Genetic Evidence for a Mixed Composition of the Genus Myoxocephalus (Cottoidei: Cottidae) Necessitates Generic Realignment

Sculpin fishes belonging to the family Cottidae represent a large and complex group, inhabiting a wide range of freshwater, brackish-water, and marine environments. Numerous studies based on analysis of their morphology and genetic makeup frequently provided controversial results. In the present work, we sequenced complete mitochondrial (mt) genomes and fragments of nuclear ribosomal DNA (rDNA) of the fourhorn sculpin Myoxocephalus quadricornis and some related cottids to increase the power of phylogenetic and taxonomic analyses of this complex fish group. A comparison of the My. quadricornis mt genomes obtained by us with other complete mt genomes available in GenBank has revealed a surprisingly low divergence (3.06 ± 0.12%) with Megalocottus platycephalus and, at the same time, a significantly higher divergence (7.89 ± 0.16%) with the species of the genus Myoxocephalus. Correspondingly, phylogenetic analyses have shown that My. quadricornis is clustered with Me. platycephalus but not with the Myoxocephalus species. Completely consistent patterns of divergence and tree topologies have been obtained based on nuclear rDNA. Thus, the multi-gene data in the present work indicates obvious contradictions in the relationships between the Myoxocephalus and Megalocottus species studied. An extensive phylogenetic analysis has provided evidence for a closer affinity of My. quadricornis with the species of the genus Megalocottus than with the species of the genus Myoxocephalus. A recombination analysis, along with the additional GenBank data, excludes introgression and/or incorrect taxonomic identification as the possible causative factors responsible for the observed closer affinity between the two species from different genera. The above facts necessitate realignment of the genera Myoxocephalus and Megalocottus. The genetic data supports the two recognized genera, Myoxocephalus and Megalocottus, but suggests changing their compositions through transferring My. quadricornis to the genus Megalocottus. The results of the present study resolve the relationships within a complex group of sculpin fishes and show a promising approach to phylogenetic systematics (as a key organizing principle in biodiversity research) for a better understanding of the taxonomy and evolution of fishes and for supplying relevant information to address various fish biodiversity conservation and management issues.


Fish Specimens and Sequencing
The environmental status of the fourhorn sculpin My. quadricornis is considered as stable; it is listed under Least Concern on the International Union for Conservation of Nature's (IUCN) Red List of Threatened Species [37]. The other sculpin fishes studied here are not included in the IUCN Red List of Threatened Species; they are not listed as endangered, vulnerable, rare, or protected species in the Russian Federation, either. The sampling points are located beyond any protected areas. The sites of the field studies are not privately-owned or protected.
The specimens of My. quadricornis were collected from Sukhoe More Bay, Dvina Bay, White Sea  Table S1). The genomic DNA was extracted using a KingFisher Flex System and a set of reagents of a MagMAX DNA Multi-Sample Kit (ThermoFisher Scientific, Waltham, MA, USA).
The complete mt genome and rDNA sequences were amplified in five and three overlapping fragments, respectively, using the Phusion High-Fidelity DNA Polymerase (ThermoFisher Scientific). Libraries were prepared using an Ion Plus Fragment Library Kit and unique adapters (Ion Xpress, Waltham, MA, USA) with pre-fragmentation on a Covaris M220 Focused ultrasonicator. Ready libraries were sequenced on an Ion S5 sequencing platform (ThermoFisher Scientific) at the Far Eastern Federal University (Vladivostok, Russia). The complete mt genomes obtained were initially annotated using the MitoFish Web Server [38] and further manually adjusted in MEGA 7 [39] by comparing them with mt genomes of other sculpin fishes. The rDNA sequences were aligned using SPAdes 3.14.1 [40] with a correction of IonTorrent data using the IonHammer tool available in the SPAdes software and annotated using sequences of other Scorpaeniformes fishes available in GenBank. The nucleotide sequences were deposited in GenBank under the accession numbers MT303953-MT303954, MT890585-MT890586, MT906795-MT906796, MT909822-MT909823 (mt genomes) and MT497857-MT497873 (rDNA). Additional complete mt genomes and separate COI gene sequences of sculpin fishes were accessed from the GenBank National Center for Biotechnology Information (NCBI) genetic sequence database [41] (see Table S2 for accession numbers).

DNA Sequence Analysis
The nucleotide sequences were aligned using the MUSCLE [42] and MAFFT v. 7 [43] software. The alignments were analyzed for evidence of recombination using various recombination detection methods provided in the RDP4 program [44]. The DnaSP v. 5 [45], PROSEQ v. 2.9 [46], and MEGA 7 [39] programs were used for interspecific analysis of divergence. Phylogeny reconstructions were based on the mt genome, rDNA, and COI gene alignments using the maximum-likelihood methods available in IQ-TREE [47,48]. ModelFinder [49] or jModelTest [50] were used to find the best-fit model of nucleotide substitution under the maximum likelihood criterion. The General Time Reversible + gamma + invariant sites (GTR + G +I; [51]), Tamura-Nei (TN93; [52]), and Hasegawa-Kishino-Yano (HKY; [53]) models showed the lowest Akaike Information Criterion (AIC; [54]) value (189,144.4265, 22,798.8538, and 6315.5052) and Bayesian information criterion (BIC; [55]) scores (189,529.5437, 23,075.7355, and 6625.4694) for the mt genome, rDNA, and COI gene alignments, respectively, and they were selected for further phylogenetic reconstructions. The ultrafast maximum likelihood bootstrap analysis [56] consisted of 1000 replicates. Partitioned analyses were performed in PartitionFinder 2.1.1 [57] that allowed the overall rate to vary between partitions including each protein-coding gene (see Table S3 for the best model of substitution under the maximum likelihood criterion).
The complete dataset was also analyzed by the Bayesian inference using MrBayes v. 3.2.7a (released March 6, 2019; [58]) under the GTR + G + I model. Analyses were performed as two independent runs, each with four incrementally heated Metropolis-coupled Monte-Carlo Markov Chains. Output trees and data were sampled for every 500 generations. Likelihood values reached a plateau within 15,000-25,000 generations. A total of 4002 trees in two files were read and 3002 of them were sampled. For the mt genomes, the log likelihood values increased from below −148,989.341 to around −94,582.436 in the first 5000 generations and then to around −94,479.814 after 1,000,000 generations. The likelihood of the best state for "cold" chain of run 1 was −94,452.62, and the likelihood of the best state for "cold" chain of run 2 was −94,452.71. The average standard deviation of split frequencies was 0.000824 after 1,000,000 generations, indicating stationary conditions. In convergence diagnostic, the Potential Scale Reduction Factor (PSRF) [59] was between 1.000 and 1.011 for all parameters, thus, indicating a good sample from the posterior probability distribution. The log likelihood values and standard deviations of split frequencies, as well as the values of convergence diagnostic for the rDNA and COI genes, are provided in the Text S1 file.
The tree topologies for the sculpin fishes obtained by the maximum-likelihood method and by the Bayesian inference were very similar. The close congruence can be explained by the fact that the three separate datasets were relatively straightforward and included the complete mt genomes, rDNA, and COI sequences of sculpin fishes only. The quality of alignment was high, and the length of alignment was long enough (a total of 19.1 kb for the mt genomes, 6.3 kb for the rDNA, and 1.1 kb for the COI gene). As has been shown by many authors, the relative efficiencies of different methods applied for obtaining correct tree topology are very close under these conditions [60]. The difference between the methods applied to the sculpin fishes (our data) resulted in the slightly different topologies and bootstrap values and did not cause changes in the relationships between My. quadricornis and its close relatives. To be conservative, we here provide the lowest bootstrap values obtained by the maximum-likelihood method.

Complete Mitochondrial Genomes
The mt genomes (120× coverage) of My. quadricornis, P. allisi, and A. zanderi are 16 (Figure 1; Cluster 1). The level of divergence between My. quadricornis and Me. platycephalus (3.06 ± 0.12%) matches the average value of divergence between different species within the same genus (intrageneric level) of cottids [16][17][18]. Another group of sequences ( Figure 1; Cluster 2) contains the Myoxocephalus species only, including My. polyacanthocephalus, My. jaok, and My. scorpius. The level of divergence (Dxy = 0.0314 ± 0.0012) between the sequences within this group is slightly higher, but still matches the value of divergence for different species within the same genus (intrageneric level). The level of divergence between clusters 1 and 2 is significantly higher (Dxy = 0.0789 ± 0.0016), matching the values of divergence, albeit the lowest ones, between different genera within the family Cottidae (intrafamily level) [16][17][18]. Comparable values of divergence based on complete mt genomes have been obtained in the pairwise comparisons between the other genera of the cottids studied ( Figure 1 The species belonging to the genera Argyrocottus and Porocottus are clustered with the genus Myoxocephalus ( Figure 1). The level of divergence (Dxy) between A. zanderi and P. allisi (0.1093 ± 0.0021), as well as between these two species and the Myoxocephalus cluster 2 (0.0950 ± 0.0017), matches the value of divergence between different genera within the family Cottidae (intrafamily Another group of sequences ( Figure 1; Cluster 2) contains the Myoxocephalus species only, including My. polyacanthocephalus, My. jaok, and My. scorpius. The level of divergence (D xy = 0.0314 ± 0.0012) between the sequences within this group is slightly higher, but still matches the value of divergence for different species within the same genus (intrageneric level). The level of divergence between clusters 1 and 2 is significantly higher (D xy = 0.0789 ± 0.0016), matching the values of divergence, albeit the lowest ones, between different genera within the family Cottidae (intrafamily level) [16][17][18]. Comparable values of divergence based on complete mt genomes have been obtained in the pairwise comparisons between the other genera of the cottids studied ( Figure 1 The species belonging to the genera Argyrocottus and Porocottus are clustered with the genus Myoxocephalus ( Figure 1). The level of divergence (D xy ) between A. zanderi and P. allisi (0.1093 ± 0.0021), as well as between these two species and the Myoxocephalus cluster 2 (0.0950 ± 0.0017), matches the value of divergence between different genera within the family Cottidae (intrafamily level). The species belonging to the genus Microcottus is clustered with the genus Megalocottus ( Figure 1). The level of divergence between M. sellaris and the Megalocottus cluster 1 (0.0829 ± 0.0025) also matches the intrafamily level.

Nuclear Ribosomal Genes
The 6.3 kb fragment of the nuclear rDNA sequences of the genera Myoxocephalus and Megalocottus shows completely consistent patterns of divergence and tree topologies with those observed for the mt genomes. There are two clusters containing Myoxocephalus species (Figure 2; Cluster 1 and Cluster 2). The first cluster contains both My. quadricornis and Me. platycephalus with the difference between them being very low (D xy = 0.0018 ± 0.0005). The second cluster contains the Myoxocephalus species only, including My. jaok, My. stelleri, My. brandtii, and My. polyacanthocephalus, with the average distance (D xy ) equal to 0.0037 ± 0.0005. The difference between Clusters 1 and 2 is significantly higher (D xy = 0.0071 ± 0.0010). The members of the genera Microcottus and Porocottus form separate clusters outside of both Myoxocephalus and Megalocottus, with relatively high genetic distances (D xy ) varying from 0.0122 ± 0.0015 to 0.0160 ± 0.0017 and a significant bootstrap support ( Figure 2). The tree topologies obtained for the Myoxocephalus + Megalocottus and Microcottus + Porocottus clusters are slightly different when based on the rDNA fragments ( Figure 2) and mtDNA (Figure 1), which can reflect specific evolutionary trends in nuclear and mitochondrial genomes. The phylogenetic discordance between nuclear and mitochondrial genomes is not a rare phenomenon among fishes [61] and many other organisms [62]. level). The species belonging to the genus Microcottus is clustered with the genus Megalocottus ( Figure 1). The level of divergence between M. sellaris and the Megalocottus cluster 1 (0.0829 ± 0.0025) also matches the intrafamily level.

Nuclear Ribosomal Genes
The 6.3 kb fragment of the nuclear rDNA sequences of the genera Myoxocephalus and Megalocottus shows completely consistent patterns of divergence and tree topologies with those observed for the mt genomes. There are two clusters containing Myoxocephalus species (Figure 2; Cluster 1 and Cluster 2). The first cluster contains both My. quadricornis and Me. platycephalus with the difference between them being very low (Dxy = 0.0018 ± 0.0005). The second cluster contains the Myoxocephalus species only, including My. jaok, My. stelleri, My. brandtii, and My. polyacanthocephalus, with the average distance (Dxy) equal to 0.0037 ± 0.0005. The difference between Clusters 1 and 2 is significantly higher (Dxy = 0.0071 ± 0.0010). The members of the genera Microcottus and Porocottus form separate clusters outside of both Myoxocephalus and Megalocottus, with relatively high genetic distances (Dxy) varying from 0.0122 ± 0.0015 to 0.0160 ± 0.0017 and a significant bootstrap support ( Figure 2). The tree topologies obtained for the Myoxocephalus + Megalocottus and Microcottus + Porocottus clusters are slightly different when based on the rDNA fragments ( Figure 2) and mtDNA (Figure 1), which can reflect specific evolutionary trends in nuclear and mitochondrial genomes. The phylogenetic discordance between nuclear and mitochondrial genomes is not a rare phenomenon among fishes [61] and many other organisms [62].
Nuclear ribosomal genes are frequently used in molecular systematics [63,64]. Both small (18S) and large (28S) ribosomal genes, characterized by slow rate of evolution, have been used for phylogenetic studies on distantly related species to resolve macroevolutionary issues such as the origin of tetrapods and general vertebrate phylogeny [65]. Despite rDNA genes are less sensitive than mtDNA genes in analysis of close groups (species or genus level), nevertheless, data obtained using nuclear ribosomal markers are important for comparative purposes. As an independent source of genetic information, our rDNA data confirms the results obtained based on the mtDNA sequences. Nuclear ribosomal genes are frequently used in molecular systematics [63,64]. Both small (18S) and large (28S) ribosomal genes, characterized by slow rate of evolution, have been used for phylogenetic Genes 2020, 11, 1071 7 of 15 studies on distantly related species to resolve macroevolutionary issues such as the origin of tetrapods and general vertebrate phylogeny [65]. Despite rDNA genes are less sensitive than mtDNA genes in analysis of close groups (species or genus level), nevertheless, data obtained using nuclear ribosomal markers are important for comparative purposes. As an independent source of genetic information, our rDNA data confirms the results obtained based on the mtDNA sequences.

GenBank COI Gene Dataset
As in the case of mt genome and rDNA data, the COI gene cluster 1 contains both the Myoxocephalus species (My. scorpioides, My. octodecemspinosus, My. aenaeus, and My. quadricornis) and Me. platycephalus (represented by two subspecies, Me. p. taeniopterus and Me. p. platycephalus; see [20]) (Figure 3). The average difference (D xy ) between the species of Cluster 1 is 0.0387 ± 0.0047 (using three randomly picked sequences per species). The COI gene Cluster 2 contains the species belonging to the genus Myoxocephalus only, including My. stelleri, My. ochotensis, My. jaok, My. polyacanthocephalus, My. brandtii, and My. scorpius (Figure 3). The average difference (D xy ) between the species of Cluster 2 is 0.0265 ± 0.0037. The average levels of divergence within both clusters are relatively low, varying within 2.7-3.9%, which fits into the range of interspecific divergence for sculpin fishes [16][17][18]36]. The level of divergence between Clusters 1 and 2 is significantly higher, D xy = 0.0826 ± 0.0091, indicating the intergeneric level of divergence for cottids [16][17][18] (see also the Section 3.1). Thus, the COI dataset containing multiple GenBank sequences shows a significantly higher divergence (8.3%) between the clusters than within them (2.7-3.9%), which agrees with the analyses based on the complete mt genomes and the rDNA sequences.
The GenBank COI sequences demonstrate also a close similarity of the three other Myoxocephalus species, My. aenaeus, My. octodecemspinosus, and My. scorpioides, to the My. quadricornis plus Me. platycephalus cluster (Figure 3, Cluster 1). This suggests that the above-listed three species may also belong to the genus Megalocottus. The close affinity between My. cf. scorpiodes and Me. platycephalus was also reported previously by Knope [19] based on mitochondrial (cytb) and nuclear (the first intron of the nuclear S7 ribosomal protein) markers. However, the available data are still insufficient to draw any reliable conclusion.

Alternative Hypotheses: Recombination and Species Misidentification
The phylogenetic inconsistency that we detected may reflect the historical hybridization event(s) between the Myoxocephalus and Megalocottus species which could result in interspecific recombination of their mtDNA (as it has been found in other organisms including fishes; see, e.g., [66,67]), or it may be due to incorrect taxonomic identification. We, therefore, analyzed the mt genome alignments for evidence of recombination using various recombination detection methods provided in the RDP4 program [44]. All the methods failed to reveal any signal of recombination between the mt genomes of the Myoxocephalus and Megalocottus species, thus rejecting hybridization as a possible explanation for the anomalous similarity between the My. quadricornis and Me. platycephalus mt genomes.

Alternative Hypotheses: Recombination and Species Misidentification
The phylogenetic inconsistency that we detected may reflect the historical hybridization event(s) between the Myoxocephalus and Megalocottus species which could result in interspecific recombination of their mtDNA (as it has been found in other organisms including fishes; see, e.g., [66,67]), or it may be due to incorrect taxonomic identification. We, therefore, analyzed the mt genome alignments for evidence of recombination using various recombination detection methods provided in the RDP4 program [44]. All the methods failed to reveal any signal of recombination between the mt genomes of the Myoxocephalus and Megalocottus species, thus rejecting hybridization as a possible explanation for the anomalous similarity between the My. quadricornis and Me. platycephalus mt genomes.
To exclude the probability of species misidentification, we collected all available GenBank COI sequences for the genera Myoxocephalus and Megalocottus (Table S2) and, again, obtained two significantly diverged groups of sequences containing the Myoxocephalus and Megalocottus species (Figure 3; Cluster 1 and Cluster 2). It is important that the My. quadricornis sequences (MT303953 and MT303954) obtained in the present study are identical or very similar to the GenBank COI sequences (KJ128649, MG422966, EU524915, KX145213; Figure 3, see also Table S2 for additional sequences absent in Figure 3) reported by many other authors (Table S2), which also rejects misidentification as the possible causative factor responsible for the observed close affinity between My. quadricornis and Me. platycephalus.

Discussion
The mt genome and nuclear rDNA data have revealed a mixed composition of the genus Myoxocephalus, presumably including also a Megalocottus species. The observed inconsistencies are supported by the GenBank COI sequences representing all the available Myoxocephalus and Megalocottus species. Both mitochondrial and nuclear markers indicate close evolutionary relationships between the genera Myoxocephalus and Megalocottus (as well as other studied sculpins belonging to the genera Microcottus, Argyrocottus, and Porocottus). Close associations between these genera were also inferred from morphological [5,6,[13][14][15] and genetic [16][17][18][19]21,22,68] studies. However, the present genetic data does not support the synonymization of Megalocottus and Myoxocephalus, as was proposed by a number of authors (e.g., [10,11,[69][70][71]). The genetic distance between them is significantly larger than the average interspecific distance within the genus Myoxocephalus (see above) and fits into the range of intergeneric distances detected for sculpin fishes [16][17][18], nevertheless, having the lowest values (see also the Results section). Instead, our genetic data indicate that the composition of the genera Myoxocephalus and Megalocottus needs revision. In particular, the closer affinity of My. quadricornis to the genus Megalocottus than to the genus Myoxocephalus, detected using both mt genomes and rDNA genes, strongly suggests the necessity of generic realignment by transferring My. quadricornis to the genus Megalocottus. The morphological, ecological, and life-history traits of the species do not contradict the genetic data obtained and also support the necessity of generic realignment. My. quadricornis and Me. platycephalus share a number of common features, including some ecological preferences and morphological traits, that distinguish them from the Myoxocephalus species [5,6,8,11,12]. In particular, the topography of the seismosensory system, which is a highly specific character used in fish taxonomy [6,12], is similar between My. quadricornis and Me. platycephalus on the one hand and differs them from the rest of the Myoxocephalus species on the other hand [6].
Our genetic data are consistent with the earlier reports of Cowan [12,72] who investigated relationships within the genus Myoxocephalus based on 45 morphological and 35 biochemical (isozymes) characters. The study placed My. quadricornis in the most basal position of the trees containing 13 Myoxocephalus species and constructed using morphological [12] and combined morphological and biochemical [72] data. Our results also agree with the data obtained by Knope [19], who showed that Myoxocephalus is not a monophyletic genus. The six Myoxocephalus species included in that study formed a monophyletic clade with high statistical support, except for one specimen identified as Triglops quadricornis (an unaccepted name of My. quadricornis; see, e.g., [4]), which grouped with a sister clade including My. cf. scorpioides and Me. platycephalus [19]. Smith and Busby [21] (p. 345), referring to a previous study by Kontula and Väinölä [25], indicated that " . . . Myoxocephalus may need minor revision." Using the cytb, atp6, and atp8 mt genes, Kontula and Väinölä [25] found that the five marine Myoxocephalus species, included in their phylogenetic assessment, formed two main clades, while My. quadricornis was deeply nested within the Myoxocephalus genus. Based on these observations, the authors claimed that My. quadricornis (as well as its subspecies, the deepwater sculpin My. q. thompsonii) should be retained in Myoxocephalus without placing it in a separate genus, as it was proposed, for instance, by Neyelov [6]. Mecklenburg et al. [9] (Figure 4E, p. 118) and Mecklenburg and Steinke [28] (Figure 12, p. 176) obtained very similar tree topologies using the COI gene and also concluded that My. quadricornis should remain in Myoxocephalus.
The datasets of Kontula and Väinölä [25], Mecklenburg et al. [9], and Mecklenburg and Steinke [28] were limited only to the Myoxocephalus species and contained neither members of the genus Megalocottus nor other close sculpin genera (e.g., Microcottus, Argyrocottus, and Porocottus). The two divergent Myoxocephalus clades, revealed in their works, completely correspond to our Clusters 1 and 2 (Figures 1-3). However, due to insufficient sample of species, the authors could not match the clades to representatives of Myoxocephalus and Megalocottus and recognize Myoxocephalus as a paraphyletic group containing both Myoxocephalus and Megalocottus species. When a dataset is represented by an insufficient sample of species, the hypothesis of monophyly can be mistakenly supported (see, e.g., [73], p. 298). Consequently, the conclusion of Kontula and Väinölä [25], Mecklenburg et al. [9], and Mecklenburg and Steinke [28] is not supported by our extended dataset containing species of fourteen genera (including Myoxocephalus and Megalocottus) and obtained based on both complete mt genomes (16.68 kb) and a long fragment of rDNA (6.33 kb), which represent more reliable markers than the short mtDNA fragments (0.65-1.98 kb) used by those authors.
Recently, Mecklenburg et al. [74] (Vol. 1, p. 210) noted that "DNA barcodes place M. platycephalus as a sister species to Myoxocephalus quadricornis in a larger all-inclusive clade of Myoxocephalus species". The data obtained in the present study agrees with the first part of the statement made by Mecklenburg et al. [74]. Indeed, we show that Me. platycephalus and My. quadricornis are sister species, but we also show that Myoxocephalus is not an "all-inclusive clade". Instead, it consists of two evolutionary distinct lineages at the intergeneric level of divergence (see Figures 1-3), which is clearly evidenced by a comparative analysis of interspecific and intergeneric distances for Myoxocephalus, Megalocottus, and other sculpins (see the Results section). No such analyses were reported in the works of Kontula and Väinölä [25], Mecklenburg et al. [9,74], and Mecklenburg and Steinke [28].
The observed discordance may be explained in part by incorrect taxonomic identification, as was noted by McCusker et al. [29]. For instance, Smith and Busby [21] reported that Myoxocephalus was paraphyletic relative to Microcottus. We reanalyzed Smith and Busby's data and found that the Mi. sellaris specimen was misidentified and actually represented My. scorpius. The very low distance (0.86%) between the My. scorpius CytB nucleotide sequence (GenBank accession number MK321578) and the corresponding sequence (KM057906) of Smith and Busby's "Microcottus" clearly indicates that these authors misidentified Mi. sellaris. The two other sequences of Mi. sellaris obtained independently by Togashi [75] (LC125750) and Balakirev, Kravchenko, and Semenchenko (present study) show a very substantial divergence (around 11%) between Mi. sellaris and My. scorpius, which explains the clear resolution of these species (Figure 1). Thus, the observation of Smith and Busby [21] that Microcottus is nested within Myoxocephalus is erroneous, being based on incorrect species identification (other examples of incorrect taxonomic identification see, e.g., in [28,33]). Another problem potentially causing different tree topologies arises from DNA barcoding, which is still used and recommended for fish identifications [9,[27][28][29]. However, the DNA barcoding approach is frequently not enough to obtain reliable resolution. This problem is highly debated (for recent reviews see, e.g., [76,77]).
In accordance with the principle of priority in the International Code of Zoological Nomenclature (ICZN; Article 23) [78], Neyelov [6] proposed to place My. quadricornis in a separate monotypic genus, Triglopsis Girard 1851 (which is currently an unaccepted junior synonym of the genus Myoxocephalus Tilesius 1811 [2][3][4]), based mostly on the seismosensory system topography and other morphological characters. In this case, Me. platycephalus should also be transferred to the genus Triglopsis due to the genetic data obtained in the present work and the principle of priority, because Triglopsis Girard 1851 precedes Megalocottus Gill 1861. However, this hypothetic transfer contradicts the classification of Neyelov [6], who, in spite of the close relationship between My. quadricornis and Me. platycephalus [5,6,8,11,12], did not include both species in the same genus Triglopsis [6].
The type species of the genus Triglopsis is the deepwater sculpin T. thompsonii Girard 1851 (currently a subspecies of My. quadricornis [25]) inhabiting the North American continental lakes. This sculpin is a highly specialized form adapted to freshwater habitats. It originated from an Arctic marine lineage of the fourhorn sculpin My. quadricornis that was driven south to freshwater habitats by early glacial advances within the Early to Middle Pleistocene or earlier [79,80]. The divergence time between T. q. thompsonii and My. quadricornis is estimated at around one million years (Myr) ago [25,81] based on the COI, cytb, atp6, and atp8 mt genes. In turn, the Arctic My. quadricornis lineage diverged from the ancient Myoxocephalus forms from the Pacific, including closely related Megalocottus lineage, in the late Pliocene (see [5,6,14,82] and references therein), which agrees well with the estimated divergence time (≈7.9 Myr ago) between the Pacific and Arctic-Atlantic Myoxocephalus species [81].
In accordance with the principle of priority in the International Code of Zoological Nomenclature (ICZN; Article 23) [78], both species, My. quadricornis and Me. platycephalus, should be placed in the genus Triglopsis Girard 1851. However, this would contradict the basic assumption of phylogenetic taxonomy [73,83,84], because Triglopsis represents a derived form relative to the ancestral Myoxocephalus and Megalocottus forms. Thus, at the moment, we find it unreasonable to resurrect the genus Triglopsis for both My. quadricornis and Me. Platycephalus; instead, we consider the transfer of My. quadricornis to the genus Megalocottus as a feasible alternative, which is consistent with the evolutionary history of both Pacific (by origin) forms [5,6,14,82].
Taking in account the above circumstances, it would be premature to provide the diagnosis, description, and classification of My. quadricornis within the traditional Linnaean system. The compositions of the genera Myoxocephalus and Megalocottus inferred in the present work are delimited based on a phylogenetic approach according to the concept of phylogenetic systematics [73,84]. Further studies using complete mt genomes and nuclear genes, along with a more detailed morphological analysis and representative sample of species, are necessary to build a natural classification of these sculpin fishes that would satisfy the requirements of both the traditional Linnaean system and the phylogenetic systematics.

Conclusions
The present data, obtained through the analysis of pairwise genetic distances and phylogenetic reconstructions with high bootstrap support, indicates an intergeneric level of divergence between Megalocottus, Myoxocephalus, Microcottus, Argyrocottus, and Porocottus. It also shows some obvious inconsistencies in the relationships among the sculpin fishes belonging to the genera Myoxocephalus and Megalocottus (Figures 1-3), which has always been a controversial and long-debated issue [6,[9][10][11]28,71,74]. The multi-gene approach provides evidence for a close affinity of My. quadricornis with the species of the genus Megalocottus (but not with Myoxocephalus) and strongly suggests My. quadricornis to be placed in the genus Megalocottus.
We believe that the proposed taxonomic realignment based upon both mitochondrial and nuclear genes is robust. However, we also expect that the genera Myoxocephalus and Megalocottus, as well as other cottids, will continue to be subject to taxonomic changes in the future. In particular, the GenBank COI sequences suggest that three other Myoxocephalus species (besides My. quadricornis), My. aenaeus, My. octodecemspinosus, and My. scorpioides, may also belong to the genus Megalocottus (and, consequently, be removed from the genus Myoxocephalus; see above). To resolve this issue, a large-scale genetic analysis including complete mt genomes and nuclear genes is necessary. The next-generation sequencing (NGS) provides extensive full nuclear genome data, which is highly efficient for identifying speciation events and taxonomic relationships among non-model fish species. It turned out that a few candidate adaptive genes may demonstrate high divergence and create a barrier for the gene flow leading to reproductive isolation between species, which may be identical for the rest part of the nuclear genome [85][86][87][88]. A targeted NGS approach makes it possible to identify a suite of loci that can be used in future research to investigate the genomic basis of adaptation, differentiation, speciation, and phylogenetic relationships between sculpins, the most challenging group of fish for genetic and taxonomic studies.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4425/11/9/1071/s1, Table S1: Sites of specimen collection for the study, Table S2: Species names and GenBank accession numbers, Table S3: The best model of substitution under the maximum likelihood criterion, Text S1: Values of log likelihood, standard deviations of split frequencies, and convergence diagnostic for the rDNA and COI genes.