Demersal fish diversity and molecular taxonomy in the Bering Sea and Chukchi Sea

DNA barcoding by sequencing a standard region of cytochrome c oxidase subunit I (COI) provides an accurate, rapid method for identifying different species. In this study, we provide a molecular taxonomic assessment of demersal fishes in the Bering Sea and Chukchi Sea based on DNA barcoding, and a total of 123 mitochondrial COI partial fragments with a length of 652 bp were obtained. The consensus among all sequences was determined by alignment via a BLAST search in GenBank. Phylogenetic relationships were reconstructed based on neighbor-joining trees and barcoding gaps. The 39 species investigated in this analysis were distributed among 10 families. Five families within Scorpaeniformes including 19 species accounted for almost half of the species. The next largest group was Perciformes, with 9 species, followed by Pleuronectiformes and Gadiformes, with 5 species each, and the smallest number of species belonged to Rajiformes. At the family level, Cottidae was the largest family, followed by Zoarcidae, accounting for 8 species. The other eight families—Gadidae, Pleuronectidae, Psychrolutidae, Agonidae, Liparidae, Ammodytidae, Hexagrammidae, and Rajidae—accounted for a smaller proportion of species. In brief, our study shows that DNA barcodes are an effective tool for studying fish diversity and taxonomy in the Bering Sea and Chukchi Sea. The contribution of DNA barcoding to identifying Arctic fish species may benefit further Arctic fish studies on biodiversity, biogeography and conservation in the future.


Introduction
The correct identi cation of species is a prerequisite for studying sh diversity. Traditional morphologybased identi cation systems rely mostly on expert experience and the integrity of samples (Li et al. 2017).  Ward et al. 2005). This technology is free from excessive dependence on experience and can allow the automation and standardization of specimen identi cation to be realized. It provides a powerful supplement to traditional taxonomy and species identi cation methods. DNA barcodes can be used not only to identify whole sh but also to identify fry, roe, sh meat, sh ns, sh products or other body fragments that are di cult to identify based on morphology (Smith et al. 2008;Ward et al. 2005). Therefore, the use of DNA barcodes as an accurate and effective method of species identi cation is currently favored by an increasing number of researchers. Recent studies have indicated that this technology is highly reliable and e cient in many sh groups, including freshwater shes (Keskin et  Moreover, it is widely used in a variety of elds, such as biodiversity assessment, sh larva identi cation and shery management (Gao 2015;Panprommin et al. 2020).
The Bering Sea is located at the northernmost tip of the Paci c Ocean, while the Chukchi Sea is the marginal sea of the Arctic Ocean. The two seas are connected through the narrow Bering Strait. The seasonal ice-covered Bering and Chukchi Sea shelves are among the largest continental shelves in the world. These high-latitude shelf systems are highly productive during both the ice melt and open-water periods (Huntington et al. 2020). As seawater warms and the extent of sea ice declines, the vulnerability of the ecosystem to environmental change is thought to be high (Grebmeier et al. 2006a DNA extraction, ampli cation and sequencing A total of 123 genomic DNA samples were extracted, including 12 from the 6th Arctic expedition, 25 from the 8th Arctic expedition, and 86 from the 9th Arctic expedition. Back muscle tissue of each sh was obtained and preserved in a centrifuge tube with 95% ethanol. A TransGen kit (Easy Pure Marine Animal Genomic DNA Kit) was used to extract the genomic DNA of the Arctic sh, which was then stored at 4°C for later use. The primers used to amplify the CO gene fragment were F1:5′-TCAACCAACCACAAAGACATTGGCAC-3′ and R1:5′-TAGACTTCTGGGTGGCCAAAGAATCA-3′. The PCR system had a volume of 25 μL, containing 2.5 μL of 10× PCR buffer (including Mg 2+ ), 2 μL of dNTPs (2 mmol•L -1 ), 1 μL of each primer, 0.25 μL of Taq DNA polymerase, 1 μL of the extracted DNA, and deionized water to the nal of 25 μL. The thermal cycling program included an initial denaturation step of 4 min at 95°C followed by 30 cycles of 0.5 min at 94°C, annealing for 0.5 min at 52°C, and extension for 0.5 min at 72°C, with a nal step of 10 min at 72°C. Negative controls were included in all ampli cation reactions to con rm the absence of contaminants. The PCR products were visualized on 1.0% agarose stained with gel green (Biotium, Hayward, CA, USA), and successful ampli cation products were sent to Personalbio for puri cation and sequencing.

Data analysis
The original data obtained by sequencing were manually compared with the corresponding sequencing peak map to check for errors to ensure the accuracy of the data. The DNASTAR Lasergene software package was used to edit and align the sequences. All high-quality sequences were compared with the NCBI BLAST program to determine the species identity of the samples. Sequence similarity greater than 98% was the criterion for identi cation at the species level, and a similarity lower than 98% was used for identi cation at the genus level (Wong and Hanner 2008). Neighbor-joining (NJ) analysis implemented in MEGA 7.0 based on the K2P model with 1000 bootstrap replicates was employed to both calculate the genetic distances and examine the relationships among taxa.

Morphological analysis
A total of 123 specimens were collected during three CHINAREs. Most of them were adults and could be directly distinguished. However, there were also some juvenile and incomplete specimens, which were di cult to identify on the basis of morphological characteristics. These specimens were identi ed as Limanda sp., Hippoglossoides sp., Lycodes sp., Ammodytes sp., Hemilepidotus sp. and Liparis sp., etc.

Page 5/14
A total of 123 mitochondrial CO gene DNA fragments were successfully ampli ed using primers. No stop codons, deletions or insertions were observed in any of the sequences after alignment. The length of the ampli ed CO gene was 652 bp. The number of haplotypes identi ed in each species ranged from 1 to 6.

Species identi cation by phylogenetic analysis of CO sequences
The phylogenetic tree constructed by the NJ method is shown in Figure 1 (different-colored bands indicate different families). The same morphological species of sh formed cohesive units. All highquality sequences were identi ed by BLAST searches and comparisons in GenBank, and the similarity was higher than 98% (Table 1). A total of 39 sh species belonging to 5 orders, 10 families and 23 genera were identi ed through DNA barcoding analysis in this survey. Among these species, 19 species of Scorpaeniformes accounted for 48.72% of the total species. Additionally, 9 species of Perciformes accounted for 23.08% of the total number of species, and 5 species from each of Pleuronectiformes and Gadiformes accounted for 12.82% of the total number of species. The smallest number of species was found in Rajiformes, which included only one species (Figure 2). At the family level, the number of Cottidae species was largest, at 9, accounting for 23.08% of the total number of species, followed by Zoarcidae, with 8 species, accounting for 20.51%. The other eight families-Gadidae, Pleuronectidae, Psychrolutidae, Agonidae, Liparidae, Ammodytidae, Hexagrammidae, and Rajidae-accounted for smaller proportions. At the genus level, the number of species from the genus Lycodes was greatest, at 8. Based on the NJ tree, all species from the same family were clustered together, indicating that the families were all monophyletic except for Cottidae, in which Hemilepidotus papilio was sister to other genera ( Figure 1); thus, the NJ analysis recovered the family Cottidae as paraphyletic.

Genetic distance and barcoding gaps
The intraspeci c distances ranged from 0% to 0.35%, and the minimum interspeci c distances of the species were greater than 2% except for Liparis tunicatus vs L. fabricii (1.43%), Hippoglossoides elassodon vs H. robustus (0.62%), H. elassodon vs H. dubius (0.54%), H. robustus vs H. dubius (0.54%), and Icelus spatula vs I. spiniger (0.69%). Nevertheless, the minimum interspeci c distance of all species was still greater than the maximum intraspeci c distances. Thus, it was obvious that there were barcode gaps in the genetic distance between intraspeci c distances and interspeci c distance (Figure 3).

Discussion
Correct species identi cation is the foundation for revealing sh diversity. Traditional morphological identi cation methods require the experience of high-level classi cation experts and sample preservation integrity. Among the samples utilized in this study, there were some juveniles and damaged individuals, The existence of barcode gaps increases the effectiveness of DNA barcodes for identifying species. In this study, all of the obtained minimum interspeci c distances were greater than the maximum intraspeci c genetic distances. A value of 2% has been suggested as a threshold value between species and genus divergence (Ward 2009 The phylogenetic tree constructed based on the obtained sequences showed cluster formation; clustering in the phylogenetic tree can help detect problems and is a valuable tool, especially for closely related species without obvious morphological differences (Dettai et al. 2011). Although barcode analysis is mostly used to delimit species boundaries, there are obvious phylogenetic signals within CO sequence information (Hebert et al. 2003a;Ward et al. 2005). In the phylogenetic tree based on the NJ method obtained in this study, different individuals of each species were clustered together. However, it should be noted that at the family level, H. papilio of Cottidae and Agonidae formed a separate branch. This may be the reason why Mecklenburg et al. (Mecklenburg et al. 2010) indicated that the internal relationships of the Cottidae are obscure and not well de ned. However, H. papilio was represented in the specimens by only one specimen, and additional specimens will certainly be necessary to further clarify the relationship between H. papilio and Cottidae.

Conclusions
This study shows that DNA barcoding is an accurate and e cient method of species identi cation. A total of 123 sh collected from the Bering Sea and the Chukchi Sea were identi ed by DNA barcoding.
Thirty-nine species from ten families were characterized; all species were identi ed correctly. We also observed low interspeci c divergence (< 2%), probably associated with recent speciation. It is recommended that other molecular markers be included to develop unique DNA barcodes that are suitable for Arctic sh. In follow-up studies, it is necessary to combine morphology-based identi cation systems with DNA barcoding to identify species because morphological identi cation alone may not be su ciently robust. In addition, our work provides important information for further studies regarding the biodiversity, biogeography and conservation of Arctic shes.
Declarations Figure 1 Neighbour-joining (NJ) tree constructed using CO gene sequence. Bootstrap values higher than 50 are indicated along the branches.

Figure 2
The percentage of sh species collected from the Bering Sea and Chukchi Sea comprising different orders.