Enhancing the 12S rRNA reference library of South African chondrichthyans to aid in eDNA metabarcoding

Abstract

Background

The rapid expansion of environmental DNA (eDNA)-based biodiversity monitoring has heightened the need for accurate and well-curated taxonomic reference databases. Although mitochondrial cytochrome oxidase I (COI) is widely used, the mitochondrial 12S ribosomal RNA gene (12S rRNA) has increasingly demonstrated strong taxonomic assignment success in chondrichthyan metabarcoding applications. The South African coastline supports one of the most diverse chondrichthyan faunas globally. However, as in many regions, coverage of 12S rRNA reference databases has remained limited. Well-represented reference libraries are critical for improving the accuracy of eDNA-based species detection. Here, we aimed to expand the regional 12S rRNA reference database for South African chondrichthyans, thereby supporting both local and global eDNA applications for widely distributed species.

Results

A total of 36 chondrichthyan species, represented by 69 individuals, were sequenced to expand the 12S rRNA reference library for sharks, rays and chimaeras from South Africa’s Indian and Atlantic Ocean coastlines. A newly developed primer set (V05F_898E / teleoR_Elas), designed to overlap with the commonly used Elas02 region, achieved consistent amplification across diverse lineages. The new data increased regional reference coverage from 50.3% to 69.1% and added 18 species globally. Sequence authenticity was confirmed through comparison with GenBank, and Bayesian phylogenetic analyses showed that 97.2% of sequences clustered within their expected family-level lineages, although some uncertainties remained in groups with low mitochondrial divergence or complex taxonomic histories. Percent identity analyses revealed high intraspecific similarity and clear interspecific separation across taxa. However, comparative in silico analyses of the shorter Elas02 fragment demonstrated reduced phylogenetic resolution and less consistent clustering among closely related taxa. While the Elas02 region is suitable for accurate amplicon sequence variant (ASV) matching in eDNA metabarcoding, the longer 12S rRNA fragment generated here provides improved phylogenetic resolution and may aid in the identification of unknown or ambiguous ASVs.

Conclusion

The expanded 12S rRNA reference library and associated phylogenetic analyses expanded genetic resources for South African chondrichthyans and support global applications. The newly developed marker can be incorporated in future eDNA studies dealing with elasmobranchs, since it can have the added benefit of providing more phylogenetic resolution than the short Elas02 fragment.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40850-026-00272-4.

Background

Environmental DNA (eDNA) metabarcoding has become an important tool in biodiversity monitoring, allowing for the non-invasive detection of species across diverse ecosystems [1–3]. Among the genetic markers used for this purpose, mitochondrial genes have proven particularly effective due to their high copy number in the environment, and their coverage in reference databases, that can provide accurate species identification [4]. Along these lines, the cytochrome c oxidase subunit I (COI) and 12S rRNA genes have been the marker of choice for taxonomic barcoding [5–7]. Although COI has been a frequently used marker (FishBOL; [5]) more recent metabarcoding studies highlight limitations such as a low amplification success for some taxa due to primer specificity problems [4, 8]. Differently, the 12S rRNA gene often provides better amplification success across species [9, 10] and thus has the ability to provide more taxonomic coverage and at the secondary level can also differentiate between closely related taxa [6, 11]. It is therefore no surprise that the 12S rRNA marker has proven particularly beneficial in eDNA metabarcoding applications [10, 12, 13].

The MiFish U and E primer sets [6] have been widely implemented in 12S rRNA eDNA metabarcoding studies targeting marine teleosts and elasmobranchs [6, 14, 15]. While effective in many contexts, these primers have demonstrated significant amplification biases in highly biodiverse coral reef environments, where the DNA of more abundant species can overshadow that of rarer taxa such as elasmobranchs [16, 17]. In fact, a study by Czachur et al. 2021 [18] further highlighted this issue by failing to detect any elasmobranch DNA along the entire South African coastline using the MiFish E primer set. To address these shortcomings, the Elas02 primer set [14] was developed which is a modified taxon specific version of the MiFish E primer set. This primer has demonstrated greater success in amplifying elasmobranch DNA [19, 20] and has shown utility for tracking species distributions and assessing community composition [21, 22].

One of the key advantages of the 12S rRNA gene is its hypervariable regions, which can provide species level taxonomic resolution for fish similar or even better than the COI gene [4, 23]. However, in sharp contrast to the COI reference data currently available on GenBank, the 12S rRNA marker is not that well represented in these databases. Currently, ~ 17% of the globally estimated 1,250–1,300 chondrichthyan species, have a 12S rRNA reference barcode available (GenBank, accessed 30/04/2025). There are especially large amounts of missing data for endemic and range-restricted taxa [4, 24]. The absence of a more comprehensive 12S rRNA reference database directly constrain species-level taxonomic assignment of sharks, rays and chimaeras and can lead to the inflation of ambiguous calls that can mask the true regional species diversity [4, 24]. To overcome this limitation, it has been argued that regional or locality-specific 12S rRNA reference libraries are needed to improve detection and assignment of elasmobranchs in a specific area [23, 25].

Based on the annotated species list by Ebert et al. 2021 [26] and a GenBank search (accessed 30/04/2025), only 56.7% of sharks (selachii; 63 of 111 species) in South African waters have a 12S rRNA reference barcode available (Table 1). The situation is even more pronounced for rays (batoids), where just 43% of species (31 of 72 species) possess reference barcodes. South African chimaeras (holocephali) show a slightly higher coverage, with 62% of species (5 of 8 species) represented [26]. While these figures cover approximately half of known elasmobranch species, a substantial proportion will be undetectable when employing eDNA metabarcoding in the region. There is thus an urgent need for improving the regional elasmobranch 12S rRNA reference libraries in South Africa [15]. This study aims to improve the 12S rRNA reference barcode coverage for chondrichthyans by selectively sampling and sequencing ~ 50% (41 of the 85 missing taxa) of the 12S rRNA region that is currently lacking in the reference database on GenBank. Importantly, however, 30 of these species have a wide distribution outside of South African waters and it is thus envisaged that the addition of these regional sequences will also advance other eDNA studies conducted elsewhere across the globe.

Table 1.

Summary of 12S rRNA reference sequences available for South African chondrichthyan groups, showing the number of species with full, partial or no sequences and those sequenced in this study

	Full	Partial	Not available	Sequenced for this project
Selachii	63	9	41	29
Batoids	28	3	41	6
Holocephali	5	0	3	1
Total	96	12	85	36

Partial sequences entail short reference barcodes (e.g. < 180 bp) which cover a partial section of the full 12S rRNA sequences (e.g. 1080 bp)

Materials and methods

Sourcing of existing data and tissue samples

To select species lacking a 12S rRNA reference barcode, a manual search on GenBank (https://www.ncbi.nlm.nih.gov/) was conducted for chondrichthyan species found in South Africa [26], by searching “12S”, filtering for “Chondrichthyes” and then sorting the 3409 results by “Organism Name”. This list was then manually compared with the South African chondrichthyan species list provided by Ebert et al. 2021 [26]. A maximum of eight 12S rRNA reference barcodes per species were downloaded as FASTA files and named with Accession No. and species name for further use (see Supp. Table 1). Tissue samples were then sourced from as many of the species possible that are not currently represented in the GenBank database (see Supp. Table 2). Tissue samples of 41 different South African chondrichthyan species were successfully sourced. For 19 species, a single specimen was sampled, 10 species were sampled in duplicate (two conspecifics) and 11 were sampled in triplicate (three conspecifics). Carcharhinus brachyurus were represented by four specimens, Carcharhinus leucas by five and Carcharodon carcharias by 6. These specimens produced 96 novel reference sequences. Chondrichthyan tissue samples were obtained from previous sampled specimens in the Evolutionary Genomics Group and also from Aletta Bester-van der Merwe (Dept of Genetics, Stellenbosch University) and the South African Institute of Aquatic Biodiversity (SAIAB) and Shark Spotters and were sampled under permits provided by the Department of Forestry, Fisheries and the Environment (DFFE; RE2010/71,2011/55, 2012/38, 2013/41, 2014/39, 2015/50, 2016/31, 2017/22, 2018/46, 2019/56, 2019/56, 2020/48, 2021/61, 2022/14, 2023/42, 2025/72). All tissue samples were preserved in 100% Ethanol.

DNA extractions

Before extraction, a 20 mg to 30 mg piece of Ethanol preserved tissue was rinsed with distilled water and then incubated in a 1.5 ml Eppendorf^® tube at 30 °C for one hour to allow evaporation of ethanol. Total genomic DNA (gDNA) extractions were performed following the tissue protocol of the E.Z.N.A^® Tissue DNA Kit (Omega Bio-tek) or the same protocol associated with the DNEasy Blood and Tissue extraction kit (Qiagen). Final elution volume was 150 µl for the Omega Bio-tek E.Z.N.A^® Tissue DNA Kit and 100 µl for the Qiagen DNEasy Blood and Tissue extraction kit. The extracted DNA were stored in PE Buffer (Qiagen) at − 20 °C until further processed. Before PCR, DNA was quantified on a Qubit^® 4 fluorometer using the Qubit™ 1X dsDNA High Sensitivity (HS) Assay Kit (Thermo Fisher Scientific) and all samples with concentrations over 120 ng/µl were diluted 1:2 or 1:10, which has been shown to improve amplification success [1, 27, 28].

Primer design for 12S rRNA marker

To optimize the amplification of the 12S rRNA gene, a literature review of published 12S rRNA primers previously used in fishes [29–33] was conducted. The Elas02 primer set produces a relatively short amplicon of approximately 175 bp. Although this would be sufficient to detect species in eDNA samples, alternative primer sets targeting longer 12S rRNA fragments were explored to obtain more informative reference barcodes. The suitability of primer sets published by Douady et al. 2003 [31], Greig et al. 2005 [32], Milan et al. 2020 [30] and Cawthorn et al. 2012 [33] was first assessed through in silico evaluation against 44 whole mitochondrial genomes of South African chondrichthyan species (see Supp. Table 3). The genome sequences were downloaded from GenBank and then trimmed to include the 12S rRNA region. These sequences were aligned using MUSCLE in Geneious Prime^® (Version 2025.0.3). Two primers were selected that seemed most optimal across the 44 chondrichthyan species and slightly modified. The selected teleoR [29], was modified to have a degenerate base at the 5′ position and comprised 5′ TTTCCAGTACACTTACCATG 3′, and the selected V05F_898 [30] was modified by removing the first adenine at the 5′ end and comprised 5′ AATCTCGTGCCAGCCACC 3′. Expected amplification fragment length of the V05F_898E and teleoR_Elas ranged from 391 to 702 bp long. The shorter Elas02 primer set was then also aligned with the reference sequences to ensure that overlapping regions are incorporated (the Elas02 forward primer is predicted to anneal 5 bp upstream from V05F_898).

PCR amplification & sanger sequencing

Each PCR reaction was performed in a Multigene Optimax Thermal Cycler (Labnet International, Inc.) with the Supertherm JMR-801 kit (JMR Holdings) and included 25mM MgCl₂, 2 µl 10x Buffer, 200µM dNTPs, 0.1mM of each primer, 1 µl Taq, and 1 µl of gDNA in a 20 µl volume. PCR conditions comprised 2-minute denaturation at 95 °C, followed by 30 cycles of: 30 s at 95 °C, 30 s at 53 °C, and 1 min at 72 °C, with a final extension of 5 min at 72 °C. The amplification success and quality of the PCR product was assessed via 1% agarose gel electrophoresis in the presence of Ethidium Bromide. Samples showing faint or no PCR products were subjected to a fresh PCR cycle, with the DNA concentration altered, number of cycles increased to 50 and the annealing temperature reduced to 50 °C.

Post-PCR clean-ups and sanger sequencing were carried out by the Central Analytical Facility (CAF) at Stellenbosch University using the Nucleofast 96-well post-PCR clean-up plate (Macherey Nagel) or the AMPure XP magnetic bead PCR cleanup system (Beckman Coulter) following the manufacturer’s instructions (protocols automated on a Tecan EVO150 robotic workstation). Cycle-sequencing reactions were set up with the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Thermo Fisher Scientific) and analyzed on an Applied Biosystems 3730xl or 3500xl Genetic Analyzer (Thermo Fisher Scientific). Reverse reads were converted to their inverse complements using Geneious Prime (Version 2025.0.3) and a consensus sequence for each individual was generated. These were compared to chondrichthyan sequences on GenBank using BLASTn to confirm authenticity.

Phylogeny of newly sequenced chondrichthyan specimen

Reference sequences for the phylogenetic analyses were obtained from the NCBI Nucleotide database using the query “(12S[All Fields] AND rRNA[All Fields]) AND “Chondrichthyes“[Organism]”. For each species, up to eight representative sequences were selected. To ensure inclusion of sequences derived from South African specimens, an additional search with the following query “(12S[All Fields] AND rRNA[All Fields] AND South Africa[All Fields]) AND “Chondrichthyes“[Organism]” was performed. The 190 selected reference sequences were downloaded in FASTA format and imported into Geneious Prime and added to the 69 newly sequenced reference barcodes. To ease with the alignment, sequences were grouped taxonomically by order and then family before alignment in the MUSCLE alignment algorithm in Geneious Prime (Version 2025.0.3) (93% similarity threshold, allowing for gaps). Ten sequences downloaded from GenBank lacking sufficient overlap (at least 200 bp) were removed at this stage. A total of 249 sequences were then trimmed to a maximum alignment length of 699 bp. The resulting family alignments were then aligned amongst their respective orders (e.g. Myliobatiformes) which then in turn were aligned within the batoid, selachii and holocephali. Final sequence alignments were manually optimized. The resulting alignments for batoids (102 sequences), selachii (119 sequences) and holocephali (28 sequences) were then extracted and exported in FASTA format to create the national 12S rRNA reference library for South African chondrichthyans. Headers were then adjusted using RStudio (Version 4.3.1) to include Kingdom, Phylum, Class, Order, Family, Genus and Species.

To authenticate the sequences obtained phylogenetic analyses were conducted. For the selachii and batoids, Callorhinchus capensis (GenBank: NC_014284) was used as the outgroup, whereas Carcharodon carcharias (GenBank: PV919455) was used as the outgroup for the holocephali dataset. The selachii 12S rRNA alignment (746 bp; 81 taxa) was subjected to a Bayesian analyses using MrBayes v3.2.7 [34]. The GTR + F+I+R4 best fit substitution model with four discrete gamma rate categories, as identified by IQ-TREE v2.4.0 [35–37], were used to specify priors. For the batoid (728 bp; 27 taxa) and holocephali (698 bp; 6 taxa) TIM2 models were identified as the optimal models. However, MrBayes v3.2.7 [34] does not support the TIM2 model, which is why GTR + F+I+G4 and GTR + F+G4 were chosen for the batoid and holocephali sequences respectively. Two independent runs of four Metropolis-coupled Markov chains (MCMC) were performed for each dataset for 5,000,000 generations, sampling every 1000 generations. The first 25% of samples were discarded as burn-in and remaining trees were summarized to obtain the majority-rule consensus tree with posterior probabilities on nodes. Convergence was assessed by the average standard deviation of split frequencies and by effective sample sizes (ESS) of parameters. All trees were visualized and edited using iTOL v7.3 [38].

To compare phylogenetic inference among sequences obtained with the commonly used Elas02 primers and those generated with the primers designed in this study, the phylogenetic inference was repeated as above but trimmed to the same size as what would be generated by the eDNA metabarcoding approach. This procedure was conducted by using Test with Saved Primers function in Geneious Prime (Version 2025.0.3). The batoid Elas02 alignment (243 bp) was subjected to Bayesian analyses using the GTR + F+I+G4 model of evolution. The selachii Elas02 alignment (230 bp) utilized the TIM2 + F+G4 as prior and the holocephali Elas02 alignment (216 bp), utilized the GTR + F+I model.

Statistical analysis

To correlate the accuracy of the taxonomic assignments made by newly generated sequences, the sequences in the updated reference library were used by first employing the Bayesian consensus tree generated in MrBayes v3.2.7 [34]. The tree was imported into R using the ape package [39]. Species names were extracted from tip labels and monophyly was evaluated with the is.monophyletic() function [40]. The most recent common ancestor of each species was identified with getMRCA() to verify consistency in clade structure. Afterwards, pairwise cophenetic distances were calculated using cophenetic() to quantify tree based intraspecific divergence [40]. Short or incomplete sequences (n = 32; selachii: 6, batoids: 23, holocephali: 3) were identified and removed prior to further analyses and 217 sequences remained. Sequence cleaning was performed by calculating the non-gap character count for each sequence using the Biostrings package [41]. The number of non-gap positions was extracted with the as.matrix() function. Sequences with non-gap lengths below 80% of the median alignment length were removed using a custom R function [41]. Percent identity comparisons were then computed to quantify sequence level similarity within and between species. Alignments were imported with Biostrings and converted to matrices, after which pairwise percent identity was calculated as the proportion of identical nucleotides across positions for each sequence pair. Distance matrices were reshaped into long format using the reshape2 package [42] and annotated with taxonomic information using stringr and dplyr from the tidyverse collection [43]. To assess differences in percent identity between intra and interspecific comparisons, a Wilcoxon rank sum test (wilcox.test()) was performed for batoids, selachii and holocephali. Differences between intra- and interspecific percent identities were assessed using Wilcoxon rank-sum tests. To assess how well the shorter Elas02 fragment perform in a similar scenario, the analyses were repeated by using the pairwise cophenetic distances from the trees constructed using Bayesian analyses. Intra- and interspecific percent identities obtained from the short Elas02 fragment were compared to those from the longer 12S rRNA fragment using Wilcoxon tests to evaluate differences between datasets. All analyses were conducted in RStudio (Version 4.3.1).

Results

Newly developed 12S rRNA primer set

In silico primer–template tests in Geneious indicated that the primers from Douady et al. [31], Greig et al. [32] and Cawthorn et al. [33] were not complementary for the application of barcoding of South African chondrichthyan species, as they did not align with the tested mitochondrial genomes. The forward primer V05F_898 [29] and the reverse primer teleoR [30] showed a greater number of complementary bases to the South African chondrichthyan reference sequences. The forward primer V05F_898 also overlapped with the forward primer of the Elas02 primers set, but in the interest of complementarity it was shifted by 5 bp towards the 5′ end. The teleoR primer had two mismatches, one of them in first position at the 3’ end, which posed high risk of reducing binding efficiency [44–46]. The mismatches were present in all 44 reference sequences across batoid, selachii and holocephali species. The V05F_898 forward primer also had a consistent mismatch in all tested chondrichthyan sequences, by the 5’ end. The newly adjusted primer pair V05F_898E and teleoR_Elas did not have any mismatches with the tested sequences.

Taxonomic expansion of reference database

Ninety-six tissue samples representing 41 different South African chondrichthyan species were amplified. Despite several attempts, Oxynotus centrina, Holohalaelurus favus, Haploblepharus pictus, Carcharhinus altimus and Raja ocellifera did not yield quality sequences or failed to amplify entirely. Of the 36 newly added species, six belong to the batoid group, 22 to selachii, and one to holocephali. All new sequences have been uploaded to GenBank (GenBank Accession No.: PX741113, PX741114, PX741117 – PX741183). In addition, seven selachii species already had existing 12S rRNA barcodes available in public databases which originated from specimens collected outside of Southern Africa (see Supp. Table 1). The inclusion of the new data resulted in an increase from 50.3% to 69.1% for the listed South African chondrichthyan species now available as reference barcodes. The alignment of the sequences to a panel of 13 chondrichthyan mitochondrial genomes (GenBank Accession No.: MT663280.1, NC_057525.1, NC_057057.1, KY067540.1, MH648005.1, KP006330.1, KU873081.1, NC_025954.1, AB560489.1, ON557293.1, NC_043830.1, NC_035491.1, MF983538.1) showed that the amplified region corresponds to positions 325 to 1198 of the reference genomes. This region overlaps with the 12S rRNA gene, which spans approximately 950 bp in vertebrate mitochondrial genomes (GenBank: NC_002012.1) and also overlaps with the Elas02 eDNA barcode region.

Phylogeny trees of South African batoids, selachii and holocephali

The batoid tree comprised 102 sequences, including 11 newly generated reference barcodes (Fig. 1). The selachian dataset included 119 sequences, of which 55 were newly sequenced (Fig. 2), and the holocephalan tree contained 28 sequences, including three newly generated barcodes (Fig. 3). These lineage-specific trees provided a framework for assessing phylogenetic placement and validating the taxonomic identity of the reference sequences.

Fig. 1.

Phylogenetic tree of batoid species found in South Africa using 12S rRNA reference barcodes sourced from GenBank as well as including newly created barcodes from this study (*). Black dots at nodes indicate significant posterior probabilities (≥ 0.95). The colours represent different batoid families occurring in South Africa as per inset

Fig. 2.

Phylogenetic tree of selachii species found in South Africa using 12S rRNA reference barcodes sourced from GenBank as well as including newly created barcodes from this study (*) inferred from Bayesian analysis. Black dots at nodes indicate significant posterior probabilities (≥ 0.95). The colours represent different selachii families occurring in South Africa as per inset

Fig. 3.

Phylogenetic tree of holocephali species found in South Africa using 12S rRNA reference barcodes sourced from GenBank as well as including newly created barcodes from this study (*) inferred from Bayesian analysis. Black dots at nodes indicate significant posterior probabilities (≥ 0.95). The colours represent holocephali families occurring in South Africa

The Bayesian phylogenetic tree recovered batoid species clustering largely according to their recognized families (Fig. 1). The newly sequenced Neoraja stehmanni (NS1) did not form a clade with the remaining Rajidae reference sequences. The placement of the two newly sequenced Cruriraja hulleyi sequences (CH2 and CH3) and Cruriraja parcomaculata (CP1) belonging to Gurgesiellidae and disrupts the monophyly of the Rajidae sequences. Within Gymnuridae, the newly sequenced Gymnura natalensis (GN1) barcode did not cluster with the Taeniura lymna reference sequences included in the dataset. Of the 11 batoid families the 12S rRNA data failed to support the monophyly of Pristidae, Gurgesiellidae, Rajidae, Mobulidae, Gymnuridae and Dasyatidae (54.5% of the families).

Of the 25 selachian families represented in the dataset, 21 formed clades consistent with their recognized family assignments. Just four families (16.0%) were not monophyletic (Alopiidae, Lamnidae, Triakidae and Carcharhinidae). The porbeagle shark (Lamna nasus) did not group with the other Lamnidae sequences rendering the Carcharhinidae paraphyletic. Nevertheless, all newly generated reference barcodes for selachian species (n = 55) were placed within their respective family level clades. Monophyly was assessed for all species represented by at least two sequences in the selachii dataset (n = 17 species). Thirteen species (76.5% of all sampled) formed monophyletic clusters in the Bayesian phylogeny. Four species did not form monophyletic groups: Carcharhinus leucas, Carcharhinus limbatus, Haploblepharus fuscus and Squalus acutipinnis.

The Bayesian phylogenetic tree based on 12S rRNA sequences for holocephali clearly resolves the three recognized families (Callorhinchidae, Chimaeridae and Rhinochimaeridae) into distinct and well-supported clades (Fig. 3). No misplacements or ambiguous affiliations were detected, and all reference sequences, including the three newly generated sequences for Callorhinchus capensis, grouped as expected according to current taxonomic classification. With the exception of Rhinochimaera africana, monophyly was detected in five of the six species (80%).

Using the shorter Elas02 fragment, phylogenetic resolution was significantly reduced (Fig. 4). In the batoid phylogeny (Fig. 4A) several Mobula species were represented by individual or weakly supported nodes. In addition, internal branches were generally much shorter, resulting in star-like topologies in parts of the tree. Consistent with this pattern, family-level monophyly based on the Elas02 fragment was only partially recovered, but the same three batoid families (Plesiobatidae, Rhinidae, Rhinobatidae) were resolved as monophyletic. In the Selachii phylogeny (Fig. 4B), members of Pentanchidae (e.g. Haploblepharus, Halaelurus and Holohalaelurus) were distributed across different parts of the tree rather than forming a monophyletic entity. Overall, ten selachian families (71.4%) were recovered as monophyletic and four (28.6%) as non-monophyletic, although the affected families differed from those in the full-length 12S rRNA phylogeny. The holocephalan phylogeny (Fig. 4C) was least affected supporting the monophyly of conspecific taxa. All three holocephalan families were recovered as monophyletic in both datasets.

Fig. 4.

Bayesian phylogenetic trees constructed using the shorter Elas02 (12S rRNA) fragment for (A) Batoidea, (B) Selachii, and (C) Holocephali, which can be directly compared to the corresponding full-length 12S rRNA phylogenies shown in Figs. 1, 2 and 3. Tip labels represent individual sequences, including both reference and Sanger-generated sequences. Sequences generated in this study are highlighted in yellow. Black dots at nodes indicate significant posterior probabilities (≥ 0.95)

Utility of the 12S rRNA dataset

Percent identity comparisons showed clear separation between intra and interspecific sequence similarity across all taxonomic groups (Fig. 5). After removing 32 short sequences (< 80% of median non-gap length) 79 batoid, 113 selachian and 25 holocephalan sequences remained. Intraspecific comparisons consistently showed high sequence identity (> 99.6%) with limited variance, whereas interspecific comparisons exhibited substantially lower identity values (86.1%). For batoids, intraspecific comparisons had a mean percent identity of 99.8% with values ranging from 98.7% to 100%, whereas interspecific comparisons had a mean percent identity of 86.1% and ranged between 78.4% and 100%. For selachians, intraspecific comparisons had a mean percent identity of 99.6% ranging from 96.2% to 100%, whereas interspecific comparisons had a mean percent identity of 88.8% percent and ranged from 82.1% to 100% percent identity. For holocephalans, intraspecific comparisons had a mean percent identity of 99.8% percent with values between 98.5% and 100% percent, while interspecific comparisons had a mean percent identity of 89.7% and ranged from 87.2% to 99.1% percent. Wilcoxon rank sum tests confirmed significant differences between intra and interspecific identity values for all three groups (p < 0.001). The lowest intraspecific identity in selachians was observed in Carcharhinus leucas (96.2%), while the highest was observed in Carcharhinus brachyurus (100%). The lowest interspecific variation was observed between Notorhynchus cepedianus and Rhizoprionodon acutus (82.1%) while sequences of Squalus blainville and Squalus acutipinnis were identical (100%). In batoids the lowest intraspecific percent identity was observed in Taeniura lymna (98.7%) while Acroteriobatus annulatus experienced 100% intraspecific identity. Neoraja stehmanni and Himantura uarnak had the lowest interspecific identity in batoids (78.8%) while Bathytoshia lata and Bathytoshia brevicaudata had the highest (97.7%). In holocephalan sequences Rhinochimaera africana (98.5%) had the lowest intraspecific percent identity while Harriotta raleighana had the highest (100%). Interspecific percent identity was the lowest between Hydrolagus africanus and Callorhinchus capensis sequences (87.2%) while Rhinochimaera atlantica and Rhinochimaera africana had the highest similarity in bases (99.1%).

Fig. 5.

Violin plots for intraspecific and interspecific percent identity distributions for batoids, selachians and holocephalans based on pairwise distance matrices generated from 12S rRNA MUSCLE alignments

Percent identity comparisons based on the short Elas02 fragment showed clear differences between intra- and interspecific sequence similarity across all taxonomic groups. Intraspecific identity values were consistently high, with mean values of 99.7% in batoids, 98.8% in selachii, and 99.2% in holocephali, and showed relatively low variation. In contrast, interspecific identity values were lower and more variable, with mean values of 84.2% (batoids), 84.4% (selachii), and 85.4% (holocephali), and minimum values ranging from approximately 71% to 78%. Wilcoxon rank-sum tests confirmed significant differences between intra- and interspecific comparisons for all groups (p < 0.001). Comparisons among groups showed no significant differences in intraspecific identity values (batoids: p = 0.156; selachii: p = 0.307; holocephali: p = 0.206), whereas interspecific identity values differed significantly across all groups (all: p < 0.001).

When comparing the short Elas02 fragment to the longer 12S rRNA fragment, intra- and interspecific identity values remained clearly separated in both datasets. However, the magnitude of separation between intra- and interspecific comparisons was reduced in the Elas02 dataset, with lower mean interspecific identity values and a broader overlap between the two comparison categories. Wilcoxon tests further indicated that interspecific percent identity distributions differed significantly between the Elas02 and full-length 12S rRNA datasets (p < 0.001).

Discussion

By sequencing an additional 69 individuals representing 36 species, this study increased the 12S rRNA reference coverage for South African chondrichthyans from 50.3% to 69.1% and also added 29 new sequences beyond the region. This substantially not only expand the genetic data available for systematics but also provide a better genetic baseline or biodiversity assessment through eDNA applications. These new sequences now add verified 12S rRNA sequences that can be incorporated into future updated taxonomic reference libraries at a global scale.

Validation of primers

Validating a newly designed 12S rRNA primer set proofed to be essential for generating accurate 12S rRNA reference sequences for chondrichthyans. Several widely used primer sets [31–33] showed to have insufficient sequence complementarity to many local taxa, which could lead to poor performance in PCR amplification with species also occurring in South African waters. The V05F_898 and teleoR primer combination [30] initially appeared most promising, but mismatches with chondrichthyan mitochondrial genomes pointed to potential reduced amplification efficiency [45, 46]. By altering the primer sequences slightly (producing the V05F_898E and teleoR_Elas pair) we produced consistently strong matches, and the subsequent PCR validation confirmed the broad applicability and robustness of the newly proposed primer pair. Since few primer-template mismatches can reduce amplification success, particularly in diverse groups such as sharks and rays, consistent with previous work [44–46], our results highlight that customized primer development is advised when targeting taxonomically complex chondrichthyan assemblages [47, 48]. The later holds even across phylogenetically distance lineages such as the lanternsharks (Etmopteridae; Squaliformes), stingrays (Dasyatidae; Myliobatiformes) and chimaeras (Callorhinchidae; Holocephali) [49, 50]. The adapted primer pair presented herein can in future be applied in global studies and used to sequence diverse chondrichthyan lineages, particularly with the idea to improve reference data bases. Since it was specifically designed to overlap with the region amplified by the Elas02 primer pair, the created barcodes (~ 699 bp) overlap with the commonly used Elas02 primers and these sequences are thus useful in eDNA metabarcoding studies focusing on chondrichthyan species across different ecosystems and oceans [19, 20, 51–53]. While the newly designed primer pair successfully amplified longer 12S rRNA fragments from tissue-derived DNA, it is important to note that amplification of such long fragments from environmental DNA is generally constrained by DNA degradation [7, 54]. Consequently, the primers developed in this study are not intended for direct application in eDNA metabarcoding workflows. Instead, their primary utility lies in generating high-quality reference barcodes that overlap with shorter metabarcoding regions, such as the Elas02 fragment, thereby improving taxonomic assignment accuracy in eDNA studies.

Phylogenetic resolution of the 12S rRNA marker

Chondrichthyan systematics are still a matter of debate, and the phylogeny remains controversial [55, 56]. The Bayesian phylogenetic reconstructions using 249 12S rRNA sequences of 131 species generated for batoids, selachii and holocephali recovered most recognized families as distinct and well supported clades, demonstrating strong correspondence with current morphological and molecular systematics [50, 57, 58]. Although the purpose of this study was not to resolve Chondrichthyan systematics (especially not based on a single mtDNA gene), it is interesting to note that the majority (97.2%) of newly generated 12S rRNA sequences were placed within their expected family-level lineages. This partly confirms that 12S rRNA barcodes can represent accurate taxonomic assignments at the higher taxonomic level and is useful for metabarcoding applications. Despite this higher-level taxonomic stability (Rhinobatidae, Rhinidae and Mobulidae formed well supported monophyletic clades), deviations from expected placements were identified in that Cruriraja hulleyi did not group with other Gurgesiellidae, disrupting the monophyly of Rajidae. These deviations highlight lineage specific limitations of the mitochondrial 12S rRNA region and emphasize the need for careful interpretation in groups with known low divergence [50, 59, 60]. Historical misidentifications of voucher specimens [57, 61, 62], particularly for morphologically similar deep-water taxa [63, 64], could be promoting these inconsistencies. Cruriraja also failed to form a monophyletic group, which is reflected in the anomalous placement of C. hulleyi and C. parcomaculata. The taxonomy of the genus Cruriraja was uncertain until recently when Compagno et al. 2007 [63] suggested that records labelled as C. triangularis were likely misidentified and represented C. parcomaculata. They also noted that Eastern and Western Cape specimens appeared to belong to an undescribed species, which was later resolved when Aschliman et al. 2012 [64] formally described Cruriraja hulleyi. Taxonomic assignment in batoids remains challenging [26] due to extensive morphological convergence, overlapping character states and the presence of cryptic species complexes that obscure true evolutionary relationships [65, 66]. This problem is especially evident in groups such as Rajidae and Dasyatidae, where phenotypic plasticity and conserved body plans hinder reliable identification just relying on morphology [67]. This leads to uncertainty in current classifications and constant changes [57, 68]. Therefore, species boundaries and family-level placement within this group remain uncertain, and the usage of a single 12s rRNA reference database needs to be treated with caution especially for applications such as eDNA metabarcoding [23, 57, 69].

Within selachii, most families were strongly supported as monophyletic. However, Lamna nasus (GenBank: NC_033911.1) did not cluster with the rest of the Lamnidae sequences. Since BLAST searches in GenBank confirmed the species assignment (100% identity), misidentification of the specimen is unlikely. The placement therefore most plausibly reflects the limited phylogenetic resolution of the short 12S rRNA fragment for resolving deeper lamniform relationships [50]. The placement of the Alopiidae more basal to the Lamnidae and interrupted by Mitsukurinidae and Megachasmidae is in line with previous studies [50, 55, 56]. The order Lamniformes have long presented difficulties for phylogenetic inference [50, 55]. The combination of deep evolutionary age, uneven diversification and low species richness in Lamniformes produces weak phylogenetic signal, making deeper relationships in the order difficult to resolve [50]. Da Cunha et al. 2017 [55] used complete mitochondrial genomes of 73 chondrichthyan species to reassess higher level phylogenetic structure, providing one of the most robust mitogenomic frameworks available for the group. Despite relying on a much shorter fragment of the mitochondrial genome, the 12S rRNA marker in this study (0.55–0.69 kb) recovered several of the same broad patterns reported in their analysis. The placement of Lamna nasus however, remains flawed and should not be considered as true indication of phylogenetic affinity. The disruption of Carcharhinidae by the Sphyrnidae, which separates the sliteye shark (Loxodon macrorhinus) and the milk shark (Rhizoprionodon acutus) from the main carcharhinid cluster, is in line with the literature [50, 55, 70]. Even though there remains uncertainty about the paraphyletic nature of the Sphyrnidae within Carcharhinidae, the hammerhead sharks have been established as sister group to the requiem sharks [55]. Interspecific resolution was poor within the Carcharhinidae family, with minimal genetic divergence observed among congeneric taxa such as Carcharhinus brachyurus and Carcharhinus leucas (> 99% identity). This is consistent with prior reports of mitochondrial homogeneity [50, 71], incomplete lineage sorting and even hybridization within this genus (e.g. Carcharhinus galapagensis and Carcharhinus obscurus [72]).

Similarly, high interspecific similarity between Mobula birostris and M. alfredi (> 99%) reflects well-documented cases of cryptic diversity within mobulid rays [73]. Non-monophyly was detected for the Mobulidae and on species level in four mobulid taxa (M. alfredi, M. eregoodootenkee, M. kuhlii and M. mobular). These limitations present significant challenges for eDNA-based biodiversity assessments for this group. Misassignments could lead to false positives or ambiguous identifications, especially for ecologically and commercially important chondrichthyan species occurring in South African waters [74]. In comparison, holocephali exhibited well-resolved phylogenetic structure. Callorhinchidae, Chimaeridae and Rhinochimaeridae formed distinct, strongly supported clades, supporting the discriminatory capacity of 12S rRNA for these lineages. The relatively low species richness of holocephali in South Africa compared to elasmobranchs may contribute to the greater phylogenetic stability observed in this group [49].

In contrast to the longer 12S rRNA fragment, phylogenetic reconstructions based on the short Elas02 fragment showed reduced topological complexity and less consistent clustering of taxa, as shown here in the Pentanchidae genera distributed across the tree rather than forming a cohesive clade. These patterns are consistent with the expected limitations of short mitochondrial fragments [75, 76], which often lack sufficient phylogenetic signal to resolve relationships among closely related taxa due to low sequence divergence and a limited number of informative sites [4, 50]. Despite this reduced resolution, the Elas02 fragment maintained a clear separation between intra- and interspecific sequence similarity, supporting its utility for species detection in eDNA metabarcoding applications [14]. However, the reduced phylogenetic resolution observed here highlights that short amplicons such as Elas02 are better suited for taxonomic assignment than for resolving evolutionary relationships [77, 78], particularly in taxonomically complex groups such as elasmobranchs [50].

In comparison, the longer 12S rRNA marker provides stronger discriminatory capacity at the genus and family levels, as demonstrated by the largely consistent recovery of monophyletic groupings in this study. However, integrating more than one marker is advised to resolve ambiguous placement of certain taxa [50, 55]. For eDNA metabarcoding in South African waters, these findings imply that selachian and holocephalan detections are likely to be taxonomically robust, whereas batoid groups with comparatively lower interspecific divergence should be interpreted with caution. The use of multiple genetic markers for eDNA taxonomic assignment has been widely recommended, as reliance on a single short mitochondrial fragment can limit species-level resolution in taxonomically complex groups such as elasmobranchs [4, 6, 79–81]. Multilocus approaches that combine markers such as 12S rRNA, COI, Cyt b or NADH2 have repeatedly been shown to improve assignment accuracy and reduce false positives by increasing the number of informative sites available for distinguishing closely related species [4, 80, 81].

Importance of national reference libraries

eDNA metabarcoding and accurate taxonomic assignment are fundamentally dependent on the breadth and quality of the underlying reference library [6, 82]. To support reliable taxonomic identification, a comprehensive South African 12S rRNA chondrichthyan reference library was compiled, comprising 249 sequences from 131 species, including 102 batoids, 28 holocephalans and 119 selachians (40 batoid, 6 holocephalan and 85 selachian species). It was pointed out, that in many regions (e.g. USA, Europe), publicly available databases lack complete or reliable coverage of local fauna, increasing the risk of ambiguous assignments, false positives and inflated diversity estimates [25, 83–85]. Region-specific reference libraries mitigate these issues by anchoring taxonomic assignments to verified local biodiversity rather than relying on incomplete or geographically biased global datasets [23, 83, 86], thereby improving assignment accuracy and strengthening the value of eDNA for ecological monitoring and conservation planning [12, 87, 88].

Following the addition of newly generated 12S rRNA barcodes from South African specimens, only ten of the previously recognised 16 chondrichthyan families remain without species-level representation (see Supp. Figure 1). Reference sequences were added for 9 batoid, 27 selachii and 1 holocephali species, including two sequences of Squatina africana, the only angel shark species (Squatinidae) recorded in South Africa. Seven Pentanchidae species were added, resulting in 50% coverage of this family in South African waters, while the inclusion of three Scyliorhinidae species (Cephaloscyllium sufflans, Scyliorhinus capensis and Poroderma africanum) increased coverage for this family to 100%. Coverage of Rajidae increased from 14.3% to 28.6% with the addition of six species, and a reference sequence for Gymnura natalensis now represents the only Gymnuridae species occurring in South Africa. Overall, 21 families now show complete species-level coverage, highlighting the benefits of systematic, voucher-based sampling.

The achievement of full coverage in 21 families underscores the value of national reference libraries grounded in local biodiversity, which allow controlled expansion, targeted gap filling and alignment with conservation priorities [25, 83]. Similar regional initiatives include the ELASMOMED DNA project in the Mediterranean [86] and the combined barcoding and eDNA metabarcoding approach applied in Mauritania’s Banc d’Arguin [23], which added 42 and six species respectively. In this study, the addition of 69 newly generated sequences provided 12S rRNA barcodes for 36 additional South African species, of which 30 also occur outside of South Africa (Supp. Table 4), which helps to improve taxonomic resolution, filling key gaps and strengthening the reliability of eDNA-based biodiversity monitoring.

Because our reference sequences originate from Southern African specimens, together with targeted GenBank barcodes derived from the region, the library captures local haplotypic variation and improves assignment accuracy. This is particularly important given that many chondrichthyans, including highly mobile pelagic species, exhibit strong population structure and regionally distinct mitochondrial lineages [71, 89–91]. The inclusion of multiple reference sequences per species further enhances representation of intraspecific diversity and increases assignment confidence in metabarcoding studies [6, 82]. Collectively, our findings support the growing consensus that reference library completeness often outweighs marker choice in applied eDNA research [25, 87, 88].

All newly generated sequences were uploaded to GenBank, contributing to a shared global resource that supports collaboration and the continued development of comprehensive reference libraries. As eDNA metabarcoding becomes increasingly embedded in biodiversity monitoring and policy frameworks, investment in voucher-based, regionally targeted reference databases is essential. The remaining 30.9% of South African chondrichthyans without a 12S rRNA barcode are primarily deep-water species, rare taxa, or species with uncertain occurrence in South African waters [26]. Expanding reference coverage for these taxa remains a priority, as their absence could introduce systematic bias in future eDNA studies. Natural history collections and museum specimens are likely to play a crucial role in this effort by providing access to rare, cryptic and geographically widespread species for barcoding [83].

Biogeographic relevance and implications for global chondrichthyan monitoring

The expanded 12S rRNA chondrichthyan reference library presented herein is a valuable resource contributing to the global database that can be used in evolutionary studies, eDNA metabarcoding and taxonomy. South Africa represents a marine transition zone where the Atlantic and Indian Ocean meet [26, 92, 93] and many of the taxa included in this study are not restricted to South African waters (Supp. Table 4) and occur wider in the Atlantic (n = 28), Indian (n = 29) and Pacific Ocean (n = 10). As such, the newly generated reference sequences can support eDNA-based biodiversity assessments in a broader context than just South Africa [6, 82, 94]. Importantly, 12S rRNA barcodes are now available for deep-sea and mesopelagic species such as Centroscymnus coelolepis, Deania calceus and Cirrhigaleus asper. Deep-sea ecosystems remain among the most understudied environments globally, and their biodiversity is still poorly characterized [95, 96]. These taxa are often elusive, occur at low densities and inhabit environments that are difficult to sample using conventional survey methods, contributing to their limited representation in reference databases [26, 97, 98]. From a conservation perspective, several of the species included in this study are of concern, including the critically endangered tope shark Galeorhinus galeus [99]. Although 12S rRNA reference sequences for this species are already available (n = 4, including one from South Africa), the addition of further regionally sourced barcodes improves its representation in reference databases. More broadly, improving reference coverage for threatened and wide-ranging taxa increases the reliability of taxonomic assignments, which is a key requirement for the application of eDNA metabarcoding in biodiversity monitoring and conservation planning [100]. While this study does not directly assess ecological patterns or temporal changes in species distributions, the improved reference coverage provides a necessary baseline for future studies aiming to investigate such processes.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

KL, SA and CAM designed the study. KL analysed and interpreted the data, prepared visualization and was the major contributor in writing the manuscript. KL and SA acquired funding. SA and CAM provided scientific input and edits to the manuscript. All authors read and approved the final manuscript.

Funding

This study was supported by Haistiftung, The Rufford Grant, SharkSafe Barrier, Italtile Foundation Stellenbosch University, SHARKVOICE and SeaToken.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. The datasets generated and/or analysed during the current study have been submitted to GenBank (SUB15865149) and are available under the following link: https://www.ncbi.nlm.nih.gov/nuccore/PX741113,PX741114,PX741117,PX741118,PX741119,PX741120,PX741121,PX741122,PX741123,PX741124,PX741125,PX741126,PX741127,PX741128,PX741129,PX741130,PX741131,PX741132,PX741133,PX741134,PX741135,PX741136,PX741137,PX741138,PX741139,PX741140,PX741141,PX741142,PX741143,PX741144,PX741145,PX741146,PX741147,PX741148,PX741149,PX741150,PX741151,PX741152,PX741153,PX741154,PX741155,PX741156,PX741157,PX741158,PX741159,PX741160,PX741161,PX741162,PX741163,PX741164,PX741165,PX741166,PX741167,PX741168,PX741169,PX741170,PX741171,PX741172,PX741173,PX741174,PX741175,PX741176,PX741177,PX741178,PX741179,PX741180,PX741181,PX741182,PX741183.

Declarations

Ethical approval

Chondrichthyan tissue samples were provided by Aletta Bester-van der Merwe and South African Institute of Aquatic Biodiversity (SAIAB) under permits provided by the Department of Forestry, Fisheries and the Environment (DFFE). All methods carried out comply with relevant guidelines and regulations.

Generative AI and AI-assisted technologies in the writing process

During the preparation of this work the author(s) used ChatGPT in order to improve readability and language in certain sections of the manuscript. After using this tool/service, the author(s) responsibly reviewed and edited the content as needed.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.