We used 44 pomacentrid species of 12 genera that we collected in the shallow waters of coral reefs around Okinawa-jima Island of the Ryukyu Archipelago, Japan, for sequence analysis. We examined more than one specimen for each species whenever possible, considering possible intraspecific variations (Table 1). Before genetic examinations, we attempted accurate species identification, based on morphology for each specimen following (Nakabou 2013, Allen et al. 2015, Wibowo et al. 2017, Wibowo et al. 2018 and Tang et al. 2021). The pectoral fins on the right side were collected from the specimens and stored in 99% ethanol until use. The voucher specimens were fixed with 10% formalin and transferred to 99% ethanol and representative specimens were deposited in the Ichthyological Collection of the Okinawa Churaumi Aquarium (OCF-P10591 to OCF-P10681). The remaining specimens are stored in the authors' personal collection shown as POM in Table 1.

1Total DNA was extracted from the tissue samples using DNeasy Blood & Tissue Kit (Qiagen), according to the manufacturer’s instructions. A fragment of the first half of the mitochondrial 12S ribosomal RNA (12S) gene (approximately 750 bp), which covered the entire length of the MiFish region inside, was amplified by standard PCR using Takara Ex Taq (Takara). The primers used for PCR were L709 (5ʹ-TACACATGCAAGTCTCCGCA-3ʹ) and H1552 (5ʹ-ACTTACCGTGTTACGACTTGCCTC-3ʹ) (Inoue et al. 2000). The cycling profile in PCR amplification consisted of an initial denaturation step at 94℃ for 90 s followed by 30 cycles of denaturation at 94℃ for 30 s, annealing at 49℃ for 30 s, extension at 72℃ for 45 s and final extension at 72℃ for 5 min. The PCR products, after confirmation using 1% agarose gel electrophoresis, were purified by PEG precipitation and then cycle-sequenced using BigDye v. 3.1 (Applied Biosystems) with a cycling profile of initial denaturation at 96℃ for 60 s followed by 25 cycles of denaturation at 96℃ for 10 s, annealing at 50℃ for 5 s and final extension at 60℃ for 4 min. The direct sequence was commissioned to an external company (Macrogen Japan). The sequences we determined are available at the International Nucleotide Sequence Database (INSD) (accession numbers LC814855 to LC814984). In addition to the sequences we determined, homologous sequences of pomacentrid fish were obtained from INSD. We first obtained sequences of the first half of the mitochondrial 12S gene (725–773 bp) for 154 species. These sequences were also used to generate a separate dataset of the MiFish region (166–199 bp) by cutting off redundant regions from the sequences. We further obtained sequences for 39 species for which only sequences of the MiFish regions are available in the INSD. By combining with the sequences we determined by ourselves, we obtained 330 longer sequences of the first half of the 12S region (725–773 bp) for 165 species and 523 sequences of the MiFish region for 194 species (Table 1). Furthermore, 20 sequences of a possibly undescribed species of the genus Pomacentrus, which was referred to as “Minamiiso-suzumedai” by Motomura (2020), were included in our data. In this study, it was treated as “P. sp. minamiiso”.

To detect sequences with unreliable labels due to various errors (Claver et al. 2022), we tentatively constructed Neighbour-joining trees for sequences of the longer 12S region and those of the MiFish region, based on the Tamura-Nei distance after sequence alignment using the MEGA-X software (Kumar et al. 2018). The pairwise distances for conspecific sequences and those for interspecific, but congeneric, sequences were calculated for the longer 12S and MiFish sequences using the MEGA-X software. Then, the sequences that were assigned to strange positions in the trees and those demonstrating apparently low sequence similarity to other conspecific or congeneric sequences were scrutinised and excluded when they were uncertain. The species that were subjected to taxonomic revisions after sequence deposition to INSD were carefully consulted and scientific names were replaced with others in 36 cases according to the latest taxonomic system described by Tang et al. (2021) (Table 1).

After removing erroneous sequences, we prepared three datasets. The first was the “longer 12S dataset (725–773 bp)” that consisted of 301 sequences for 150 species (8.8% of the equences we obtained from INSD was excluded after scrutinisation; Table 1), of which 53 species have more than one sequence. It covered 70.2% (80/114) of pomacentrid species known from the Japanese waters and 36.1% of species known in the world. The second was the dataset of the MiFish region (166–199 bp) consisting of the same 301 sequences for the 150 species that was generated by cutting off redundant regions from the first dataset. The third dataset was of the MiFish region, but including additional sequences retrieved from INSD. This dataset consisted of 476 sequences for 183 species (9.3% of the sequences we obtained from INSD was excluded; Table 1), with 87 species having more than one sequence and covered 89.5% (102/114) of Japanese species and 43.8% of species in the world.

NJ analyses and genetic distance calculations were performed for the three datasets to investigate the capability of the two DNA regions (the longer 12S and the MiFish regions) for species identification. We first evaluated their capabilities according to placement of the conspecific sequences in the NJ trees. The NJ trees, based on the Tamura-Nei distance, were drawn and subjected to the bootstrap analysis with 1,000 pseudoreplications using the MEGA-X software. Ratnasingham and Hebert (2013) recognised the following four categories in evaluation of the performance of DNA barcoding of a given DNA region, based on recognition of the operational taxonomic unit (OTU) using the NJ tree framework: 1) Match, in which all sequences of a given species were placed collectively and exclusively in a single OTU, indicating that the barcoding would be successful for that species; 2) Split, in which the sequences of a given species were assigned to more than one OTU, but not intermixed with other species, indicating that barcoding would be a failure; 3) Merge, in which all sequences of a given species were placed in a single OTU together with another species, indicating that barcoding would be a failure; and 4) Mixture, in which the sequences of a given species were assigned to more than one OTU and at least one sequence is shared by another species, indicating that barcoding would be a failure. We adopted these criteria in evaluations of the capabilities of the longer 12S and MiFish regions for the pomacentrid species.

It is a fact that the NJ analyses approach is not applicable to species for which only a single sequence was included in the datasets, although we could consider barcoding a failure when the single sequence of a given species is shared by another (i.e. equivalent to “Merge” in the abovementioned four categories). In other cases, we focused on the genetic distance of a given species with its most closely-related species compared with the genetic distance generally observed within a species. Uncorrected p-distance was calculated for combinations of the relevant sequences using the MEGA-X software.

For setting an upper threshold of genetic distance within a species, we referred to genetic distances within each of the 53 and 87 species for which more than one sequence is available in the longer 12S dataset and the third dataset of the MiFish region with the larger samples. Our analyses demonstrated that the maximum intraspecific genetic distances were < 0.01 for 86.8% (longer 12S) or 80.5% (MiFish) of the species, respectively (see “Results”). Thus, we tentatively included 0.01 in the genetic distance as a threshold and judged a given species as “success” for barcoding when the genetic distance between that species and its closest relatives was > 0.01.

The NJ trees constructed, based on the longer 12S and MiFish datasets, are depicted in Figs. 1–3. In the longer 12S NJ tree, conspecific sequences of 49 of the 53 species for which more than one sequence was available exclusively formed respective clades and were judged “Match” (Fig. 1). The bootstrap values for these 49 clades were 64–99 (96.7 on average). Sequences of the remaining four species were classified as “Mixture”. In addition to these, 2 of 97 single-specimen species shared identical sequences with other species and were classified as “Merge”.

Figure 1.  

The Neighbour-joining tree constructed, based on the longer 12S dataset for 301 sequences of 150 pomacentrid species. Identical sequences are not merged and, thus, all the 301 sequences appear as terminal OTUs. Those sharing identical sequences are expressed as branch length of zero. The bootstrap values are shown for species were judged "Match".

In the NJ tree constructed, based on the MiFish dataset for the same samples, the sequences of 41 species exclusively formed respective clades (Fig. 2). Of the remaining 12 species, two were classified as “Split,” four were classified as “Merge” and six were classified as “Mixture”. In addition to these, 12 single-specimen species shared identical sequences with other species and were classified as “Merge”.

Figure 2.  

The Neighbour-joining tree constructed, based on the MiFish dataset for 301 sequences of 150 pomacentrid species. The bootstrap values are shown for species were judged "Match".

Fig. 3 shows the NJ tree constructed, based on the third dataset with 476 MiFish sequences for 183 species. Conspecific sequences of 71 of the 87 species for which more than one sequence was examined exclusively formed respective clades. The bootstrap values for these clades were 37–100 (85.7 on average). Of the remaining 16 species, two were classified as “Split”, seven were classified as “Merge” and seven were classified as “Mixture” (Table 2). In addition to these, 13 of 96 single-specimen species shared identical sequences with other species and were classified as “Merge”.