Complete samples of puffer fish have been identified by morphology. All puffer fish samples were provided by the Fisheries Research Institute of Fujian, including Lagocephalus inermis, Lagocephalus lagocephalus, Lagocephalus gloveri, Lagocephalus lunaris, Lagocephalus spadiceus, Takifugu vermicularis, Takifugu fasciatus, Takifugu xanthopterus, Takifugu bimaculatus, Takifugu flavidus, Takifugu rubripes, Takifugu oblongus, Takifugu alboplumbeus. Other fish samples used for the specificity test have also been identified in morphology, including Limanda aspera, Verasper variegatus, Verasper moseri, Platichthys stellatus, Paralichthys lethostigma, Oncorhynchus gorbuscha, Gadus macrocephalus, Sebastes schlegelii, Trachurus japonicus were obtained from Dalian Tianzheng Industrial Co., Ltd (Dalian, China). All fish materials information was listed in Supplementary Tables 1, 2. DNA oligonucleotides were synthesized by TaKaRa (Dalian, China) and set out in Supplementary Table 3. All sequences were purified by polyacrylamide gel electrophoresis (PAGE).

Fish meat samples were pulverized using a high-speed tissue masher (34BL99, Waring Blender dynamics Corp., New Hartford, CT, USA). Ground sample (200 mg) was taken for DNA extraction. DNA Extraction Kit (Code No. 9766, TaKaRa Co., Ltd., Dalian, China) was used according to the manufacturer’s recommendations. The operations were as follows: 10 mg sample materials were taken for low-temperature grinding by adding liquid nitrogen, then added 200 μL PBS buffer. Briefly, 200 μL VGB buffer, 20 μL proteinase K, and 1.0 μL carrier RNA were added and fully mixed in a 56°C water bath for 10 min. Then 200 μL 96–100% ethanol was added, and fully mixed. Placed the spin column on the collection tube, transferred the solution to the spin column, centrifuge at 12,000 × g for 2 min, and discarded the filtrate. Then, 500 μL RWA buffer was added to the spin column, 12,000 × g centrifuge for 1 min, and discarded the filtrate. And then, 700 μL buffer RWB was added to the spin column, 12,000 × g centrifuge for 1 min, and discard the filtrate. Repeat the previous step. Placed the spin column on the collection tube, and 12,000 × g centrifuge for 2 min. Placed the spin column in a new 1.5 ml RNase-free collection tube, and added 30–50 μL RNase-free dH₂O, standing at room temperature for 5 min. Centrifugation at 12,000 rpm for 5 min. The products were dissolved in H₂O.

Puffer fish COI gene sequences of mitochondrial were retrieved from the official National Center for Biotechnology Information (NCBI) database GenBank.¹ And then, these sequences as the template used for Blast analysis. In addition, the specificity of the primer and probes was tested by Blast. MEGA 4.0 software (27) was used to perform sequence alignment to screen high variability DNA fragments, examine the specificity of primers and probes, and guarantee that the primers and probes cannot theoretically amplify genes from related species. All sequence accession numbers were listed in Supplementary Table 2.

Arrange the COI sequence of the target species and the DNA sequence of the most relevant species (such as the common puffer fish species of the genus Fugu) and screen for the fragments with the greatest variability.

Considering the impact of food processing on DNA quality, the amplification efficiency of real-time qPCR analysis was improved by designing primer pairs to amplify relatively short DNA fragments. The nucleotide sequences chosen for primer design were introduced into the program “Oligocalc” (28), and the length was optimized for the resulting “salt-adjusted” annealing temperature. Then, the annealing temperatures calculated by “Oligocalc” applying the “salt-adjusted” algorithm were used as starting values for the qPCR. Four qPCR methods were based on the TaqMan probe, modified with the reporter fluorophore, 6-carboxyfluorescein (FAM), and quencher fluorophore black hole quencher (BHQ_1) at 5’ and 3’ end, respective. L. inermisand and L. lagocephalus amplification fragment sizes were 196 bp, L. gloveri was 174 bp, L. lunaris was 150 bp, and L. spadiceus was 173 bp. Finally, 18SrRNA was used as the control gene to design primer and probe for detecting DNA of all sample materials to ensure no inhibitory contaminants. All primers and probes oligonucleotide sequences were listed in Supplementary Table 3 and synthesized by TaKaRa (Dalian, China).

Real-Time qPCR analysis was carried out in QuantStudio 7 Real-time fluorescent quantitative PCR system (Applied Biosystems, VA, USA). Real-time qPCR reaction was carried in a volume of 25 μL containing 16 μL Probe qPCR Mix (Code No. 391A, TaKaRa, Dalian, China), 1 μL forward primer (0.4 pmol/μL), 1 μL reverse primer (0.4 pmol/μL), 1 μL probe (0.4 pmol/μL), and 2 μL target DNA. The reaction blend was then subjected to 45 cycles at 95°C for 5 s and 60°C for 30 s, with fluorescence acquisition at each cycle. Each sample was analyzed three times.

The specificity and cross reactivity of the detection methods were evaluated by qPCR analysis. Undiluted sample material DNA obtained from 13 closely related different puffer fish species and 9 other unrelated fish species listed in Supplementary Table 3 was used. DNA analysis of each species was repeated no less than 3 times. All sample DNA used for the test was detected with primers and probes of internal reference 18 S rRNA to avoid inhibitory substance.

In order to calculate the amplification efficiency (E) of qPCR, 8 series of dilution levels were prepared using the sample DNA of Lagocephalus inermis, Lagocephalus lagocephalus, Lagocephalus gloveri, Lagocephalus lunaris, and Lagocephalus spadiceus, including 100, 50, 20, 10, 5, 2, 1, and 0.1%. All samples were analyzed three times. Mean Ct values obtained for each point were plotted against the Log (DNA concentration (ng/μL), and a linear regression analysis was performed. Using the slope of the regression line, the qPCR efficiency was calculated using the equation E = 100 (10^–1/slope − 1) and expressed in percent. For each target, the slope of the regression curve should be between 3.9 and 2.9 corresponding to PCR efficiencies ranging from 80 to 120%. Additionally, the correlation coefficient R² of the curve is a measure of the linearity of the PCR reaction. The R² for each target should be greater than 0.98.

The LOD was experimentally determined according to accepted guidelines (28). DNA was extracted from Lagocephalus inermis, Lagocephalus lagocephalus, Lagocephalus gloveri, Lagocephalus lunaris, and Lagocephalus spadiceus, respectively, and diluted by Limanda aspera DNA extracted from slices of fish meat. Therefore, 8 series of dilution levels were prepared to simulate real samples for qPCR analysis and determine LOD, including: 100, 50, 20, 10, 5, 2, 1 and 0.1%. LOD₆ of the qPCR method represents six replicate analyses performed for each dilution point of serial dilution. At least three times must be performed under repeat conditions, yielding a total of 18 results per dilution point. The lowest dilution level at which all 18 replicates show a specific positive amplification was considered as the LOD₆. The analytical sensitivity of the qPCR method was present by LOD_95%. The LOD_95% refers to the use of the corresponding LOD₆ level, one higher dilution level, and one lower dilution level, and each level is tested 60 times. All 60 replicates showed specific positive amplification, and the lowest dilution level was considered as LOD_95% with a 95% confidence level. Statistical significance LOD_95% was calculated by Semi logarithmic regression analysis (PRISM, Graphpad Software Inc., San Diego, CA, United States), input of the corresponding number of sample materials, the number of repetitions, and the number of positive results in qPCR detection.

The robustness of puffer fish detection was checked by changing conditions of the qPCR reaction such as the qPCR instrument (CFX 96 Real-Time qPCR System, Bio-Rad Co., Ltd, Hercules, CA, USA), the concentration of primers and probes (±25%), and the annealing temperature S6. In each combination, a template in an amount of four times the LOD₆ was added to the assay, at least repeat three times in one run. Regression analysis was carried out with SPSS (statistical product and service solutions, IBM Inc.) software, and the Kruskal Wallis H test was used to evaluate the significance level of difference in results obtained by orthogonal design combination of each method. When the p > 0.05, there was no significant difference in results.

The DNA fragment of the mitochondrial Cytochrome Oxidase Sub-unit I (COI) gene was selected as the target of the developed real-time qPCR method for puffer fish species identification. The base sequence of the COI gene region has a large genetic variation among species, but a small genetic variation within species, was stable and has high identification ability, and has been widely used for fish species identification (29–31). We have searched almost all the genus Lagocephalus in NCBI, there are 18 accession numbers of COI gene sequences of Lagocephalus inermis and 9 accession numbers of COI gene sequences of Lagocephalus lagocephalus. The homology of COI genes between these two species is 100%. In addition, there are 15 accession numbers of Lagocephalus gloveri, 19 accession numbers of Lagocephalus lunaris, and 41 accession numbers of Lagocephalus spadiceu in NCBI. They have differences in nucleic acid sequences, which can realize the identification of each species. The DNA sequences of 5 species of puffer fish in the genus Lagocephalus and 8 species of puffer fish in the genus Takifugu were aligned (Figure 1), and primer sequences that could distinguish the DNA sequences of 5 species of puffer fish in the genus lepidocephalus from those of other species were searched. The sequences of five Lagocephalus and eight Takifugu were comparable to search primers that can distinguish the DNA sequences of five species of Lagocephalus, Takifugu and other puffer fish species (Figure 1). All the designed primers could theoretically exclude other species of puffer fish. However, it was worth noting that the COI sequence of Takifugu vermicularis with three accession numbers in NCBI has 100% homology with that of Lagocephalus inermis and Lagocephalus lagocephalus. It is impossible to distinguish these three species based on the COI sequence. Nevertheless, it has been revealed that the homology analysis results of the 18 S rRNA, Cytb, and COI DNA fragment sequences of Takifugu vermicularis all belong to the same group as the genus Lagocephalus, with a homology of 99–100% (32).

In order to determine the specificity of the qPCR method, the DNA of target sample materials (L. inermis, L. lagocephalus, L. gloveri, L. lunaris, and L. spadiceus) and non-target sample materials were analyzed. To exclude possible inhibitory effects in DNA preparations, samples were analyzed by qPCR using the primers and probe of internal reference of 18 SrRNA. All target and non-target fish species sample were successfully amplified on tested (Supplementary Table 4), confirming the suitability of sample DNA for qPCR assays. In the qPCR detection method of L. inermis and L. lagocephalus, the two species puffer fish could be detected at the same time, and had cross reactivity with the T. vermicularis, no false-positive signal was detected in other tested samples. The results of specificity test also verified that the sequence homology of the COI gene of L. inermis, L. lagocephalus, and T. vermicularis, indicating that the genus classification of T. vermicularis needs further exploration. In addition, in the respective qPCR detection methods of L. gloveri, L. lunaris and L. spadiceus, no cross reactivity was found in the DNA of non-target sample materials, and no false-positive signal was detected (Figure 2). These qPCR analysis results confirmed the accuracy and specificity of the detection method.

The qPCR efficiency of these four methods was analyzed by qPCR at eight consecutive dilution levels (m = 8) of DNA samples from five species puffer fish in the genus Lagocephalus including L. inermis, L. lagocephalus, L. gloveri, L. lunaris, and L. spadiceus. Each dilution level is tested at least three times. The threshold cycle value (Ct value) was compared with the DNA concentration (pg/μL) to draw a linear regression curve. According to the equation E = 100 (10^–1/slope − 1), the analysis efficiency is determined as the slope of the regression line, which shows a good linear relationship between Ct value and DNA concentration (Figure 3). The correlation coefficient R² of these qPCR methods is 0.9808–0.9903, the slope of the regression curve is −3.824 to −3.228, and the efficiency E is 82.60–104.07% (Supplementary Table 5). These results match the specifications of the common qPCR validation guidelines, with a required linearity (R²) should be ≥ 0.98, the slope of the regression curve should be between −3.9 and −2.9 corresponding to an efficiency (E) from 80 to 120%.

The sensitivity of the qPCR method was an important parameter that needs to be evaluated, especially considering the regulations that wild puffer fish were not allowed to eat cultured by unauthorized certification companies, and the detection of species that may contain low concentrations. In this study, the sensitivity of the qPCR LOD was expressed by LOD₆ and LOD_95%. It is determined by measuring the serial dilution level of DNA of five kinds of puffer fish samples in their respective detection. The LOD₆ of L. inermis and L. lagocephalus was 37.24 pg and 32.90 pg, and the LOD_95% was 40.83 pg (17.31–233.44 pg, 95% CI) and 45.64 pg (19.25–265.22 pg, 95% CI), respectively. For L. gloveri, L. lunaris, and L. spadiceus the LOD₆ was 35.99, 33.91, and 32.85 pg, and the LOD_95% was 34.79 pg (14.70–207.20 pg, 95% CI), 32.78 pg (13.85–195.23 pg, 95% CI), and 31.76 pg (13.41–189.13 pg, 95% CI), all of the five qPCR methods had highly sensitive (Supplementary Table 6 and Figure 4). This highly sensitive indicates that when the aquatic products contain 0.1% of a small amount of wild puffer fish materials without certification, they can be tracked reliably (Figure 5).

To evaluate the robustness of the qPCR methods, we used orthogonal design to slightly change the principal different experimental conditions, such as the qPCR instruments, qPCR reagents, primer, and probe concentrations, and slight deviations of PCR annealing temperature. We used a 5% DNA template sample to examine the impact of the aforementioned variables on the stability of the results. In the orthogonal design combination of each method, there was no significant difference in the Ct values of the five species of puffer fish in the genus Lagocephalus (Figure 6 and Supplementary Tables 7, 8). For L. inermis, the Ct value was 29.84 ± 0.43 (p = 0.692). For L. lagocephalus the Ct value was 30.23 ± 0.20 (p = 0.063). For L. gloveri the Ct value was 28.33 ± 0.34 (p = 0.433). For L. lunaris the Ct value was 28.59 ± 0.20 (p = 0.291). For L. spadiceus the Ct value was 28.61 ± 0.24 (p = 0.564). Data in above are mean ± s.d. (n = 3). Thus, it can be concluded that the statistical p-value >0.05 of each method’s Ct values, these qPCR methods were stable, and can be transferred to other laboratories and used in routine analysis.