2.1 Sample collection

Since 2011, we have collected a total of 616 specimens of pufferfish from the coastal waters of China (Table 1). Samples collection and use comply with the relevant regulations of the Institutional Animal Care and Use Committee.

After preliminary classification according to morphological characteristics, they were preserved in 95% ethanol for subsequent DNA barcoding identification. All voucher specimens are deposited in the National Marine Fishery Biological Germplasm Resource Bank.

2.2 Morphological identification

Firstly, the body length, weight, and other measurable traits of the specimen were measured. Further species identification was conducted based on the morphological characteristics of pufferfish described by Chen and Zhang (2015) and Su and Li (2002), including body size, number of fins, stripes, spots, body color, mouth shape, etc.

2.3 DNA extraction and quality assessment

DNA was extracted from the muscle of fish by using TIANamp Marine Animals DNA Kit (Tiangen Biotech (Beijing) Co., Ltd.). For detailed steps of DNA extraction, refer to the kit manual. DNA samples were stored at −20°C.

Nano-300 micro-spectrophotometer (Allsheng) was used to detect the concentration and purity of DNA samples, and the quality of the DNA samples was detected by 1% agarose gel.

2.4 PCR amplification and DNA sequencing

PCR amplification was in total volume of 25 μL, containing 12.5 μL Taq Mixture (2×, vazyme), 2 μL DNA template, 1 μL of each primer (10 μM), and 8.5 μL ddH2O. Thermo cycling was carried out as follows: 94°C for 5 min, 94°C for 30 s, 55°C for 30 s, 72°C for 1 min, cycle 35 times from step 2 to step 4, and finally 72°C for 10 min. PCR products were detected by 1% agarose gel, and the qualified samples were sent to BGI in Qingdao for direct bidirectional sequencing.

2.5 Data analysis

The raw sequences were assembled by seqman of Lasergene software package (Burland, 2000; Swindell & Plasterer, 1997) and aligned by Clustal W in BioEdit version 7.0.9 (Hall, 1999). After correction and alignment, the effective sequence length was 687 bp. The haplotypes were analyzed by DNASP version 5.10.01 (Librado & Rozas, 2009). Each haplotype sequence was compared for similarity in NCBI (https://www.ncbi.nlm.nih.gov/) and BOLD (http://www.barcodinglife.org/). The top matches showing ≥98% similarity were used as the preliminary identification results. Then based on Kimura-2-parameter (K2P) mode (Kimura, 1980), the genetic distance of different taxonomic levels of Tetraodontiformes was calculated.

Bothus myriaster (NC_030365) and Glyptocephalus stelleri (NC_060723) were selected as the outgroups, based on the haplotypes of COI gene of Tetraodontiformes, the Neighbor-joining tree (NJ tree) of Tetraodontiformes was constructed in Mega 7.0 (Kumar et al., 2016).

3.1 Morphological identification

According to the morphological characteristics, 616 samples were identified as 50 species, belonging to 23 genera, 6 families, and 4 Suborder (Table 1). Among them, 612 samples can be identified to species. The remaining four samples are larval and juvenile fish, which can only be identified to genus due to their atypical morphological characteristics.

Among the 23 genera of Tetraodontiformes collected in this study, the various species of Takifugu genus have very similar morphologies and are difficult to distinguish with the naked eye. For example, the morphologic characteristics of T. alboplumbeus and T. niphobles are very similar. It is generally believed that T. niphobles has obvious chest spots. The chest spots of T. alboplumbeus are not obvious, but there are 4–6 dark bands on its back. Out of 616 specimens, 21 were identified as T. niphobles. And 10 specimens have the characteristics of T. alboplumbeus, but they are not exactly the same as T. alboplumbeus because they only present two dark bands. Furthermore, it is also difficult to distinguish T. rubripes, T. pseudommus, and T. chinensis, some studies suggest T. rubripes has irregular round black spots and white patterns in front of the caudal fin on the sides of the body, T. chinensis has no such black marks while T. pseudommus has white spots scattered on a black background on the dorsal and lateral sides of the body (Baek et al., 2018; Reza et al., 2008), as shown in Figure 1. Based on these morphological characteristics, 8 specimens were identified as T. chinensis, 4 as T. pseudommus, and 27 as T. rubripes.

Additionally, Lagocephalus wheeleri (Abe, Tabeta & Kitahama, 1984) and Lagocephalus spadiceus (Richardson, 1845) can also be distinguished by morphological characteristics. The former has elliptical dorsal spinule patch, while the latter has a rhomboidal patch with a posterior extension. In this study, 82 specimens were identified as L. wheeleri, 10 were identified as L. spadiceus, and 4 were found to have the intermediate characteristics of these two species, covering discontinuous patch. The above three morphological differences can be seen in Figure 2.

3.2 DNA barcoding of the Tetraodontiformes

3.2.4 Phylogenetic relationships of Tetraodontiformes

The Neighbor-joining tree was constructed for the order Tetraodontiformes, based on the haplotypes of COI gene. The results indicate that the six families of this order in this study are divided into two branches, namely the Diodontidae and Tetraodontidae families, which converge into one branch, while the Monacanthidae, Balistidae, Ostraciidae, and Triacanthidae families converge into another branch. As expected, in the Neighbor-joining tree, T. rubripes, T. pseudommus, and T. chinensis clustered into one clade, while T. alboplumbeus and T. niphobles clustered into one clade, and L. wheeleri and L. spadiceus also clustered into one clade. This once again confirms the previously proposed hypothesis of synonyms. In addition, we found that although the interspecific genetic distance of the genus Takifugu is small, some even <0.02, except for the few species that may be synonymous with the same species mentioned above, other species of this genus have clustered into different separate clades (as shown in Figure 3).

4.1 DNA barcoding threshold of Tetraodontiformes

DNA barcoding is used to identify species at molecular level, which is independent of time, region, and individual morphology (Hebert et al., 2003). This study shows that DNA barcoding can quickly and accurately identify the species of Tetraodontiformes, without being affected by key factors such as the integrity and development stage of the sample and the taxonomic expertise of the appraiser. For those closely related species, although the COI gene sequence may have high homology, they can still be distinguished by phylogenetic tree. For example, the sequence similarity between T. flavidus (Hap 100) and T. bimaculatus (Hap 91) has reached 98.7% (678 conserved sites in the 687 bp COI gene sequence), but the two species clustered into two independent clades in the NJ tree, thus enabling effective species identification.

This study indicates that species in Takifugu genus do have closer phylogenetic relationships than other reported species. The ratio of the average interspecific to intraspecific genetic distance is only 6.21 times. Among them, the largest interspecific genetic distance is between T. oblongus and T. stictonotus, with a value of 0.045; the smallest ones are T. bimaculatus and T. flavidus, with only 0.013; there are also multiple species with interspecific genetic distances <0.02 (as shown in Table 2). From this, it can be seen that the DNA barcode classification standard proposed by Hebert et al. (2004), with an interspecific genetic distance threshold of 0.02 and a ratio of interspecific to intra-specific genetic distances greater than 10, is not entirely applicable to the species of Takifugu genus.

4.2 Quality control of sequences downloaded from databases

NCBI's GenBank and BOLD are the two most abundant databases of DNA barcode resources. In recent years, the number of DNA barcode sequences submitted to NCBI and BOLD has increased explosively. When we use these two databases for DNA barcode sequence alignment analysis, we sometimes encounter inaccurate matching results. It has been reported that there are many errors in the sequences of the NCBI database (Liu et al., 2020; Shen et al., 2013). Some studies believe that the BOLD database has conducted a more rigorous review and screening on the submitted sequences, so the data in this database is more reliable (Macher et al., 2017; Wang et al., 2009). In fact, there are also wrong sequences in the BOLD database (Lis et al., 2016).

In this study, all sequences were run BLAST in NCBI and BOLD databases. The Blast results showed that most of the query sequences were able to match the target species with the highest identities in the database, which was consistent with the morphological identification results. But there were also some unexpected matching results of the query sequences, with multiple highly similar target species, some of which were consistent with morphological identification, while others were inconsistent. For instance, the query sequence of the T. porphyreus specimen was matched to the T. porphyreus, and in addition, two COI gene sequences of T. porphyreus (KT951818, KT951819) in the database were also matched, with identities of 100%. But apart from these two sequences, it could not match with other sequences of T. porphyreus in the database. Therefore, the information on those two sequences in the database is inaccurate. In addition, the query sequences of T. chinensis, T. pseudommus, and T. rubripes showed 99.56% identities to the two mitochondrial genome sequences (NC_024199, KJ562276) of T. flavidus in BOLD and NCBI. After checking, it was found that the information of these two sequences were also incorrect. Therefore, there are some incorrect sequences of Tetraodontiformes in the NCBI and BOLD databases.

The unreliable data in the databases will directly lead to the misidentification of species. To reduce the interference of the wrong sequence in the database on the analysis results, it is suggested that the relevant sequences in the database should be strictly screened when using DNA barcoding technology to identify species. For instance, when screening DNA barcode sequences, we can select sequences uploaded by at least three different authors, or select the sequences that have already been published and provide morphological characteristics of the species. If the information in NCBI and BOLD databases is found to be inaccurate when the voucher specimens with complete morphological characteristics are available, the database or data submitter shall be contacted in time to correct the sequence information. In this study, DNA barcoding technology and morphological methods were used synchronously to identify species of Tetraodontiformes, which could ensure the accuracy of species identification. Therefore, the DNA barcode sequences of Tetraodontiformes provided in this study can be used as reference sequences for the identification of the species in this order.

4.3 Synonym phenomenon in the order of Tetraodontiformes

Early taxonomic research was mainly based on morphological characteristics. Due to the fact that the morphological characteristics of fish are not only susceptible to subjective factors but also undergo significant change at different developmental stages and environmental conditions, coupled with delayed information exchange among taxonomists in different regions, the same species may have two or more different Latin names, which is also called synonym.

This study shows that there are a lot of synonyms in Tetraodontiformes. For example, M. leiurus is the synonym of P. leiurus, but the samples morphologically identified as P. leiurus were identified as M. leiurus in NCBI, and there is no information about P. leiurus can be found in NCBI. What's more, the samples which morphologically identified as D. nigroviridis, in NCBI and BOLD databases, were identified as T. nigroviridis. In fact, in the FishBase T. nigroviridis is the synonym of D. nigroviridis. The variety of species names in the database will affect the molecular identification results of species. We consider that the unsynchronized information in the database is also one of the reasons for the confusion of species identification of Tetraodontiformes.

In addition, in the FishBase, both Lagocephalus wheeleri (Abe, Tabeta & Kitahama, 1984) and Lagocephalus spadiceus (Richardson, 1845) are effective species, and they can be distinguished from each other in terms of morphological characteristics. For example, the former has elliptical dorsal spinule patch, while the latter has a rhomboidal patch with a posterior extension. However, some studies believe that these morphological features cannot be used to distinguish the two, and L. wheeleri may be a synonym for L. spadiceus (Matsuura, 2010; Sakai et al., 2021). In this study, the results of haplotype analysis showed that they had the same haplotype and clustered into the same clade in neighbor-joining tree. Additionally, their interspecific genetic distance was only 0.003. This shows, the molecular classification results support that L. wheeleri is synonymous with L. spadiceus.

Moreover, there are many reports believe that T. rubripes, T. chinensis and T. pseudommus may be different phenotypes of the same species, that is, T. chinensis and T. pseudommus are the synonyms of T. rubripes (Cui et al., 2005; Liu et al., 1999; Park et al., 2020; Reza et al., 2008; Song et al., 2001). However, all three species are valid in Fishbase and can be distinguished by morphological characteristics. T. rubripes has irregular round black spots and white patterns in front of the caudal fin on the sides of the body, T. chinensis has no such black marks, while T. pseudommus has white spots scattered on a black background on the dorsal and lateral sides of the body (Baek et al., 2018; Reza et al., 2008). In this study, the initial morphological identification results showed that eight samples were identified as T. chinensis, 4 as T. pseudommus, and 27 as T. rubripes. However, further research has found that samples identified by morphology as the three species had the same haplotype, and the interspecific genetic distance among them ranged from 0.001 to 0.002. In additional, in the NJ tree, they clustered into one clade. To sum up, the T. chinensis and T. pseudommus are the synonyms of T. rubripes, and T. rubripes is the valid name.

And the morphologic features of T. alboplumbeus and T. niphobles are so similar that they are difficult to distinguish. It is generally believed that T. niphobles have obvious chest spots, while the chest spots of T. alboplumbeus are not obvious, and there are several dark bands on the back of T. alboplumbeus. In this study, the DNA barcode identification results also showed that the two species may be synonyms. However, since the 11 specimens identified as T. alboplumbeus do not completely have the morphological characteristics of T. alboplumbeus, it is not certain that these specimens are hybrids of T. alboplumbeus and T. niphobles. Or that T. alboplumbeus and T. niphobles are the same species, but they will undergo a morphological transformation process. Matsuura (2017) regarded T. niphobles and T. alboplumbeus as synonyms based on the same color patterns and no differences in morphological characteristics between the type specimens. Some researchers support this view as well (Okabe et al., 2019), and these two species are also considered to be synonyms in NCBI. However, in the FishBase, it is considered that both T. alboplumbeus and T. niphobles are effective species. Zhang and He (2008) showed that 12 s and cytb genes could separate T. alboplumbeus and T. niphobles, but the study did not describe the morphological characteristics of the two species. Zhou et al. (2020) showed that the mitochondrial genome can separate T. alboplumbeus and T. niphobles effectively. However, the study did not describe the morphological characteristics of the two species as well. Therefore, whether T. alboplumbeus and T. niphobles are synonyms has not been determined.

4.4 Reconstruction of Tetraodontiformes phylogenetic relationships

Many researchers have studied the relationship between families of Tetraodontiformes based on different methods (Alfaro & Brock, 2007; Breder Jr & Clark, 1947; Holcroft, 2005; Leis, 1984; Rosen, 1984; Santini & Tyler, 2010; Tyler & Sorbini, 1996; Winterbottom, 1974; Yamanoue et al., 2007, 2008). In their study, there were more or less differences in the phylogenetic tree of Tetraodontiformes. The commonality is that they unanimously agreed that Balistidae and Monacanthidae were closely related, and Diodontidae and Tetraodontidae were closely related as well.

In this study, the Neighbor-joining tree based on the haplotype of COI gene of Tetraodontiformes shows that the species of six families are split into two branches. And Diodontidae and Tetraodontidae clustered into a branch, Monacanthidae, Balistidae, Ostraciidae, and Triacanthidae clustered into another branch, as shown in Figure 3. The result that Diodontidae and Tetraodontidae are clustered in a separate branch is consistent with the research results of others. However, unlike other studies, Balistidae and Monacanthidae did not cluster into a branch in this study. The reason for this result may be that there are not enough samples collected in this study. For example, only one sample of Triacanthidae was collected. It is also possible that the COI gene sequence is not suitable for the systematic relationship study of families and higher taxonomic levels.

In the Neighbor-joining tree based on the haplotype of COI gene of Tetraodontiformes, the species of the same genus clustered together, and the same species are clustered into a clade, which is basically consistent with the results of morphological identification. Although the interspecific genetic distance of some species of Takifugu is <0.02 (as shown in Table 2), in the Neighbor-joining tree, except for T. pseudommus, T. rubripes, and T. chinensis, each species can gather into a separate clade. It indicates that the COI gene sequence is suitable for the classification and identification of genera and lower taxonomic levels.