DNA fingerprinting: an effective tool for taxonomic identification of processed precious corals

Precious coral species have been used to produce jewelry and ornaments since antiquity. Due to the high prices at which corals are traded, coral beds have been heavily fished. Hence, fishing and international trade regulations were put in place. However, poaching remains extensive and mislabeling of products is common. To this date, the control of precious coral exploitation and enforcement of trade rules have been largely impaired by the fact that species of processed coral skeletons can be extremely difficult to distinguish even for trained experts. Here, we developed methods to use DNA recovered from worked precious coral skeletons to identify their species. We evaluated purity and quantity of DNA extracted using five different techniques. Then, a minimally invasive sampling protocol was tested, which allowed genetic analysis without compromising the value of the worked coral objects. We found extraction of pure DNA possible in all cases using 100 mg skeletal material and over half of the cases when using “quasi non-destructive” sampling with sampled material amount as low as 2.3 mg. Sequence data of the recovered DNA gave a strong indication that the range of precious coral species present in the trade is broader than previously anticipated.


Introduction
trained experts based on morphological characteristics, and proper analytical tools to conclusively identify the species of worked precious corals are still lacking 6,12,18,19 .
The various analytical methods tested to distinguish precious coral species based on skeletal material were either unable to provide clear-cut distinction among the different coral species (i.e. trace element analysis, 20 ; X-ray fluorescence spectroscopy and Raman spectroscopy, 21 ), or were not improved to become a standardized and easy-to-use tool i.e. immunolabeling, 22 . As a novel approach, Cartier, et al. 23 recently proposed DNA analysis to distinguish species, assuming that coral DNA molecules can be trapped in the organic material or adhered to the CaCO3 crystals during the skeleton formation.
Genetic analyses have become a powerful analytical tool to elucidate the species identity and trace the geographic origin of various valuable artefacts of biogenic origin. These include processed products of tortoise shell 24 , snake skin 25 , fur 26,27 , ivory 28,29 or tiger bones 30 . Of greatest relevance to this present study, Meyer, et al. 31 reported quasi-nondestructive species identification of pearls based on DNA analysis, where so little amount of pearl material was used for the analyses that the gemological value of the pearl was not compromised. Particular biogenic materials require specific DNA extraction methods, moreover, we anticipate that DNA preserved in precious corals skeletons to be present in very small amounts and highly fragmented due to the lengthy skeleton-formation process and because the majority of corals are already dead when fished 17,32-34 . In the present study, we aim to explore whether precious coral skeleton fragments cut, carved and polished for jewelry could be taxonomically identified through genetic analysis. We compare five different DNA extraction methods to find the method producing the highest purity and quantity of DNA. We then apply the most successful technique to extract DNA using a minimally destructive sampling method and amplify and sequence the recovered DNA to taxonomically identify the coral samples. We demonstrate that genetic analysis of gem-quality precious corals is an efficient method to assess their species identity.

Results
Comparison of DNA retrieved from worked precious corals with five extraction methods We evaluated which one of five candidate DNA extraction protocols is the most suited for precious coral skeletons. Each of the five tested methods (abbreviated as "W", "F", "B", "E", "Y") have earlier proven to be useful to extract DNA from biomineralized material. DNA was extracted from each of a set of 25 worked coral skeletal samples with all five techniques, and DNA purity and quantity were assessed using real-time quantitative PCR (qPCR) technology.
To test DNA extract purity, we assessed PCR inhibition with qPCR using an internal amplification control molecule. Three extraction methods, "F", "E" and "Y", resulted in DNA with no detectable PCR inhibition effect from any of the tested 25 samples (Fig. 1, Supplementary Results S1). In contrast, a PCR inhibition effect was observed in 15 out of 25 samples extracted with the "B" method. Of these, complete inhibition of the PCR was observed in one case. Inhibition was also detected in three DNA extracts produced with the "W" method.
Of these, no PCR product was observed at all in one sample.
Absolute quantity of the DNA obtained with the five extraction techniques was tested using qPCR with a standard curve from a dilution series of a standard template DNA molecule with known concentrations. Throughout these analyses, the average qPCR efficiency was 88.5 % (± 3.6% standard deviation) and coefficient of determination for the calibration curve was R 2 = 0.9947 (± 0.0035 standard deviation), respectively.
The five extraction methods yielded highly varying amounts of DNA ( Fig. 1, Supplementry Results S1.). Methods "E" and "Y" both yielded PCR amplifications for all 25 samples. Method "W" yielded PCR product for 13 samples, while methods "F" and "B" both yielded PCR product for 21 samples. Overall, there was concordance among the amplification results; the 13 samples that amplified with method "W" also amplified with methods "F" and "B", and the latter two methods amplified DNA of the very same 21 samples. Strong significant correlation was found between the copy numbers obtained from the same coral items with the "E" and "Y" methods (r=0.97, t=19.223, df=23, p<0.001). The DNA yield was higher with method "Y" than with method "E" (595 versus 944 molecules per mg coral skeleton with "E" and "Y", respectively; paired t-test: t = -2.8832, df = 24, p = 0.008). Focusing on the best performing "Y" method, DNA concentrations ranged three orders of magnitude; three samples had over 10 3 DNA copies in each mg of skeleton material. In contrast, in five other samples this value was below 10 (Fig. 1). Figure 1. Results of the DNA extract purity and quantity measurement experiment and taxonomic identification of 25 worked precious coral samples. Five methods were used to extract DNA from equal amounts of material from each sample. PCR inhibition measurement and absolute template quantification was performed with quantitative real-time PCR. Two short mitochondrial DNA fragments were sequenced and each specimen was taxonomically assigned.
DNA extraction with "quasi non-destructive" sampling of worked precious coral skeletons We developed a "quasi non-destructive" technique to take material for analysis from the worked corals with minimal weight loss and virtually invisible effects of the sampling (Fig. 2).
A new set of 25 worked coral samples were sampled in this manner; removed material amounts ranged from 2.3 mg to 13.1 mg and were 7.9 mg on average. Modifications were applied to the lysis step of the "Y" extraction method compared to the original protocol, which resulted in an essentially complete dissolution of the coral powder. This allowed the amount of DNA that remained trapped in the undissolved powder to be kept to the minimum. Out of the 25 "quasi non-destructively" sampled worked coral objects, 16 gave qPCR amplicons at least twice ( Fig.   3, Supplementary Results S1). Another two samples produced amplification only once and were omitted from further analyses. DNA copy numbers calculated per mg of coral skeleton were in the same range as in the case of the extractions carried out from c. 100 mg material using the "Y" method. However, the presence of unsuccessful amplifications and lower average copy number (160 DNA copies) recovered per mg of coral skeleton indicates that DNA recovery from low amount samples is less effective than from standard material amount, despite the amendments made in the DNA extraction protocol.

Taxonomic assignment of worked precious corals
We sequenced amplicons of the large ribosomal RNA gene subunit (LR) and the putative mismatch repair protein (MSH) fragments originating from a total of 41 worked coral skeletons. In our entire DNA sequence dataset, sequence of three OTUs did not align with neither with the LR nor the MSH reference sequences. NCBI BLAST search did not find any sequence entries in the NCBI database with higher than 95% sequence similarity to any of these sequences.
Length of concatenated LR and MSH sequences were between 264 base-pairs (bp) and 290 bp long per coral sample (Supplementary Results S2). Phylogenetic analysis identified 10 samples (11,14,19,22,23,31,34,38,41,45) as Corallium rubrum, of which nine had sequences identical to either of two the reference C. rubrum sequences, and one (11) had a single variable site (Fig. 4). Six samples (9,17,20,21,28,35) were identical with reference samples of Corallium japonicum, but also with the reference samples of C. nix and C.   Posterior probability value is displayed after each tree node.

Discussion
Technical advancements and the growing body of reference DNA data have made genetic analyses a powerful tool to combat poaching, illegal trading and mislabeling of animal products 35 . Application of genetic barcoding was suggested by Ledoux, et al. 36 as a forensic tool to identify species of corals. Acknowledging that the discriminatory power of standard species barcoding markers (e.g. the cytochrome c oxidase subunit I gene) is poor to distinguish the closely related precious coral species, these authors suggested development of custom designed species identification markers. Moreover, if the aim is to distinguish coral skeleton samples, then the high portion of fragmented DNA will call these markers to be as short as possible. A further challenge is if sampling of the coral skeleton is to be done with minimal material loss, and as consequence, the chosen DNA extraction method has to be capable of recovering DNA from small sample amount.
In our quest to find an optimal method to recover DNA from worked coral skeletons, we tested the performance of five DNA extraction methods, each on equal amount of coral material from the same set of 25 worked coral samples. We found two methods, protocol "E" and "Y" that yielded DNA that was successfully amplified and sequenced from all of the 25 tested corals. Methods "E" and "Y" are two similar techniques developed for the extraction of DNA from ancient eggshells and ancient bones. They only slightly differ in their lysis buffer ingredients and the type of DNA-binding silica column used for the purification of the recovered DNA molecules 37,38 . These methods produced similar amounts of DNA, however method "Y" produced slightly higher DNA yield, particularly in the samples that had < 50 DNA copies per mg coral powder. The three other tested DNA extraction methods did not result in amplifiable DNA from all samples, which may be due to their inability to recover DNA coupled with PCR-inhibitory effect of co-extracted substances, which was detected in some extracts, PCR inhibition was not detected in any extracts produced with methods "W", "E" and "Y". By using these methods, PCR inhibition seems to be overcome in precious corals, unlike in other types of corals, where it led to technical challenges 39 .
DNA concentration of the extracts differed largely; while in certain samples <10 copies per mg material was recovered, in some others this reached up to the order of magnitude of 10 3 copies per mg material. The large variation in DNA preservation of the samples may be determined by their varying ages: corals are often fished decades after their death 17,32,33 and coral skeletons maybe stored for long before they get processed 6 . However, without specific knowledge about the age of the samples this remains hypothetical.
Our test to choose the best DNA extraction protocol from potential methods was based on 100 mg of coral skeleton, which is a standard amount used for extracting DNA from pulverized material with the applied protocols. The essence of precious material testing would be to use as little material as possible, ideally using a "quasi non-destructive" sampling method.
This means that the sampling area is not visible and the sampling does not cause significant weight loss of the coral object. Worked coral skeletons can be separated into two main types; the ones that have a hole drilled through the item (generally those that are strung for bracelets or necklaces) and the ones that do not have a hole, instead generally have flat reverse or bottom sides (those that are mounted to a frame and used as pendants, i.e. cabochons, or the carved figures used as ornaments). We performed "quasi non-destructive" sampling using a drill with a 0.8 mm diameter diamond engraver head taking care not to heat up the sampled object (no hard pressing of the drill and regular pauses to let the drill head cool down). With careful handling, it was possible to take sample material by slightly widening the internal surface of the ca. 1 mm wide drill-holes completely invisible by eye. From the cabochons, a thin layer was removed from the reverse side; therefore the visible front side remains unaffected by the sampling. Assuming approximately 3.8 kg/dm 3 density of the precious corals 9 , the removed 2.3-13.1 (in average 7.9 mg) mg sampled powder per sample corresponds to a 0.7 -3.5 (2.1) mm 3 volume loss of the items.
We were able to repeatedly produce PCR products for altogether 16 out of the 25 "quasi non-destructively" sampled worked coral skeletons. We cannot determine a threshold for the minimum amount of material necessary for successful genetic testing; the two samples processed with the lowest weight of coral powder, 2.3 mg and 2.6 mg, respectively, both produced results. Although it was not possible to genetically analyze all samples with the minimally destructive method, there might be a good chance that when analyzing several samples from a batch of samples, at least some will produce results.
We expected that the DNA sequences we generate will cluster together with reference sequences of one of the eight species listed by CIBJO as relevant in the jewelry industry. Our taxonomic identification markers allowed us to distinguish all of these eight precious coral species from the others with the exception of Pleurocorallium elatius and P. konojoi. However, unexpectedly, we found a much higher diversity within our samples, with several of our sequences not grouping together with the reference sequences of the eight species. Hence, we repeated the phylogenetic analysis with an extended reference sample set. The results of this analysis show that samples could clearly be identified as Corallium rubrum. The samples grouping together with Corallium japonicum also grouped together with two other species, C. nix and C. tortuosum, which, however, have white and pink color, respectively, unlike the dark red color of C. japonicum 44,40 . Hence, we confidently identify these red corals as C. japonicum.
Samples that grouped together with the Hemicorallium references all had identical sequences with multiple Hemicorallium species. As a consequence, these samples could be identified only to the genus level as Hemicorallium. A part of these samples (i. e. 3, 5, 10, 13 47 . Of these species, the latter two are well known in the jewelry industry, while the former is a recently described species known from single area of the West Pacific 41 . To distinguish these species, the coloration of the skeletal axis may give a partial solution. In particular, the color of P. carusrubrum is red, P. elatius varies from white to dark pink, while P. konojoi is always pure white 4,6,9 . Consequently, our specimens identified as one of these species with pink shading may be identified as P. elatius, while our samples with white color are determined as P. elatius / P. konojoi. Of our multiple samples within the Pleurocorallium clade that did not group together with the species traditionally accepted as present on the coral market (P. elatius, P. konojoi and P. secundum ), two samples (18,46) formed an individual clade and were identified to the genus level as Pleurocorallium. DNA sequences of the other samples were all identical with the sequences of the Pleurocorallium niveum samples. This species was described from waters surrounding the Hawaii islands, which is a historically important coral fishing area 42, 43 .
The 41 samples that we managed to genetically analyze from 50 samples of a single collection is not representative to draw conclusions about the entire jewelry industry, but it indicates that there may be more species present in the trade than the eight precious coral species commonly listed as part of the jewelry industry (cf. 9,10,18 ). This is conceivable, if we consider that in the Pacific Ocean different precious coral species may co-occur and coral fishing does not seek to individually separate them based on species.

Conclusions
This study is a proof of concept that genetic analysis can be an effective tool to taxonomically identify precious corals worked for jewelry. We demonstrated that while 100 mg coral skeletal material is sufficient for successful DNA extraction in all cases, DNA sequencing and taxonomic assignment were possible with minute amounts of "quasi non-destructive" samples in more than half of the cases. Among the worked precious corals examined in this study, DNA sequence analyses revealed several samples belonging to precious coral species previously not considered to be present in the jewelry industry. Future research should focus on broadening the reference data by sequencing multiple specimens for each species identified by experts to substantiate their intra-and interspecific genetic diversity. This will be an essential step in developing genetic tests to become a reliable and standardized method to promote sustainable use of precious corals in the jewelry industry.

Study species
The precious corals relevant in the high-end jewelry industry are Octocorallid Anthozoans that belong to the Alcyonacea order and Coralliidae family. Recent phylogenetic studies confirmed the existence of three genera in the family; Corallium, Hemicorallium and  Table 1, while further details on their distribution, taxonomy, harvesting and conservation are presented in Supplementary Table S3.

Genetic markers used in the study
We expected that the DNA extracted from the coral skeletons would be highly degraded.
Therefore, we used markers developed on the mitochondrial genome, which is present in each cell in multiple copies and offers the best chances of achieving positive results for fragmented DNA. Octocoral mitochondrial genomes have an exceptionally low rate of evolution and standard taxonomic markers are unable to distinguish closely related species 48,49 . Hence, we developed two genetic markers with the criteria that they are, at the same time, short to be suitable for degraded DNA and highly variable to maximize our ability to identify the precious coral species to the lowest possible taxonomic level. We expected each analyzed sample to originate from one of the eight precious coral species listed by CIBJO, thus developed our markers with the aim that they are capable of distinguishing these eight species. The two markers were used in combination to taxonomically identify the coral skeleton samples.
Furthermore, one of them was used to test purity and DNA quantity of our extracts.
The two mitochondrial markers were developed based on DNA sequence data of Tu, et al. 45 , which is the most detailed study on precious coral phylogeny to this date. Marker selection and procedure of designing PCR primers are detailed in Supplementary Methods S4.
Following examination of the phylogenetic resolution of multiple short mitochondrial genome fragments, we developed the two set of primers for the large ribosomal RNA gene subunit (LR) and the putative mismatch repair protein (MSH), respectively. The LR marker was used for the assessment of DNA extract purity and DNA quantification. Phylogenetic analysis using the LR and MSH markers showed that these two short markers were able to reconstruct the phylogenetic relationships obtained by much longer sequences, and they allowed the distinction of each of the eight precious coral species from each other, except for Pleurocorallium elatius and P. konojoi, which are not possible to conclusively distinguish based on the data of Tu, et al. 45  Methods S5 and crushed in a metal mortar with a metal pistil to produce crude coral powder, which was then transferred to a porcelain mortar and ground to fine powder. The coral skeleton powder was divided into five aliquots of equal weight, 100 mg ± 1 mg in general, except for four samples that had less available powder (Supplementary Table S6). The powder aliquots were used to extract DNA using five different extraction methods, which have proven to successfully recover DNA from biomineralized material ( Table 2). For each method, we followed the protocols cited in Table 2. All DNA extracts were eluted in 100 µl and stored at - Table 2. The five different methods tested to extract DNA from precious coral skeletons worked for jewelry. The abbreviated method name is followed by references of relevant studies where the method was used to extract DNA from calcified material, the chemical composition of the demineralization buffer, the method used for DNA binding and DNA purification and, finally, the exact protocol followed.  We compared the DNA quantities gained with the extraction methods for which DNA was successfully amplified for all 25 samples with a correlation test and paired t-test in R 57 .

Method
"Quasi non-destructive" sampling, DNA extraction and quantification In the previous experiment, 25 samples were completely pulverized and five DNA extractions were carried out with different methods from each. The aim was to select the most suitable technique for extracting DNA from coral skeletons. In the current experiment, the best performing DNA extraction technique was used with "quasi non-destructive" sampling of processed corals. We define "quasi non-destructive" sampling as taking material for analysis from the worked objects without compromising its gemological value. A new set of 25 worked coral samples were selected from the SSEF coral collection for this experiment (named samples   26-50, Supplementary Table S8), and each was thoroughly cleaned as described in Supporting Methods S3. Two main types of samples were sampled differently: i) beads with drill-holes: the inner surface of the drill-hole was carefully widened (Fig. 2a); ii) worked items with no existing drill-hole: a small layer of the surface of the back side was removed (Fig. 2b). We used 0.8 mm diameter diamond engraver bit heads attached to a Dremel 4000-4 rotary tool (Dremel, Racine, WI, USA). Rotation speed was set to 10,000 rpm and coral drill-powder was left to drop in 1.5 ml collection tubes.
DNA was extracted from the quasi non-destructively sampled drill-powder of the 25 samples with the "Y" method. The material amount obtained by the "quasi non-destructive" sampling was far lower than the 100 mg used in the experiment comparing extraction methods, therefore we slightly modified the "Y" protocol to accommodate it to the low material amount.
In particular, 200 µl lysis buffer was added to the coral powder and incubated at 56 °C for one hour with mixing, then another 100 µl lysis buffer was added. The lysis-mixture was then incubated again with mixing at 56 °C for one hour and then at 37 °C for additional 65 hours.
The lysate was then mixed with 450 µl 1 × TE buffer and 3750 µl PB buffer (Qiagen) and the entire volume of the mixture was centrifuged through a MinElute (Qiagen) column, which washed then washed with PE buffer and the DNA was eluted in 35 µl EB buffer (Qiagen).

DNA amplification and sequencing
We sequenced the qPCR products of the LR fragment generated for the DNA quantity assessment. From each sample, one of the triplicates was selected for sequencing. The MSH region was then amplified for all 25 DNA samples extracted with the "Y" method in our DNA extraction test and the 16 DNA extracts from the "quasi non-destructive" sampling that gave amplification products for the LR region. The MSH was amplified in singlicate for each sample with identical reaction setup and cycling conditions as described above for the LR region.
The 16S and MSH PCR products were purified with the AMPure bead system (Beckman