Scleractinian mt genomes representing 71 species, 36 genera, and 15 families were downloaded from the NCBI Genome resource² on May 1, 2021 (Supplementary Table 1). Downloaded mt genomes were aligned using mafft v7.475 with the - -adjustdirection option (Katoh and Standley, 2013). Then, prospective PCR primer regions, highly conserved in all 71 scleractinian species, were visually identified for each mitochondrial gene (13 protein-coding genes and 2 rRNA genes). Primer pairs were designed using Primer3Plus (Untergasser et al., 2007). Primer pairs were selected so that PCR products would be <550 bp for use with the Illumina platform, with a maximum sequencing read length of 300 bp (with paired-end sequencing, and more than 50 bp of overlap length between read 1 and read 2 sequences). After inspection of aligned whole mt genome sequences, we designed a primer pair for the 12S ribosomal RNA gene (12S) and a primer pair for the cytochrome c oxidase subunit I gene (CO1) in sequence regions that are highly conserved among all scleractinians examined. We also examined all other mt genes, but their nucleotide sequences were not conserved, so it was impossible to develop primers for these genes. Only 12S and CO1 are sufficiently conserved to develop universal primers for all scleractinians. Expected PCR-amplified sequences from the 71 mt genomes were extracted using Seqotron version 1.0.1 (Fourment and Holmes, 2016). To check the number of unique sequences in the expected PCR-amplified regions, these sequences were clustered using CD-HIT-EST version 4.6 with 100% nucleotide percent identity (Li and Godzik, 2006). For in silico comparison with previously reported universal primers, we used primers for 12S reported in Chen and Yu (2000), primers for CO1 reported in Fukami et al. (2008), and metabarcoding primers for CO1 reported in Nichols and Marko (2019).

Five liters of seawater from the surface and from 2–3 m depth were collected at each of three points along a coral reef offshore of Onna Village, Okinawa, Japan, where coral coverage was around 50–90% and the genus Acropora was dominant, on 29 April, 2021 (Figure 3A and Supplementary Table 2). Twenty 50-mL aliquots of seawater were filtered through 0.45-μm Sterivex filters (Merck), for a total of 1 L from each sample. eDNA in seawater was extracted following the Environmental DNA Sampling and Experiment Manual Version 2.1 (Minamoto et al., 2021; Supplementary Table 3).

Extracted eDNA samples were PCR-amplified using Scle_12S_Fw (5′-CCAGCMGACGCGGTRANACTTA-3′) and Scle_12S_Rv (5′-AAWTTGACGACGGCCATGC-3′) primers to amplify 12S rRNA genes (Figure 1A), and Scle_CO1_Fw (5′-ATTGTNTGRGCNCAYCATATGTTTA-3′) and Scle_CO1_Rv (5′-CCCATAGARAGNACATARTGAAA-3′) primers for CO1 genes (Figure 2A), using genomic DNA isolated from an Acropora tenuis colony as a positive PCR control. Tks Gflex™ DNA Polymerase (Takara) was used for PCR amplification. PCR cycling conditions were 1 min at 94°C, followed by 35 cycles of 10 s at 94°C, 15 s at 60°C (12S) or 55°C (CO1), and 30 s at 68°C, with an extension of 5 min at 68°C in the final cycle. PCR products were extracted and cleaned with a FastGene Gel/PCR Extraction Kit (NIPPON Genetics). Amplicon sequencing libraries of cleaned PCR products were prepared using a KAPA Hyper Prep Kit (NIPPON Genetics Co., Ltd.) without fragmentation. Libraries were multiplexed and 300-bp paired-end reads were sequenced on a MiSeq platform (Illumina) using MiSeq Reagent kit v3 (600 cycles).

First, low-quality bases (Phred quality score <20) and Illumina sequencing adapters were removed from raw sequence reads using cutadapt version 3.4 (Martin, 2011). Sequences ≥200 bp were retained. Merging of paired reads, unique sequence identification, chimera removal, and denoising (error-correction) were performed using USEARCH, version 11.0.667 (Edgar, 2010) as follows. Each pair of quality- and adapter-trimmed read 1 and read 2 sequences were merged (assembled) with the “fastq_mergepairs” command and the “-fastq_minovlen 30” option. For each sequencing sample, merged read sequences were dereplicated using the “fastx_uniques” command with a minimum abundance of 10 reads (-minuniquesize 10), and denoising and chimera removal were performed using the “unoise3” command for each sample with a minimum abundance of 10 reads (-minuniquesize 10). Then, denoised (error-corrected) operational taxonomic units, called ZOTUs (zero-radius operational taxonomic units), were prepared for each sample.

ZOTU sequences from all samples were concatenated and clustered using CD-HIT-EST version 4.6 with 100% nucleotide identity (Li and Godzik, 2006). Clustered unique ZOTU sequences were used for the database for mapping. Merged sequences from each sample were mapped to the clustered ZOTUs and numbers of mapped sequences for each ZOTU were counted using the USEARCH “otutab” command with a percent identity of 99% (-id 0.99).

To identify ZOTUs that originated from scleractinians, we selected ZOTUs satisfying the following criteria: (1) taxonomic assignments of ZOTUs using the NCBI NT database (downloaded on July 9, 2021) were performed using Assign-Taxonomy-with-BLAST³, and ZOTUs with the best estimated scleractinian taxonomy were selected. (2) BLASTN searches against expected PCR amplified regions of 12S and CO1 genes from the 71 downloaded scleractinian genomes were performed using BLASTN version 2.12.0 + and ZOTUs with an e-value ≤ 1e^–20, percent identity ≥90%, and query coverage ≥95% were selected.

After selecting ZOTUs from Scleractinia, mapped reads to ZOTUs from the same genera were combined. To remove possible errors and contamination, we removed scleractinian genera restricted to the Atlantic Ocean and only genera with more than 0.1% of the total number of mapped reads in a given sample were considered. To infer similarities between samples and sampling locations, hierarchical clustering based on the ward D method was performed using percentages of coral genera detected in each sample using the pheatmap command⁴ in R version 4.1.2 (R Core Team, 2015).

We aligned whole mitochondrial genome sequences of 71 scleractinian species (aligned sequences are available as Supplementary Figures). After we checked aligned sequence regions of 13 protein-coding genes and 2 rRNA genes, we designed primer pairs for the 12S gene, which amplified ∼366–465 bp (Figures 1A,B), and the CO1 gene, which amplified ∼296–302 bp (Figures 2A,B). Due to the short amplified sequence lengths, fewer unique sequences and distinguishable genera were identified among the expected amplified sequences from 71 scleractinian species than with previously reported universal primers for the 12S and CO1 genes (Chen and Yu, 2000; Fukami et al., 2008; Nichols and Marko, 2019; Figures 1B,C, 2B–D). However, conservation of nucleotide sequences of primer loci among the 71 scleractinians was higher than of previously reported primers (Figures 1B,C, 2B–D), indicating that the designed primers for 12S and CO1 have potential to efficiently amplify DNAs of various scleractinians.

We isolated eDNA from three locations (surface and 2∼3 m depth), for the six seawater samples (Figure 3A). Then we performed PCR amplification to check whether designed primer pairs could amplify scleractinian genes from seawater eDNA. We obtained PCR products of expected sizes from all eDNA samples with 35 PCR cycles (Figure 3B). Using cleaned PCR products, we performed amplicon sequencing using an Illumina MiSeq sequencer (300-bp, paired-end reads). We obtained ∼345–528 Mbp (570–872 k reads) from 12S amplicons (Supplementary Table 4) and ∼439–612 Mbp (740–1,031 k reads) from CO1 amplicons (Supplementary Table 5). After quality and adapter trimming, each read pair was merged to obtain full length sequences of PCR amplicons. These merged reads were used for clustering and denoising to prepare ZOTUs from each sample. Then, merged and clustered ZOTUs with 100% nucleotide identity were used to create a database for mapping the merged reads. We obtained 2,193 unique ZOTUs from 12S amplicons and 11,171 unique ZOTUs from CO1 amplicons, which also included sequences from non-scleractinians, respectively. Among those, 1,179 ZOTUs from 12S and 3,366 ZOTUs from CO1 were taxonomically identified as scleractinian (Supplementary Table 6). We further selected 1,151 ZOTUs from 12S and 3,356 ZOTUs from CO1 that also showed significant sequence similarities with expected PCR-amplified regions of 12S and CO1 from the 71 scleractinian sequences (BLASTN e-value ≤ 1e^–20, percent identity ≥90%, and query coverage ≥95%). Thus, we concluded that these ZOTUs genuinely originated from scleractinian 12S rRNA and CO1 genes. Numbers of ZOTUs from each eDNA sample are also provided in Supplementary Tables 4 (12S), 5 (CO1).

92.7% of 12S amplicon reads from eDNA samples were properly mapped to ZOTUs from scleractinians (Figure 4A and Supplementary Table 4). In contrast, 72% of CO1 amplicon reads from eDNA samples mapped to ZOTUs from scleractinians, whereas 21.8% of reads mapped to ZOTUs from non-scleractinians (Figure 4A and Supplementary Table 5). Twenty-three genera were detected from 12S amplicons and 14 genera from CO1 amplicons, respectively (Figure 4B). Eleven genera were detected from both, and 12 genera were exclusively detected from 12S and three genera from CO1 amplicons (Figure 4B). When we used Acropora tenuis genomic DNA as a PCR positive control, almost 100% of reads were mapped to ZOTUs from Acropora (Figure 5). In all six eDNA samples, almost half of the reads mapped to Acropora for both the 12S and CO1 PCR amplicons. The next most abundant genus was Pocillopora (12S) and Montipora or Pocillopora (CO1), respectively (Figure 5 and Supplementary Tables 7, 8). Hierarchical clustering of reads mapped to each genus showed that eDNA samples from both amplicon types were clustered by location, but not by sampling depth (Figure 5).

Although the distance between sites 2 and 3 is only 730 m (Figure 3A and Supplementary Table 2), hierarchical clustering based on the percentage of reads mapped to each genus using 12S and CO1 primers was able to separate the three sampling locations, suggesting that this eDNA survey is sufficiently sensitive to differentiate coral communities closer than 1 km. In addition, surface and 2–3-m sample pairs clustered together (Figure 5), indicating that eDNA does not vary based on sampling depth at these locations. No differences were observed in an eDNA survey in western Australia between surface and bottom eDNA samples; thus, sampling from the surface is sufficient (Dugal et al., 2021). In addition, no significant differences in marine metazoan diversity patterns were detected from eDNA samples collected at different depths (Ip et al., 2021). Our results provide additional support for these findings.

Primer pairs that we developed for the 12S and CO1 genes also amplified non-scleractinian genes (Supplementary Table 6); however, 92.7% of 12S PCR amplicons and 72% of CO1 amplicons were from scleractinian mt genes (Figure 4A). Nichols and Marko (2019) developed a primer pair for CO1 genes (HICORCOX_F1 and HICORCOX_R2) and performed eDNA metabarcoding of Hawaiian corals. In our analysis, nucleotide conservation of probable HICORCOX_F1 and HICORCOX_R2 binding regions was quite low. 71% of nucleotides for HICORCOX_F1 and 65% for HICORCOX_R2 primer positions were identical among the 71 scleractinian species examined here (Figure 2C). In contrast, percentages of identical nucleotides in Scle-CO1-Fw and Rv primers were 84 and 87%, respectively (Figure 2B), which were also higher than for another universal primer set for the CO1 gene (Fukami et al., 2008), 75% for AcMCOIF and 60% for AcCOIR (Figure 2D). The greater conservation of our Scle-CO1-Fw and Rv primer sequences may be related to the higher percentage of sequencing reads that mapped to scleractinian CO1 genes (72%, Figure 4A) than percentages of DNA reads mapped to anthozoan 16S (61%) and CO1 markers (51%) (Nichols and Marko, 2019). Although it may be related to differences of benthic fauna, the percentages of scleractinians among PCR amplicons using primer pairs, CoralITS2 and CoralITS2_acro, for nuclear ITS-2 genes were around 50% (46.7 and 57.5%, respectively) (Alexander et al., 2020), and those primer pairs sometimes required more PCR cycles (45 cycles), depending on sampling sites (Dugal et al., 2021). Note that percentages of identical nucleotides for Scle-12S-Fw and Rv primers were 86 and 95%, respectively (Figure 1B), higher than those of other 12S universal primers, 75% for ANTMTSSU-F and 85% for ANTMTSSU-R (Chen and Yu, 2000; Figure 1C) and primers mentioned above for the CO1 gene, indicating high specificity of Scle-12S-Fw and Rv primers for scleractinian mt genes.

With regard to the percentage of PCR amplicon sequences that properly mapped to scleractinian mt genes, our 12S primers (92%) discriminate much more effectively than our CO1 primers (72%) (Figure 4A). The number of scleractinian genera detected using 12S primers from eDNA (23 genera) was also larger than for CO1 (14 genera); however, three genera were detected exclusively by our CO1 primers (Figure 4B), possibly reflecting differences in numbers and species of sequences deposited in the NCBI database to date. Therefore, to maximize the number of coral genera that can be detected using eDNA, it may be suitable to use both 12S and CO1 primers. If eDNA surveys aim to investigate animals other than stony corals at the same time, previously reported metabarcoding primers would be more useful. However, if eDNA surveys need to target only stony corals, the method reported here would yield the best data. Nonetheless, it will be important to verify the effectiveness of our primers by deploying them in various coral reef environments. We hope that the primer pairs we developed for 12S and CO1 genes will become a powerful tool for coral reef monitoring using eDNA and that they will be applied to coral reefs globally.