ichthyogeographic district to the Tyrrhenian Sea basin, where R. rubilio occurs. Dactylogyrus crucifer was originally described from Rutilus rutilus, but Šimková et al. 12 collected this species also from Leuciscus idus and Scardinius erythrophthalmus and therefore suggested that D. crucifer represents a generalist species. In our study, D. crucifer was only collected from Rutilus species (R. rutilus from the Czech Republic and R. lacustris from the Ponto-Caspian area), which supports the association between Rutilus hosts and D. crucifer and even indicates that the occurrence of this parasite on other cyprinid species may be the result of accidental infection. Both Rutilus species parasitized by D. crucifer originated and live in sympatry in the Black Sea and Caspian Sea basins 58 , which may promote the host switching of D. crucifer between these two sister Rutilus lineages.
Interestingly, we showed that Dactylogyrus sp. 4 from D. adspersus and D. sekulovici from P. pictum clustered together (group 1). Both Dactylogyrus species seem to be host specific -at least, there are no previous records of these two species from other cyprinid species. Regarding the morphology of the hard parts, these two Dactylogyrus species differ in the shape of their MCOs. While Dactylogyrus sp. 4 has hard parts morphologically similar to those of D. erhardovae from Rutilus, it shares with D. sekulovici only the shape of the haptoral connective bars (see Pugachev et al. 45 for morphology of D. sekulovici). Two cyprinid species -namely, D. adspersus and P. pictum -are representatives of two phylogenetically unrelated ancient lineages 26 , but have a similar geographical distribution, i.e. they are restricted to the rivers of the Adriatic Sea Basin. Pachychilon pictum occurs only in the Albanian ichthyogeographical district 59 ; D. adspersus inhabits the central Adriatic (Dalmatian) district, which shares only two species with the Danubian basin [59][60][61] , and is probably linked to the Adriatic district by underground connections 16 . The paraphyly of the Dactylogyrus species from P. pictum suggests their multiple origin on this host. The phylogenetic proximity of D. sekulovici to Dactylogyrus sp. 4 suggests a host switch between two cyprinid species living in the same area of the central Adriatic region. The second host-specific parasite of P. pictum is D. ivanovichi 44,45 . Its phylogenetic position suggests a different origin (when compared to D. sekulovici), likely also resulting from a host switch. Dactylogyrus ivanovichi is phylogenetically closely related to D. auriculatus from Abramis brama. The two species exhibit MCOs with an identical structure and differ only in the positioning of the VA and in the root lengths of haptoral anchor hooks 45 . These two species, like the two species of the sister clade (clade 3), secondarily lost their connective haptoral ventral bar 45 . The phylogenetic proximity of D. ivanovichi and D. auriculatus and the morphological similarities in copulatory organs between D. ivanovichi and Dactylogyrus spp. of A. brama suggest that D. ivanovichi originated from a recent host switch from the widely distributed A. brama, and then adapted its attachment organ to new host species. Other Dactylogyrus species from P. pictum, namely D. martinovici and D. petkovici, are phylogenetically closely related to Dactylogyrus sp. 5 of P. macedonicum. Dactylogyrus martinovici, D. petkovici, and Dactylogyrus sp. 5 exhibit haptoral hard parts with an almost identical shape but differ in the shapes of their copulatory organs. This is in congruence with Šimková et al. 6 , suggesting similar adaptations of the haptor among Dactylogyrus species parasitizing phylogenetically related hosts. We can hypothesize that these three species evolving from the same ancestor have for a long time been associated with Pachychilon and that D. martinovici and D. petkovici emerged as a result of more recent intra-host duplication followed by reproductive isolation. In contrast, D. ivanovichi and D. sekulovici are the result of earlier host switching between cyprinid species of different genera living in contact zones and of subsequent speciation. Finally, another Dactylogyrus species from P. pictum, D. rosickyi, exhibits a different phylogenetic position when compared to the aforementioned Dactylogyrus of Pachychilon spp., which suggests a different origin for this species.
Regarding Dactylogyrus from Barbus spp., our analyses did not fully resolve the phylogenetic relationships between these species, but in general all species are clustered in three well or moderately supported groups (G-I).
In total, we collected 5 different Dactylogyrus species from 10 Barbus hosts. The most common was D. dyki, parasitizing 8 Barbus species and representing one clade in our phylogenetic analysis. Šimková et al. 43 observed significant interpopulational phenotypic plasticity and molecular variability among D. dyki isolated from 3 Barbus species, which is in accordance with the present study. The monophyly of the group including D. dyki specimens was supported. However, low support for D. dyki from B. strumicae was found and these specimens were recognized as a different species by species delimitation analysis. Following the suggestion of Šimková et al. 43 , D. dyki from Barbus spp. could represent a species complex of several morphologically similar species. The confirmation of this hypothesis requires further morphological reevaluation of Dactylogyrus representatives from all Barbus hosts, including those from B. meridionalis in Western Europe and B. tyberinus from the Apennines. We inferred some paraphyly concerning D. balkanicus. Whilst Dactylogyrus specimens of B. prespensis and B. rebeli were clustered together, specimens from B. plebejus appeared to be phylogenetically related to D. dyki. The sister status of these two species is supported by the similar shape of the sclerotized parts of their haptors (both species share a small triangular connective ventral bar), and also the remarkably similar shape of their MCOs 45  to another endemic Barbus, or (2) parasite duplication on recent Barbus species in this region and a host switch to the phylogenetically and geographically closest Barbus species. According to our phylogenetic analyses, D. petenyi, D. malleus, and D. prespensis form a well-supported group, namely group G. These three Dactylogyrus species parasitizing Barbus species share similar morphologies of the copulatory organs and haptoral hard parts. Surprisingly, specimens of D. petenyi do not form a monophyletic group. Species delimitation analysis suggests that each representative of group G represents a single species. Specimens of D. crivellius from different host species formed a monophyletic group. Our phylogenetic analyses support a monophyletic group including D. crivellius from Balkan Barbus spp., D. carpathicus from B. barbus, and Dactylogyrus sp. 2 and Dactylogyrus sp. 3. These four species exhibit the same morphology of a ventral bar with 5 extremities, a typical feature of Dactylogyrus spp. from Luciobarbus. Species with this morphology are considered as the 'carpathicus' 42 or 'cornu' 45 type. This supports the hypothesis that haptoral hard parts are more suitable for resolving the phylogeny of monogeneans; that is, haptor morphology is similar between closely related species 6,63,64 .
The phylogenetic position of D. omenti among Dactylogyrus species parasitizing Barbus and Luciobarbus was already suggested by Benovics et al. 65 . Even though its exact phylogenetic position is not fully resolved, our result suggests that this species is phylogenetically closer to D. petenyi and D. prespensis than to the aforementioned species which share the 'cornu' type of haptoral ventral bar. Adding more Dactylogyrus species from Iberian, North African, and Middle Eastern Barbus and Luciobarbus in a phylogenetic reconstruction and assessing coevolutionary scenarios involving these parasites and their hosts could better resolve the relationships within this group of Dactylogyrus.
Several well-supported phylogenetic groups (J-L) were formed exclusively by Dactylogyrus species of the 'ergensi' type of copulatory organ, or, in the case of D. tissensis, the 'chondrostomi' type of copulatory organ 47 . While the MCO and VA among Dactylogyrus spp. belonging to groups J-L are very similar, these species differ in the shapes and sizes of their haptoral hard parts. All Dactylogyrus species of groups K and L parasitize species of the genera Alburnoides and Chondrostoma. The species status of D. caucasicus parasitizing Alburnoides and that of D. dirigerus parasitizing Chondrostoma were supported by species delimitation analysis. Surprisingly, Rutilus-specific D. rutili belonging to the phylogenetically distant group J possesses the same type of copulatory organ as D. caucasicus and D. dirigerus. This suggests that a similar copulatory organ morphotype can emerge independently several times during the evolution of Dactylogyrus species in evolutionarily distant hosts (such are Rutilus, Chondrostoma, and Alburnoides 26 ). Rohde 2 hypothesized that the rapid evolution of morphological variation in copulatory organs is considered as a mechanism for avoiding hybridization. In contrast, similar types of copulatory organs in Dactylogyrus species may be recognized in different host lineages, as shown in the present study. Then, species with a similar MCO morphotype could be found within congeneric hosts only if these parasite lineages had diversified recently (e.g. D. ergensi and D. dirigerus of Chondrostoma).
High numbers of southern European endemic Dactylogyrus species were strictly host specific and/or distributed only in one region. However, some of them were collected from a wide range of cyprinid hosts. Dactylogyrus vistulae is the species with the widest host range in the Balkans. In addition to the host range for this parasite revealed in this study, the presence of D. vistulae was also reported from R. rutilus in Finland 66 and from V. vimba in the Czech Republic 8 . Genetic distances between specimens collected from different host species correlated with   geographical distances, suggesting the geographical structure of D. vistulae populations, rather than some association with the phylogenetic relatedness of the host species. For example, D. vistulae from C. phoxinus appears to be genetically more similar to D. vistulae from hosts in the same or close ichthyogeographical region than to D. vistulae collected from geographically separated congeneric Chondrostoma. Since D. vistulae is widely distributed and relatively easily distinguishable from other Dactylogyrus spp. on the same hosts (on the basis of morphological characters and its large body size 45 ), it could potentially represent a suitable model for population studies that could elucidate the origin of this species and the distribution pattern between phylogenetically distant hosts or between two host species from different regions. Another species with a wide distribution range is D. rarissimus. It was originally considered as a specialist of R. rutilus 6,12,67 ; however, we collected this species in the Balkans from phylogenetically well-separated genera: Rutilus, Alburnus, Pelasgus and Telestes. In this case, the Mantel test did not reveal a significant correlation between genetic and geographical distances, even as specimens collected from T. alfiensis and P. laconicus in Peloponnese (the only representatives of D. rarissimus from the Ionian ichthyogeographic district) are genetically the most different from northern populations originating from the Albanian district (such as R. ohridanus). We measured only a very small genetic difference between D. rarissimus from R. rutilus and D. rarissimus from R. lacustris (similarly to that measured for D. crucifer), which supports the recent divergence of these Rutilus species or, alternatively, a more ancient separation followed by recent contact. All these results suggest that D. rarissimus is a true generalist species parasitizing several cyprinid genera. We investigated the correlation between genetic and geographical distances among D. folkmanovae individuals. In contrast to D. vistulae and D. rarissimus, D. folkmanovae was reported as a generalist parasite of S. cephalus and R. rutilus 8,67 ; however, it is generally reported in Squalius species 12 16 , or through the introduction of non-native species/populations into the Balkan region. Fish introduction has been a very common occurence in the Balkans and includes both exotic, and native species from geographically near localities 68,69 . River drainages 70,71 and also isolated karstic drainages are affected, where non-native species such as S. cephalus and R. rutilus have been introduced 72 . Low molecular variability between Czech and Bosnian-Herzegovinian populations of D. folkmanovae may favour the hypothesis of the natural dispersion of the fish via river connections. However, the investigation of other European populations and the use of other genetic markers suitable for population genetics of Dactylogyrus are necessary to reveal the distribution patterns of widespread Dactylogyrus species. In addition, the extent of parasite transfer from introduced species to endemic species needs to be studied further to reduce the possible risk of parasite introduction to already threatened native species.
In this study, we revealed interpopulation genetic variability within endemic Balkan Dactylogyrus species. The intraspecific genetic distances could also be linked to the morphological variability which was suggested for other monogenean taxa [73][74][75] . Concerning Dactylogyrus, morphological variability among the haptoral hard parts of a given Dactylogyrus species was recorded even within a single host specimen of L. maghrebensis 71 , but without any molecular variability, suggesting phenotypic plasticity and/or selection within a specific microhabitat. On the other hand, as documented above, our molecular data also revealed potential complexes of cryptic species, formerly considered to be a single species solely on the basis of a morphological approach. According to species delimitation analysis, the 38 Dactylogyrus species included in the analysis may in fact represent 47 species. This finding is in accordance with previous studies, in which delimitation analyses were incongruent with classical taxonomy 76,77 . In our study, Dactylogyrus sp. 2 and Dactylogyrus sp. 3 from L. graecus and L. albanicus, respectively, were shown to be morphologically indistinguishable species; however, molecular data suggest that they are actually two different species (which is also supported by species delimitation analysis). A similar result was revealed for other Dactylogyrus species, such as D. rutili, which seems, on the basis of delimitation analysis, to represent three species parasitizing three host species, and D. dyki, which seems to represent six potential species on 10 Barbus host species. Our future aim will be to undertake the morphometrical reevaluation of taxonomically important traits in combination with the use of molecular data in order to resolve the potential species complexes previously recognized within Dactylogyrus 76 .

Material and Methods
Parasite sampling. From 2014 to 2017, individuals from 63 cyprinid fish species were sampled from 47 different localities in the Balkan Peninsula and the Czech Republic (Table 5, Fig. 4). Approximately 90% of all endemic cyprinid species in the Balkans were processed in this study 15 . Fish were dissected using the standard methods described by Ergens and Lom 78 and their Dactylogyrus species were collected. More precisely, Dactylogyrus specimens were removed from the gills, mounted on slides, and covered in a mixture of glycerine and ammonium picrate (GAP 79 ) for further determination. All applicable institutional, national and international guidelines for the care and use of animals were followed and approved by the Animal Care and Use Committee of the Faculty of Science, Masaryk University in Brno (Czech Republic). Identification at the species level was performed using an Olympus BX51 microscope equipped with phase contrast optics. Dactylogyrus species were determined using Pugachev et al. 45 on the basis of the size and shape of the hard parts of the attachment organ (the haptor) and the reproductive organs (MCO and VA). Some Dactylogyrus specimens from each cyprinid species investigated were bisected using fine needles under a dissecting microscope, and the body part with the haptor was individually preserved in 96% ethanol for further DNA extraction. The remaining body part, i.e. that including the hard parts of the respective reproductive organ, was mounted on a slide for species determination.
DNA extraction, amplification, and sequencing. Individual parasites were removed from the ethanol and dried using a vacuum centrifuge. DNA was extracted using the standard protocol (DNeasy Blood & Tissue Kit, Qiagen, Hilden, Germany). Partial 18S rDNA and the the entire ITS1 region were amplified using the primers S1 (5′-ATTCCGATAACGAACGAGACT-3′) and IR8 (5′-GCTAGCTGCGTTCTTCATCGA-3′) 80 , which anneal to the 18S and 5.8S rDNA respectively. Partial 28S rDNA was amplified using the following primers: forward C1 (5′-ACCCGCTGAATTTAAGCA-3′) and reverse D2 (5′-TGGTCCGTGTTTCAAGAC-3′) 81 . Each amplification reaction for partial 18S rDNA and the ITS1 region was performed in a final volume of 15 µl, containing 1.5 units of Taq polymerase, 1X buffer, 1.5 mM MgCl 2 , 0.2 mM of each dNTP, 0.5 µM of each primer, and 2.5 µl of DNA (20 ng/µl). PCR was carried out using the following steps: 2 min at 94 °C, followed by 40 cycles of 1 min at 94 °C, 1 min at 53 °C, and 1 min 30s at 72 °C, and 10 minutes of final elongation at 72°C. The PCR for partial 28S was performed using the same conditions as described in Šimková et al. 82 . The PCR products were checked on 1% agarose gel and purified using ExoSAP-IT kit (Ecoli, Bratislava, SK) following the standard protocol. Purified products were directly sequenced using the PCR primers and BigDye Terminator Cycle Sequencing kit (Applied Biosystems, Foster City, CA). Sequencing was performed on an ABI 3130 Genetic Analyzer (Applied Biosystems). New sequences were deposited in GenBank (their accession numbers are shown with asterisks in Table 5).
Phylogenetic analyses. DNA sequences were aligned using fast Fourier transform in MAFFT 83 . The sequences were trimmed to concur with Dactylogyrus sequences obtained from GenBank. The sequences for 14 Dactylogyrus species from central European cyprinids were obtained by sequencing in this study or acquired from GenBank (see Table 5 for accession numbers).
Genetic distances between specimens of selected Dactylogyrus species collected from different host species were computed using sequences of partial 18S rDNA combined with ITS1 region. Uncorrected pairwise distances were calculated in MEGA 7 84 .
Gaps and ambiguously aligned regions were removed from the alignment using GBlocks v. 0.91 85 . Phylogenetic analyses using maximum likelihood were computed with RaxML v8.1.X 86 , and by means of Bayesian inference with MrBayes 3.2 87 . For each analysis, jModelTest 2.1.10 was employed to select the most appropriate model of DNA evolution 88,89 using the Bayesian information criterion (BIC). Trees obtained by ML analyses were validated using 1000 bootstrap iterations. Bayesian inference was performed using the Metropolis-coupled Markov chain Monte Carlo algorithm, with 2 parallel runs of 1 cold and 3 hot chains. This was run for 10 7 generations and trees were sampled every 10 2 generations. 30% of all saved trees were discarded as a relative burn-in period according to the standard deviation split frequency value (<0.01).
Phylogenetic reconstruction including all sampled Dactylogyrus species was based on concatenated sequences of partial 18S rDNA and partial 28S rDNA (Fig. 1). The resulting phylogram was rooted using the evolutionarily divergent lineage of Dactylogyrus species parasitising Carassius gibelio and Cyprinus carpio 12 . To resolve the phylogenetic relationships among specific subgroups, partial subtree analyses were performed using partial 18S rDNA combined with the ITS1 region and partial 28S rDNA. Optimal evolutionary models were selected for each marker using BIC, each model including an alpha parameter for the gamma distribution (G) accounting for rate heterogeneity across sites and/or a proportion of invariable sites (I).
Species delineation in the final trees was carried out using a PTP (Poisson Tree Processes) model 90 . This approach was applied to the BI tree computed from concatenated partial 18S rDNA, 28S rDNA, and the partial ITS1 region, and run for 5 × 10 5 generations. 30% of the resulting trees were discarded as burn-in. PTP can give species delimitation hypothesis based on gene trees inferred from molecular sequences, modelling the speciation or branching events in terms of the number of mutations. This method does not require an ultrametric input tree or a sequence similarity threshold as input, but uses only the tree resulting from either phylogenetic reconstruction.
The Mantel test 91 to test the correlation between genetic and geographical distances was performed in R 92 using the mantel function in the vegan package 93 .

Data Availability
All new sequences of Dactylogyrus obtained during this study were deposited in NCBI GenBank under accession numbers MG792838-MG793066. Appropriate accession numbers according to Dactylogyrus species and specific rDNA regions are presented in Tables 1-3. Since whole fish specimens were completely processed during parasitological dissection, additional specimens of each analysed host species were collected from the same locality and fish vouchers were deposited in the ichthyological collection of the National Museum in Prague (Czech Republic). Voucher specimens of the sequenced Dactylogyrus species (excluding undescribed species) are deposited in the Finnish Museum of Natural History in Helsinki (available under the accession numbers MZH KN10850-989).