Quantitative real‐time polymerase chain reaction (PCR) and droplet digital PCR duplex assays for detecting Zostera marina DNA in coastal sediments

The sequestration of atmospheric CO2 in seagrass meadows as organic carbon (OC) has been attracting more attention as a means for climate change mitigation and adaptation. A direct method to detect seagrass DNA in coastal sediments, which is essential to unravel long‐term seagrass‐derived OC accumulation, was developed based on environmental DNA (eDNA) detection techniques. Quantitative real‐time polymerase chain reaction (qPCR) and droplet digital PCR (ddPCR) were applied to quantify Zostera marina DNA in coastal sediments, using species‐specific primers and dual‐labeled probes for one nuclear and one chloroplast gene. Suitable pretreatments and methods for extracting Z. marina DNA from coastal sediments were examined and their applicability to environmental samples was demonstrated. Surface sediments collected from Z. marina meadows contained about 2000 times more Z. marina DNA than the unvegetated tidal‐flats in the Seto Inland Sea. Moreover, both qPCR and ddPCR successfully detected Z. marina DNA in ancient sediments (up to 5000 calibrated years before present), evidencing that Z. marina DNA can be preserved in temperate coastal sediments for several millennia. In addition, qPCR and ddPCR results obtained in the present study were highly correlated, although the latter was more accurate than qPCR, particularly at low eDNA concentrations in ancient sediments. This work opens avenues to explore and clarify the process of the sequestration of OC produced by Z. marina and demonstrate the presence of past seagrass meadows from several millennia.

Healthy coastal marine environments, in particular those including marine angiosperms, have been suggested as the most important blue carbon sinks, as they can absorb carbon dioxide (CO 2 ) at a similar rate to the most productive tropical rain forest at the areal basis, and store a significant part of the absorbed carbon in ocean sediments (L opez-S aez et al. 2009;Nellemann et al. 2009). Blue carbon, which is the biological carbon captured by marine living organisms (sensu Nellemann et al. 2009), has been recognized both as a scientific and a governmental concern, regarding CO 2 levels reduction and global climate change mitigation (Martin et al. 2016). Therefore, understanding the current status of blue carbon storage and estimating organic carbon (OC) accumulation rate in blue carbon sinks are needed for future global actions considering climate change mitigation and adaptation.
Seagrass meadows are one of the most efficient blue carbon sinks in worldwide oceans (Kennedy et al. 2010;Fourqurean et al. 2012;Duarte et al. 2013), as they can fix CO 2 in their tissues through their high primary production and their vegetative structure can enhance the accumulation of suspended OC into the sediment by reducing water currents. Therefore, some studies tried to provide quantitative estimations of the OC stored in the sediment of seagrass meadows, along with its accumulation rate (reviewed in Miyajima et al. 2015). Using radiocarbon and carbon stable isotope analyses, some of these studies suggested that OC derived from seagrass tissues and cells can be stored in the underlying sediments, being isolated for several millennia from the active carbon cycle involving atmospheric CO 2 . However, a direct method to calculate how long and to what extent seagrassderived OC can persist in the sediments is needed. Such a method would enable determining the long-term dynamics of seagrass-derived OC and the vestiges of ancient seagrass meadows in sediments from previous millennia. Unfortunately, Zostera marina, which is the dominant seagrass species in temperate zones, quickly disintegrates and no visible traces of its components persist in marine sediments for a long time (Reich et al. 2015).
DNA is widely present in marine sediments and plays a role like other persistent macromolecular substances such as cellulose and lignin. The concentration of DNA in the deepsea ranges from 670 ng (g dry sediment) 21 to 24,300 ng (g dry sediment) 21 (Dell'Anno et al. 1999). It has been estimated that DNA could supply from 4.7% to 47% of the nutrient requirements of natural bacterial assemblages in deep-sea sediments (Bailiff and Karl 1991;Dell'Anno et al. 2002;Dell'Anno and Danovaro 2005). Thus, DNA is actively decomposed and fragmented by bacteria and fungi present in marine sediments and, as a result, most of the DNA fragments found in ancient marine sediment cores were less than 100 base pairs (bp; Lejzerowicz et al. 2013).
In recent years, new molecular techniques such as environmental DNA (eDNA) technologies (Ficetola et al. 2008) have been developed, providing information on current and past living organisms and habitats. This technique was recently applied to trace the fate of seagrass-derived organic matter in coastal areas. Reef et al. (2017) analyzed the eDNA from seagrass meadows by next generation sequencing. However, this method is a qualitative method and difficult to determine the relationship between seagrass DNA and OC.
The detection and quantification of eDNA and RNA in water and sediment samples are performed using quantitative real-time PCR (qPCR) (Takahara et al. 2012;Thomsen et al. 2012;Pilliod et al. 2013;Turner et al. 2015;Nakayama and Hamaguchi 2016). In recent years, a third-generation PCR called droplet digital PCR (ddPCR) was developed (Hindson et al. 2011) to provide the absolute quantification of a target DNA without using a standard reference curve, and this technique is more accurate and reliable than conventional qPCR. Thus, ddPCR is usually used for clinical diagnosis of virus infection and cancer, but the methods have been expanded in an application for field science (Kim et al. 2014a,b;Doi et al. 2015;Te et al. 2015).
In the present study, new primers and dual-labeled probes (DLPs) were designed for detecting Z. marina DNA and applied in qPCR as a simple quantitative method to detect ancient Z. marina DNA in coastal sediments. In addition, an absolute quantification method of target DNA using ddPCR and our DLP sets was developed to simultaneously assess the presence and abundance of Z. marina DNA in sediments. To demonstrate the applicability of the developed methods, long sediment cores were collected from a seagrass meadow and a nonvegetated tidal flat, and analyzed for the concentration of Z. marina-derived DNA fragments by both qPCR and ddPCR. Obtained vertical distributions of Z. marinaderived DNA were then compared with that of seagrassderived OC that was estimated by the carbon isotope-based provenance analysis (Miyajima et al. 2015(Miyajima et al. , 2017.

Materials and procedures
Design of primers and DLPs for detecting Z. marina DNA Specific Z. marina primers and DLPs were based on two DNA regions that have been demonstrated to accurately identify Z. marina samples: the nuclear internally transcribed spacer-2 (ITS-2, Olsen et al. 2004) and the chloroplast gene for maturase K (matK, Kato et al. 2003). Sequences of both regions were retrieved from GenBank (https://www.ncbi.nlm.nih.gov/ genbank/; Accession numbers: AY553598 and AB125354, respectively). Primer 3 (Rozen and Skaletsky 2000;Koressaar and Remm 2007) was used to design primers and DLPs. Candidate primers and probes were selected under the following program settings: desired optimal PCR product length of about 100 bp, including primer regions; estimated melting temperature (following Owczarzy et al. 2004) of approximately 608C for primers and 708C for DLPs. Primers and DLPs were synthesized by Sigma Genosys (Sigma-Aldrich Japan, Tokyo, Japan), and DLPs were labeled with 6-carboxyfluorescein (6-FAM) and 6carboxy-2 0 ,4,4 0 ,5 0 ,7,7 0 -hexachlorofluorescein (6-HEX) at the 5 0 end, and with Black Hole Quencher-1 (BHQ-1; Biosearch Technologies, Petaluma, California, U.S.A.) at the 3 0 end. The primer and DLP sets designed in this study are shown in Table 1.

Designed primers and DLPs specificity
The specificity of the newly designed primers and DLPs (AmaITS and AmaMatK) was tested using the two seagrass species inhabiting the Seto Inland Sea, Z. marina and Zostera japonica, the wild terrestrial vascular plants Japanese silver grass (Miscanthus sinensis) and common reed grass (Phragmites australis) inhabiting the margins of the Seto Inland Sea, and rice (Oryza sativa, Japanese type) and common wheat (Triticum aestivum) as agricultural representatives.
To avoid cross contamination, all leaf sample treatments and DNA extractions were performed in separate laboratories. Samples were freeze-dried and disrupted using Tissue-Lyser (Qiagen, Valencia, California, U.S.A.), according to the manufacturer's instructions modified by Shimabukuro et al. (2012), and DNA was extracted from 50 mg of the obtained powder using the DNeasy Plant Mini kit (Qiagen).
To test the efficiency of the newly designed primers and DLPs, a duplex qPCR was performed in a final volume of 10 lL, containing 5 lL the iQ multiplex powermix (Bio-Rad, Hercules, California, U.S.A.), 0.3 lL each primer (300 nM final concentration), 0.2 lL each DLP (200 nM final concentration), 1 lL template DNA, and 3.5 lL sterilized pure water. Amplification reactions were carried out using the CFX96 Touch Real-Time PCR detection system (Bio-Rad). Before analysis, the annealing temperature was determined using the thermal gradient function of the qPCR equipment and ranged from 558C to 658C. Thus, the thermal cycling conditions consisted of 2 min at 958C, followed by 50 cycles of 10 s at 958C plus 30 s at 55-658C. The specificity of the newly designed primers and DLPs was tested in a qPCR performed as described in the previous paragraph, but by using 50 cycles of 10 s at 958C plus 30 s at 588C.

Plasmid development for the construction of the standard curve
The PCR amplicons of AmaITS and AmaMatK, amplified using the PCR primers AmaITS forward (5 0 -GGACTTGTAGTG-GGTTAA-3 0 ), AmaITS reverse (5 0 -GCAGTTTCAGAATGGTAA-3 0 ), AmaMatK forward (5 0 -AGATCCTTATCTAAAGCTAAA-3 0 ), and AmaMatK reverse (5 0 -ACCAAACCGATCAATAATTA-3 0 ) included the target region of the primer and the DLP probe sets designed in this study. Both amplicons were cloned into a TOPO PCR4 plasmid vector (Invitrogen, Carlsbad, California, U.S.A.) and chemically transformed into Escherichia coli TOP 10 competent cells (Invitrogen) cultured on Luria-Bertani (LB) agar plates (Gibco; present ThermoFisher Scientific, Waltham, Massachusetts, U.S.A.) overnight. After this period, several colonies were selected and checked by colony direct-PCR using the primers for each of the inserted PCR amplicons. Positive colonies were further cultured in 2 3 LB broth overnight and plasmids were extracted using the Qiaprep plasmid purification Kit (Qiagen). Plasmid concentration was measured with the GeneQuant II (Amersham Pharmacia, Uppsala, Sweden) spectrophotometer, and the number of gene copies in the plasmid (molecules lL 21 ) was calculated as, A 3 6.022 3 10 23 )/(660 3 B), where A is the plasmid concentration (g lL 21 ), B is the plasmid length containing the cloned sequence, 6.022 3 10 23 is the Avogadro's number, and 660 is the average molecular weight of one base pair. Plasmid standards subcloned with each target sequence were diluted 10 times and amplified using the primers and DLPs designed in this study. In all experiments, negative controls containing same sample volumes of sterile 10 mM Tris-buffered water (pH 7.6) were tested in six replicates. The qPCR was performed as described in the "Designed primers and DLPs specificity" section. We also checked the reproducibility by using the above plasmids 10-fold diluted from 10 6 to 10 0 .

Quantitation of genomic DNA in Z. marina parts
The third leaf, youngest leaf, sheath, meristematic region, rhizome, roots, and pollen of Z. marina were collected and freeze-dried. After treatment by the TissueLyser (Qiagen), DNA was extracted from 50 mg of each sample and from pollen traces using the DNeasy plant Mini Kit (Qiagen), as described in the "Designed primers and DLPs specificity" section. Gene copy number per ng of extracted DNA was determined by qPCR using the plasmid standard curve obtained in the "Plasmid development for the construction of the standard curve" section.

DNA extraction from coastal sediments
The commercial DNA extraction kits Qiamp stool Mini Kit (Qiagen), Isoil (Nippon Gene, Tokyo, Japan), Nucleospin from Soil Kit with buffer systems SL-1 and SL-2 (MACHEREY-NAGEL, Deer Park, New York, U.S.A.), and DNeasy Plant Mini Kit (Qiagen), were used to extract DNA from sediment samples, mainly following the manufacturer's instructions, and their performances were compared. Surface sediment samples used here were collected from the Z3 sites (see below), freeze-dried, and disrupted using TissueLyser (Qiagen), according to Shimabukuro et al. (2012). Each extraction kit was used to extract DNA from approximately 100 mg of powdered sediment. Gene copy number per gram of sediment was determined by qPCR using the plasmid standard curve obtained in the "Plasmid development for the construction of the standard curve" section. For qPCR, six replicates of each sample were run.

Application of qPCR and ddPCR to detect Z. marina DNA in coastal sediments
The long-core sediment samples collected from seagrass meadows (Ikunojima Island, Hiroshima, Japan; Z3) and nonvegetated tidal-flat (Nakatsu, Oita, Japan; E1) and described in Miyajima et al. (2015) were used for the assessment and crosscheck of the newly developed ddPCR-based method and conventional duplex qPCR method. Both sites are located in the Seto Inland Sea. Portions of the same batches of sediments as those used by Miyajima et al. (2015) for determining OC content, stable carbon isotope ratio (d 13 C), Tm, melting temperature. and 14 C age, were used in the present study. To minimize the risk of contamination, a strict protocol of sample manipulation in the laboratory based on Bissett et al. (2005) and Lejzerowicz et al. (2013) was developed and followed. Sediment samples were prepared and treated in a laboratory separate from that where plant samples were treated and analyzed. Sediment subsamples, which were collected from the center of core slices using sterile spatulas, were freezedried and disrupted by TissueLyser (Qiagen); 100 mg of each powdered sediment sample were then subject to DNA extraction using the DNeasy Plant Mini Kit (Qiagen). The concentration of total DNA was determined in a GeneQuant II (Amersham Pharmacia) spectrophotometer and the number of Z. marina gene copies per gram of sediment was determined by qPCR using the plasmid standard curve obtained in the "Plasmid development for the construction of the standard curve" section. All samples were run in triplicate in the qPCR. The absolute quantitation of Z. marina DNA was obtained from a ddPCR using the primers and DLPs used in qPCR. Before the assay, the ddPCR was optimized for AmaITS and AmaMatK. ddPCR mixtures contained 10 lL 23 ddPCR TM Supermix for Probes (Bio-Rad), 1 lM each primer, 0.25 lM each probe, 2 lL template DNA, and sterilized pure water up to 20 lL total volume. The quantification assay of the target DNA was carried out in a QX100 ddPCR system (Bio-Rad). After transferring the 20 lL ddPCR mixture to a droplet generator cartridge (Bio-Rad), 70 lL droplet generation oil was added to oil wells and the cartridge was covered with a rubber seal and loaded onto the QX100 droplet generator. The emulsions resulting from the droplet generator were then transferred to a 96-well PCR plate, which was sealed with Pierceable Foil Heat Seal by the PX1 PCR plate sealer (both Bio-Rad), and then subjected to amplification in a C1000 Touch thermal cycler (Bio-Rad). This amplification was carried out on a regular ramp rate of 2.58C/s, at 958C for 10 min followed by 40 cycles of 30 s at 948C plus 1 min at 548C; a final enzyme deactivation step occurred at 988C for 10 min. The fluorescent signals of FAM and HEX from amplified droplets were captured individually in the QX100 Droplet Reader and analyzed with QuantaSoft 1.6.6 software (both Bio-Rad). Positive and negative droplets were distinguished by threshold values, automatically obtained from different sample types. Target DNA concentration was reported as the number of gene copies per microliter of PCR mixture after correction with the Poisson distribution.

Statistical analyses
Statistical analyses, using Welch's t-test, nonparametric Kruskal-Wallis, and Type I regression, were performed in R 3.1.0 (R Core Team 2014) within SPSS 22.0 (IBM, Armonk, New York, U.S.A.) and p values lower than 0.05 were considered statistically significant.

AmaITS and AmaMatK optimal annealing temperatures
The amplicons of AmaITS and AmaMatK were 90 bp and 102 bp in length, respectively (Table 1). The cycle threshold (Ct) values of the both amplicons gradually increased over 59.18C (Table 2) and thus, considering detection specificity, 588C was chosen as the annealing and extension temperature for qPCR.

AmaITS and AmaMatK specificity
The in silico verification of AmaITS and AmaMatK specificity by BLAST searches indicated they were highly specific. The evaluation using qPCR and the DNA extracted from the two different Zostera species and wild and cultivated terrestrial vascular plant species from the Seto Inland Sea area revealed that AmaITS and AmaMatK only reacted to Z. marina thus being highly specific (Table 3).

Reproducibility of qPCR
Reproducibility was assessed using twenty replications from 10 6 -10 0 gene copies of both plasmid standard curves. Percent coefficient of variation (CV) of Ct values was 0.6-2.1 in AmaITS and AmaMatK (Table 4). Although the duplex qPCR systems developed in this study showed high reproducibility, this was slightly lower for gene copy numbers below 10 2 in AmaITS and 10 1 in AmaMatK regions.

Quantitation of genomic DNA in Z. marina parts
Overall, the gene copy number of AmaITS was approximately 10 times greater than that of AmaMatK in all parts of Z. marina. The highest number of gene copies for AmaITS was found in the lower part of the leaf sheath, although differences were not significant among all parts analyzed (Fig.  1). In contrast, the gene copy number of AmaMatK was significantly lower in the rhizomes and roots (Kruskal-Wallis, n 5 6, p < 0.01) of Z. marina. Thus, if only chloroplast gene copy numbers were evaluated in coastal sediments, the contribution of Z. marina would be underestimated. The ITS gene was present with a high copy number in ribosomal RNA clusters, and the results showed that this gene is easily detected even in small quantities of DNA. Pollen was haploid and thus contained fewer gene copies than the other parts (Kruskal-Wallis, n 5 6, p < 0.01), although AmaITS and Ama-MatK presented nearly identical gene copy numbers in the pollen.

DNA extraction methods for detection of Z. marina DNA from sediments
The gene copy numbers in the DNA extracted from sediment and plants were compared for four commercial DNA extraction kits, and the two buffer systems from the Nucleospin from Soil Kit. It was found that DNA was extracted from sediments as well as from plants most efficiently by the DNeasy Plant Mini Kit. In addition, gene copy numbers were significantly different among samples extracted using the Qiamp Stool Kit, Isoil, and Nucleospin from Soil Kit using buffer system SL-2 (Kruskal-Wallis, n 5 6, p < 0.05; Fig. 2). The gene copy number obtained in samples extracted with the DNeasy Plant Mini Kit was five times higher than that of samples extracted using Nucleospin from Soil Kit with buffer system SL-1, although this difference was not statistically significant. Thus, the DNeasy Plant Mini Kit was used in the following experiments, also because the DNeasy Plant Mini Kit is less costly than the Nucleospin from Soil Kit.
Detecting Z. marina genes in coastal sediments by qPCR and ddPCR The efficiency of the duplex reaction with AmaITS and AmaMatK in the ddPCR decreased over 548C and thus this was chosen as the annealing and extension temperature for the ddPCR. The quantification using qPCR was cross-checked with the absolute numbers obtained in the ddPCR, and the gene copy numbers of Z. marina AmaITS and AmaMatK genes, excluding the top 10 cm of sediment samples whose gene copy numbers were extremely high, showed a significant and positive correlation between these two methods (AmaITS: n 5 29, r 5 0.985, p < 0.01; AmaMatK: n 5 18, r 5 0.950, p < 0.01) with high R 2 values (0.970 in AmaITS and 0.902 in AmaMatK, p < 0.01 in both; Fig. 3). Thus, the qPCR developed in the present study was highly accurate and reliable. However, the number of data below the detection limit was greater in ddPCR than in qPCR, for both AmaITS and AmaMatK, in the E1 samples. The correlation between qPCR and ddPCR was relatively low for samples with low gene copy numbers because qPCR often overestimated the copy numbers in these samples. One of the advantages of ddPCR was its higher accuracy at low concentrations of target DNA compared to qPCR. Thus, ddPCR was used in the subsequent analysis of ancient DNA in the long-core samples from Z3  and E1. The vertical distributions of AmaITS and AmaMatK gene copy numbers (Fig. 4) showed that the sediments from Z. marina meadows (Z3) contained approximately 10-2000 times more copies of both genes than the sediments of the unvegetated tidal flat E1. The gene copy number at the top 30 cm of Z3 was approximately 2000 times higher than that of E1 (Welch's t-test, n 5 6, p < 0.01). In Z3, a downward decreasing trend in gene copy numbers was observed down to approximately 1 m depth (Fig. 4), and the downward decreasing rate of AmaMatK was faster than that of AmaITS. Below 1 m, gene copy numbers stabilized, and the AmaITS gene could be detected even in the layer dated to approximately 5000 calibrated years before present (yr cal BP) using radiocarbon dating (Miyajima et al. 2015). Finally, the correlation between the gene copy numbers of AmaITS and Ama-MatK and the concentration of OC derived from seagrass (Miyajima et al. 2015) was confirmed for the samples from Z. marina meadows Z3 (AmaITS: n 5 28, r 5 0.789, p < 0.01; AmaMatK: n 5 21, r 5 0.943, p < 0.01) and the R 2 was 0.623 for AmaITS and 0.889 for AmaMatK (Fig. 5), but not for those from nonvegetated tidal-flat E1 (Fig. 5).

Discussion
The first step of sequestration of OC in coastal sediments begins with a senescence of Z. marina. In higher plants like Z. marina, apoptosis occurs in the senescent seagrass cells before abscission, and their DNA molecules may be digested into about 200 bp fragments (Pietramellara et al. 2002). After abscission, most of above-ground part of the seagrass flows away from original meadow as floating materials. The lysis of dead parts of the seagrass make up the main source of DNA in the underlying seagrass meadow sediments. Chlorophyll, proteins, and RNA molecules are enzymatically degraded during senescence of plant leaf (Thomas and Stoddart 1980;Green 1994). A part of DNA molecules may escape these intracellular degradations and be released from decomposing tissues as demonstrated for the cases of tobacco and tomato leaf cells (Pot e et al. 2005(Pot e et al. , 2007. The released DNA molecules will be gradually degraded by ultra violet or DNase produced by bacteria and fungi in coastal sediments (Pietramellara et al. 2009). Lejzerowicz et al. (2013) reported the size of DNA fragments by using foraminifera-specific PCR primers for deep-sea sediments. The size of successfully amplified fragments ranged from about 1000 bp at the surface to about 100 bp in lower layers. As mentioned above, DNA fragmentation induced by apoptosis in senescence Z. marina cells would lead to fragmentation of DNA into about 200 bp. Therefore, we designed the two DLP sets (AmaITS and AmaMatK) that amplified 90 bp and 102 bp of the amplicon, respectively.
The detection efficiency of Z. marina DNA by AmaITS was higher than by AmaMatK because the number of gene copies was greater for AmaITS than for AmaMatK in all parts of Z. marina analyzed. Although the number of gene copies of AmaITS was similar in all parts of Z. marina, that of Ama-MatK was drastically lower in belowground parts, i.e., in the rhizome and roots, than in aboveground parts. In vascular Fig. 2. Gene copy numbers of AmaITS and AmaMatK detected to the DNA extracted from the sediments of seagrass meadow Z3 with several commercial kits. Error bar indicates standard deviation. "a" indicates significant difference (p < 0.05) among gene copy numbers of AmaITS in the samples extracted using the Qiamp Stool Kit, Isoil, and Nucleospin from Soil Kit using buffer system SL-2 based on Kruskal-Wallis H test. "b" indicates significant difference (p < 0.05) among gene copy numbers of AmaMatK in the samples extracted using the Qiamp Stool Kit, Isoil, and Nucleospin from Soil Kit using buffer system SL-2 based on Kruskal-Wallis H test.

Fig. 3. Correlations between the gene copy numbers of AmaITS and
AmaMatK determined by qPCR and ddPCR in sediment samples collected from Ikunojima and Nakatsu.
plants, chloroplast genes are regulated by an intricate network of genes and by environmental stress (Sugita and Sugiura 1996;Zergaes 2016). Light intensity, for example, stimulates the expression of chloroplast-related genes, which is upregulated in the aboveground portion and downregulated in the belowground portion of a vascular plant. Rhizomes and roots play a substantial role in OC sequestration because these parts are directly accumulated in Z. marina meadows sediments. In some Mediterranean seagrass meadows, Posidonia oceanica constructs a robust root system and is exceptionally efficient in accumulating organic matter in the underlying sediment (Mateo et al. 1997). The root systems of Z. marina are not as robust as those of P. oceanica, although, under reducing conditions, its rhizomes and roots may remain in coastal sediments for long periods. The abundance of Z. marina-derived DNA in coastal sediments would be underestimated if only chloroplast-related genes such as AmaMatK were examined. Ceccherini et al. (2003) reported a 98% loss of the chloroplast aadA gene in ground tobacco leaf material after 72 h compared to the 56% loss of total DNA. This suggests that the chloroplast gene AmaMatK decomposes faster than the nuclear AmaITS gene and, therefore, nuclear gene fragments such as AmaITS are a superior tracer for Z. marina-derived organic remains in coastal sediments to chloroplast gene fragments such as AmaMatK.
Several DNA extraction methods from soil and sediments have been reported (e.g., Robe et al. 2003;Zhao and Xu 2012), but most were originally developed for identification of bacteria and survey of bacterial flora. As mentioned above, macroscopic debris derived from seagrass tissues are likely to remain abundant in seagrass meadow sediments, and DNA fragments may persist in such sediments, bound to such Fig. 5. Relationship between seagrass-derived OC and gene copy numbers of AmaITS and AmaMatK in Ikunojima (Z3) and Nakatsu (E1) sediment cores. The data points below the dashed line in the graphs of site E1 correspond to the data below the detection limit of ddPCR; the OC derived from seagrass is shown as reference.

Fig. 4. Vertical distributions of AmaITS (closed circle) and AmaMatK
(open circle) gene copy numbers and the concentration of OC derived from seagrass, estimated by the d 13 C-based provenance analysis using the 3-endmember model in Miyajima et al. (2015) in Ikunojima (Z3) and Nakatsu (E1) sediment cores (green line).
debris. Consequently, the DNeasy plant Mini Kit, which is optimized for extraction of DNA from plant materials, was tested in the present study for extraction of Z marina-derived DNA from seagrass meadow sediments and compared with some extraction kits specific to soil samples. Indeed, the recovery of DNA from the surface sediments of Z. marina meadows in the Seto Inland Sea was the highest with the DNeasy plant mini kit, and this was also the case for the samples of deeper (> 1 m) cores Z3 and E1. Therefore, DNA extraction methods or kits designed for plants are appropriate to extract Z. marina DNA from coastal sediments.
The OC concentrations at the top 30 cm of sediment cores from Z3 and E1 were 830 lmol g 21 and 130 lmol g 21 , respectively (Miyajima et al. 2015). However, the gene copy numbers of both AmaITS and AmaMatK at the top 30 cm of sediment cores were approximately 2000 times higher in Z3 than in E1. The amount of the DNA gradually decreased downward in sediment cores down to about 1 m below the surface. Miyajima et al. (2015) described a downward decreasing trend in the OC observed in Z3 sediment cores down to 1 m below the surface. These diagenetic losses of OC were caused by bacterial or fungal remineralization (Pollard and Moriarty 1991). Two thirds of the OC seemed to have been lost in the top 1 m, which corresponded to a period of approximately 2500 yr in core Z3 (Miyajima et al. 2015). In contrast, gene copy number decreased two orders of magnitude from 0 m to 1 m depth, and DNA loss was much faster than OC loss. However, below 1 m, the gene copy number stabilized, which might be related to the adsorption of DNA molecules onto surface-active particles, such as clay minerals and humic substances, in coastal sediments as this has been referred to as a key factor to the stabilization of DNA fragments in deeper sediments (Stotzky 2000). The DNA molecules bound to these substances are physically and chemically protected against degradation by nucleases (Niemeyer and Gessler 2002;Cai et al. 2006aCai et al. , 2006bCai et al. , 2006cNielsen et al. 2006Nielsen et al. , 2007. Moreover, Miyajima et al. (2017) reported that fine mineral particles, associated with organic matter, were abundant in the sediment of Z. marina meadows compared to estuarine sandy sediments. They concluded that the accumulation and retention of OCrich fine sediment particles enhanced by the seagrass was one of the principal factors for OC sequestration. Most of the DNA in sediments deeper than 1 m might have been adsorbed to fine mineral particles and accumulated in sediments older than 2500 yr cal BP.
The AmaITS gene was detected in sediment samples dated ca. 5000 yr cal BP, which coincides with the formation of the current shoreline of the Seto Inland Sea (Sato 2008). Therefore, seagrass meadows seem to have appeared in this area almost as soon as the shoreline had reached the current level. The sea level around Japan gradually rose from the end of the last glacial maximum to mid-Holocene (approximately 6000-7000 yr cal BP) and was about 4 m higher than the current sea levels (Tanigawa et al. 2013). Since the Holocene marine transgression, named "Jomon marine transgression," the sea level around Japan gradually decreased, reaching the current level approximately 5000 yr cal BP. Thus, Z. marina meadows in Ikunojima (Z3) seem to have continuously accumulated sediments since around 5000 yr cal BP. Mateo et al. (1997) reported that the organic deposits produced by P. oceanica accumulated in Mediterranean sediments for several millennia and the present study showed that the organic deposits produced by Z. marina also persisted in Seto Inland Sea sediments for several millennia. These results strongly suggest that seagrasses, including Z. marina, make a significant contribution to the long-term sequestration of OC.
This study revealed that the gene copy numbers of AmaITS and AmaMatK had a moderate and high correlation to the concentration of seagrass-derived OC in the seagrass meadows (Fig. 5). Thus, the DNA-based trace techniques used here are a promising tool to trace and quantify seagrass-derived OC sequestered in marine sediments. However, the present study also showed some inconsistencies between the Z. marina gene copy numbers and d 13 C-based estimation of seagrass-derived OC. In particular, the gene copy number in nonvegetated tidal flat sediments was often under the detection limit of the current techniques, even with the ddPCR technique, albeit the presence of seagrass-derived OC was suggested by the d 13 Cbased analysis for the same samples. Most likely, the reason for this inconsistency is the overestimation of seagrass-derived OC by the conventional d 13 C-based model. In nonvegetated well-illuminated tidal flats, the sediment surface is often covered with a thin film of abundant microphytobenthos, whose d 13 C is close to that of seagrasses. If the OC derived from microphytobenthos considerably persists in the sediment, this might cause a significant positive bias in the estimation of seagrass-derived OC by the d 13 C-based model. Therefore, the DNA-based tracer technique developed in the present study can be used as a precise and complementary method to verify and support the provenance of the OC estimated by conventional isotope-based methods. Still, further studies are needed to clarify the relationship between OC and DNA in marine sediment cores of Z. marina and other seagrass meadows.
Z. marina DNA was successfully identified in the sediments collected from Seto Inland Sea coastal areas by both qPCR and ddPCR. Numerous studies have compared the detection efficiency and quantifiability between qPCR and ddPCR (Hayden et al. 2013;Hindson et al. 2013;Bharuthram et al. 2014;Boizeau et al. 2014;Kim et al. 2014aKim et al. , 2014bSze et al. 2014;Wiencke et al. 2014;Yang et al. 2014;Doi et al. 2015), and all have concluded that qPCR results were highly correlated with ddPCR results. The ddPCR has often been used in clinical practice for diagnosing infectious diseases and oncogenes, where accurate diagnosis is needed. Quantitative infectivity or diagnostic assays should have adequate sensitivity and reproducibility and be performed with sufficient replicates to ensure adequate statistical validity of the results. In general, ddPCR is more sensitive than qPCR when the sample DNA quality is poor or the concentration of target DNA is low (Henrich et al. 2012;Beaver et al. 2014). Also ddPCR performs better than qPCR because the background noise of negative controls or of samples with the low concentrations of target DNA is reduced with the former method (Sze et al. 2014). Stults et al. (2001) reported that many PCR inhibitors present in sediments samples affect qPCR quantification. Another advantage of ddPCR over qPCR is its lower susceptibility to PCR inhibitors (Dingle et al. 2013;Racki et al. 2014). Doi et al. (2015) reported that errors associated with the ddPCR analysis at low eDNA concentration in aquatic systems were smaller than that with qPCR. Kim et al. (2014a) described that ddPCR was more sensitive than qPCR for examining bacterial population dynamics in soils. For the above stated reasons, ddPCR is considered more suitable for the detection of Z. marina DNA from marine sediments than qPCR. However, ddPCR has two significant disadvantages: (1) target DNA concentrations should be less than 20,000 copies in reaction mixtures for proper quantification; and (2) ddPCR is more expensive and time-consuming in operation than qPCR. Thus, ddPCR and qPCR should be selected depending on the purpose of the survey. Our results suggest that ddPCR will be suitable for sediment core samples collected from unvegetated tidal flats or from ancient sediment cores because these samples have low gene copy densities, while seagrass meadow sediments from shallow depths can be analyzed using qPCR. When Z. marina gene copy number is to be used as a paleo or millennial seagrass indicator, ddPCR will be the best method.
The seagrass DNA detection method developed in this study have significant advantages for models of dynamics of Z. marina-derived carbon across broader scales. First, this method can provide at least quantitative information about the fraction of seagrass-derived OC among the OC stored in sediment (Fig. 5). Many available models for carbon sequestration in seagrass bed sediments do not specify the origin of OC. Actually, the OC stored in seagrass bed sediments are derived from multiple sources including both autochthonous and allochthonous ones. Some analytical techniques such as stable isotope ratios and lipid biomarkers can be used to evaluate relative importance of some representative OC sources, however, information provided by these conventional tools is often ambiguous and not quantitative, especially when there are many potential sources of OC to the sediment. In particular, it is often very difficult by these tools to trace seagrass-derived OC that is exported from the original meadows and deposited and stored in distant offshore sediments. In contrast, the DNA-based method can provide clear and definitive evidence for the presence of seagrass-derived OC. Moreover, as we demonstrated in this study, the seagrass DNA can be recovered from deep, old sediments of up to 5000 yr cal BP, which implies that the DNA-based technique can be used effectively to reconstruct the history of propagation and decline of seagrass beds and trace the fate of OC carried away from lost or degraded seagrass beds. As an example, the technique would enable us to demonstrate the differences in the OC sedimentation between the area with seagrass distribution (DNA present) and the area without seagrass distribution (DNA absent) in both the present and the past, which is essential to estimate the OC sequestration by seagrass beds in the historical fluctuation of its distribution.

Comments and recommendations
Costs and time consumption Te et al. (2015) reported that the cost to analyze a full 96well plate was US$156.00 for qPCR and US$470.00 for ddPCR in the duplex assay, while, in Japan, the cost to analyze a full 96-well plate was US$73.00 for qPCR and US$663.00 for ddPCR. These comparisons suggest that qPCR is much cheaper than ddPCR. For 96 samples, qPCR included 45 min for the preparation of reaction mixtures and 80 min for amplification, while ddPCR included 45 min for the preparation of reaction mixtures and droplets, 120 min for amplification, and 120 min for plate reading. Thus, Z. marina DNA in coastal sediment core samples was detected using qPCR in this study. As ddPCR is an absolute quantitation method, and less susceptible to PCR inhibitors than qPCR, we recommend using it for crosschecking purposes and for the sensitive detection of DNA from such samples as sediment cores collected from unvegetated tidal flats or deep millennial levels.

Development of duplex and multiplex assays
For detecting Z. marina DNA in coastal sediments, we used a duplex assay. Although the qPCR system could detect up to five fluorophores, multiplexing with sample volumes as small as 10 lL, the ddPCR system could detect only two fluorophores (FAM and HEX). Therefore, we dual-labeled our original two probes (AmaITS and AmaMatK) with FAM and HEX. Duplex or multiplex assays are powerful tools that improve the accuracy of molecular investigations regarding Z. marina in a coastal environment and, therefore, nuclearand chloroplast-related genes were used in duplex quantitation by qPCR and ddPCR. However, gene copy numbers of chloroplast-related genes differed among Z. marina parts. Therefore, we recommend using nucleic multicopy genes for more accurate duplex and multiplex quantitations of seagrass DNA in coastal sediments.