Introduction

Efficient management of fisheries resources requires prompt and precise evaluation of the spatial distribution and abundance of target species. Traditional methods for this purpose, such as net-towing, visual census, and acoustic survey, are generally effort-demanding (Murphy and Jenkins 2010). Environmental DNA (eDNA) technology can serve to evaluate the presence and abundance of aquatic animals or plants by sampling environmental water and detecting DNA fragments released from organisms. The source of eDNA can be epidermis, mucus, feces, or urine of species (Turner et al. 2014; Klymus et al. 2015; Wilcox et al. 2015). eDNA is typically detected by quantitative PCR (qPCR), using a set of species-specific primers and probe. Typical eDNA surveys can be performed by collecting only a liter or less of water (Ficetola et al. 2008; Takahara et al. 2012; Thomsen et al. 2012a, b). Advances in such methods are expected to promote less costly and non-invasive monitoring of the ecosystem for conservation and biomass estimation. Thus, eDNA technology has a great potential for the management of aquatic resources.

eDNA was first applied to vertebrates to detect the invasive American bullfrog Rana catesbeiana in Europe (Ficetola et al. 2008). Since then, it has been used to identify habitats of rare species (Thomsen et al. 2012b; Pilliod et al. 2013; Rees et al. 2014; Fukumoto et al. 2015) and to indicate other introduced species’ invasions (Takahara et al. 2013; Piaggio et al. 2014). Recent studies have shown a positive relationship between eDNA concentration and biomass in experimental surveys, e.g., Takahara et al. (2012), and field surveys in freshwater (e.g., Thomsen et al. 2012b; Pilliod et al. 2013; Eichmiller et al. 2014; Wilcox et al. 2016; Doi et al. 2017). eDNA studies have also been conducted in the marine environment (Thomsen et al. 2012a; Valentini et al. 2016; Kelly et al. 2014). Yamamoto et al. (2016) demonstrated that eDNA concentration can be used to assess the biomass of jack mackerel Trachurus japonicus, with findings confirmed using sonar. eDNA metabarcoding using universal primers enables fish fauna to be comprehensively and simultaneously surveyed (Miya et al. 2015). Yamamoto et al. (2017) detected 128 fish species by eDNA metabarcoding of water samples collected during a 1-day cruise, including 40 species observed by underwater visual census conducted in this area during the preceding 14 years. In experimental studies, the sequence reads of fish and amphibians in eDNA metabarcoding correlated with their biomass (Kelly et al. 2014; Evans et al. 2016). In the sea, the species detected by eDNA metabarcoding matched trawling catch data, and eDNA sequence read numbers were correlated with abundance data (Thomsen et al. 2016). Quantification in eDNA metabarcoding has recently been established by simultaneously using artificial DNA internal standards with known concentration (Ushio et al. 2018). As such, eDNA is being developed as a suitable tool across many fields for estimating both biomass and species diversity.

eDNA released by marine or freshwater organisms does not remain at the site of its release, but is dispersed and degraded over time. Further, information as to the timing and source of eDNA release cannot be resolved based on the eDNA itself. Thus, a firm understanding of eDNA dynamics in nature constitutes essential information for estimating distribution and abundance of aquatic organisms. Dispersion or degradation of eDNA in the natural environment has been reported in rivers (Deiner and Altermatt 2014; Jane et al. 2015; Sansom and Sassoubre 2017), but there is limited information for the marine environment (Yamamoto et al. 2016; O’Donnell et al. 2017). Owing to the multi-directional currents present in oceanic habitats, the processes of eDNA dispersion and degradation may be more complicated than those in rivers or ponds.

The present study aimed to show the spatiotemporal scale of dispersion and degradation of eDNA in a coastal marine system, using caged live striped jack Pseudocaranx dentex as a known source of eDNA in Maizuru Bay, the Sea of Japan. This species was selected as it was presumed to be absent in this area based on results of long-term underwater visual census and eDNA metabarcoding in this area (Masuda 2008; Yamamoto et al. 2017). A cage containing live fish was set off the end of a pier, and eDNA dispersion and degradation were periodically estimated with spatiotemporal monitoring.

Materials and methods

Field experiment

Hatchery-reared striped jack juveniles were purchased from the Aquaculture Research Institute of Kindai University and transported to Maizuru Fisheries Research Station (MFRS), Kyoto University on June 5, 2015. Forty-nine fish with a cumulative biomass of 426.9 g were stocked in a covered mesh cage (1.15 × 1.15 × 1.15 m, 7 mm mesh size) in a 500-l black polyethylene tank. The tank water was recirculated using four sets of aerators equipped with filtering materials (Roka Boy M, GEX, Osaka, Japan). Fish were fed with pellets (Otohime, Marubeni Nisshin Feed Co., Ltd., Tokyo, Japan) beginning 2 days after stocking.

On June 9, 2015, a cage was set ~ 1 m from the surface at the tip of a 30-m floating pier at the research station (35°28′N, 135°22′E). The area is in a protected inner bay. The sea bottom is muddy silt with a small number of artificial structures such as discarded concrete blocks. The water depth where the cage was set is ~ 8 m. The fish were not fed during the experiment. Sampling stations were set above the cage ~ 1 m from the school of fish and at six linear distant points extending northwest and northeast (six in each direction; Fig. 1). These stations were named St. 1, W10–W1000 and E10–E1000, where “W” or “E” and the number of each station name represent northwest or northeast and distance in meters from the cage, respectively. We took a course of first going to W1000, then along the transect line to W10, E10, E1000 and then back to the port. Care was taken not to run the boat across sampling stations except during sampling. Sampling at St. 1 was conducted directly from the floating pier so as not to disturb fish when the boat was between W30 and W10. Two-liter samples of sea surface water were collected using a ladle (1.4 m in length, 3.2 l in capacity) in triplicate at each station at roughly 0, 1, 2, 4, 8, 24 and 48 h after setting the cage, as well as 40 min prior at St. 1 as a control. It took 22–34 min to visit all the stations and conduct water sampling. Therefore, the actual timing of each sampling was on average ~ 14 min after the above-mentioned ‘hours’ after setting the cage. These samples were placed in a cooling box on ice to minimize the degradation of eDNA until filtration was performed.

Fig. 1
figure 1

Location of caged striped jack (St. 1) and water sampling stations along the northwest and northeast transect lines. Two triangles represent where current flow meters and thermometers were deployed. A thermometer was also set at St. 1. Circle size indicates total eDNA concentration over all time points

Full size image

The cage was removed from the sea after 48 h, and sampling was repeated at 0, 1, 2, 4, 8 and 24 h after this point. Sampling was also conducted at St. 1, 48 h after removing the cage. Current flow meters (COMPACT-EM, JFE Advantech, Hyogo, Japan) were set NW and NE of the cage, and conductivity-temperature meters (INFINITY-CT, JFE Advantech, Hyogo, Japan) were placed NW, NE, and at St. 1 at a depth of 2 m from the surface (Fig. 1).

Vertical dispersion of eDNA was also examined to confirm the validity of sampling surface water in studying horizontal dispersion. Triplicate water samples were collected using a van Dorn sampler at a distance of ~ 1 m from the edge of the cage, at the sea surface, midwater (4 m) and bottom (8 m) at 29, 32 and 37 h after setting the cage. The timing of these samplings did not overlap with those of horizontal dispersion.

Filtration and DNA extraction from water samples

Filtration was conducted in the order of water sampling. We started to filter samples within ~ 40 min (30 min on board and 10 min for transportation to the laboratory) after collection using 47-mm-diameter glass microfiber filters (nominal pore size 0.7 µm, GE Healthcare Life Science, Whatman) through an aspirator. It took ~ 30 min to complete filtration per one sampling time point. Filters containing trapped eDNA were folded, with the surface in contact with the sample facing inward, then wrapped with aluminum foil and preserved at − 20 °C. Samples from northwest and northeast transects were filtered in separate laboratories to avoid cross-contamination, with each transect provided to the opposite lab at each time point to negate handling bias. All instruments used for filtration were chlorinated using 0.1% hypochlorous acid to degenerate remnant DNA and were rinsed with tap water and distilled water prior to the next filtration step. The amount of water filtered was 1 litre for each sample. The same amount of artificial sea water (MARIN MERIT, Matsuda, Osaka, Japan) was filtered once per series of samples in each transect as a negative control. DNA was extracted according to the procedure detailed in Yamamoto et al. (2016).

Species-specific primers and probe for striped jack

Sequences of mitochondrial cytochrome b (cyt b) were obtained from the National Center for Biotechnology Information (NCBI) for Pseudocaranx dentex (Sequence ID: EF392607) and five other carangid fishes found in this area, including: Trachurus japonicus (AP003091.1), Decapterus maruadsi (EF512291.1), Seriola quinqueradiata (AB263290.1), Seriola dumerili (EF439584.1), and Kaiwarinus equula (KM201334.1). We searched for an appropriate primer binding site of five base pairs with a pair unique to striped jack at the 3′-end, and designed a primer targeting this region (Fig. 2). The sequences of primer pairs and probe were as follows: 5′-ACGCAGTTTTTATTCTGGC-3′ for the forward primer, 5′-CTAGGAAGTATAGGGC-3′ for the reverse primer, and 5′-FAM-CGTGGCTATCCTCACCTGAATCGGG–TAMRA-3′ for the probe. The PCR product size was 130 bp. Only striped jack DNA was amplified when qPCR was conducted using DNA extracted from tissue at levels of 10 and 100 pg per reaction with the above six species as template. Real-Time TaqMan® PCR (Life Technologies) was performed to measure the copy number of striped jack cytochrome b using a LightCycler 96 System (Roche Diagnostics, Mannheim, Germany). The total volume of qPCR reaction solution was 20 µl, including 2 µl DNA extract, 900 nM of each primer (final concentration), 125 nM TaqMan probe (final concentration), and 10 µl TaqMan master mix (TaqMan Gene Expression Master Mix, Life Technologies). PCR was performed according to the following thermal cycling: 2 min at 50 °C and 10 min at 95 °C, followed by 55 cycles of 15 s at 95 °C and 1 min at 60 °C. qPCR was performed with both eDNA samples and the artificial DNA fragments as a dilution series containing 3 × 101 to 3 × 104 copies. The artificial DNA of striped jack cytochrome b was pUC57 plasmids containing 330 bp of partial cyt b gene, or 130 bp of the PCR target and 100 bp of upstream and downstream regions. Prior to qPCR analysis, the plasmid was digested with a restriction enzyme (Hind III). qPCR was performed in triplicate for each eDNA sample as well as the artificial DNA. In all the runs, R2 values of the calibration curve were greater than 0.99, with the slope range between − 4.190 and − 3.350 and an intercept range between 39.66 and 43.36. In addition, a 2-µl pure water sample was analyzed simultaneously in triplicate as a PCR negative control. Different laboratory rooms were used for DNA extraction and PCR to prevent cross-contamination. The average of qPCR triplicates (including the template with 0 copies) was taken to represent eDNA concentration (number of copies per liter) in each sample.

Fig. 2
figure 2

Sequence alignment of striped jack and five other carangid fishes. The target region is a mitochondrial cytochrome b gene fragment, with a forward primer, b reverse primer, and c TaqMan probe. An asterisk indicates a site specific to striped jack. The arrow indicates the direction of DNA synthesis

Full size image

To confirm the specificity of the primer and probe design, the sequence of qPCR products that showed amplification by qPCR was analyzed and BLAST searching was conducted using the NCBI nucleotide database. In the qPCR, 64 water samples showed amplification. Sequencing revealed that 61 samples that were successful in the read out of the total 64 samples matched with the striped jack cyt b gene sequence (Sequence ID: EF392607.1). Therefore, we assumed that the designed primers and probe set were specific for striped jack.

Statistical analyses

Prior to statistical analysis, eDNA concentration was log10 (x + 1) transformed to improve homogeneity of variance. Since the vertical dispersion experiment showed that most eDNA can be assumed to remain near the surface and disperse horizontally, we applied a power regression between the distance from the cage and eDNA concentration to evaluate the dispersion. A coefficient of determination was obtained for each time point. We compared eDNA concentrations among sampling stations along the northwest or northeast transect lines using a nested ANOVA, with nesting after setting or removing the cage and time points, followed by Tukey’s HSD test. eDNA concentration was compared among time points after setting or removing the cage by nested ANOVA, with nesting in sampling stations, followed by Tukey’s HSD test. Then, we calculated the detectability of eDNA concentration at each sampling station after removing the cage as percent eDNA remaining per hour, using the average of eDNA concentration at 0 and 1 h time points after removing the cage, as follows:

$$ {\text{Dtime}}\; = \;100\; \times \;{{\text{A}} \mathord{\left/ {\vphantom {{\text{A}} {\text{B}}}} \right. \kern-0pt} {\text{B}}}, $$

where A is the average eDNA concentration 1 h after removing the cage, B is the average eDNA concentration immediately after removing the cage, and Dtime is detectability (%/h). Detectability of eDNA over distance was calculated using data between St. 1 and W10 or W30, and St. 1 and E10 or E30 at all time points after setting the cage, as follows:

$$ {\text{Dst}}\; = \;100\; \times \;{{\text{A}} \mathord{\left/ {\vphantom {{\text{A}} {\text{B}}}} \right. \kern-0pt} {\text{B}}}, $$

where A is the average eDNA concentration at each station, B is the average eDNA concentration at St. 1, and Dst is detectability in distance, presented as a percentage.

We also compared eDNA concentrations at three different depths (0, 4, and 8 m) at St. 1 using ANOVA followed by Tukey’s HSD test. R statistical software version 3.2.0 (2015.04.16; R Core Team 2015) was used for these analyses.

Results

Striped jack DNA was not detected in samples collected at St. 1 before setting the cage, implying that neither wild nor anthropogenically originated striped jack DNA was present. No striped jack DNA was detected in any negative controls during filtration or PCR, indicating that there was no contamination during the experiment.

The concentration of eDNA gradually decreased with distance from the cage (Fig. 3). Amplified samples were confirmed in 39 of the 273 (13 stations × 7 time points × 3 replicates) samples collected at 0–48 h after setting the cage (Fig. 3). Among these amplified samples, 28 (72%) were located within 30 m of the cage. Meanwhile, a substantial proportion of samples (9 out of 21, after setting the cage) at St. 1 showed no detection of DNA (0 copies). eDNA was detectable for up to 1 h after removing the cage. The detectability of eDNA concentration at each sampling station was 0–36.5%/h after removing the cage (Table 1). These results indicate that eDNA concentration at a station decreased over 63.5% in 1 h. At 0–1 h after removing the cage, target eDNA was detected in 18 out of 78 samples (13 × 2 × 3). Among these, 17 (94%) were located within 30 m of the cage. Taken together, amplified samples were confirmed in 57 out of the total 351 samples at the total of 0–48 h after setting the cage and 1–2 h after removing. Among the amplified samples, 45 (79%) were obtained within 30 m of the cage. However, after 2–48 h, eDNA was not detected in any samples. The decrease in eDNA concentration from its source was fitted to the power regression model (Figs. 3, 4), and the coefficient of determination for each time point and direction ranged from 0.00 to 0.93.

Fig. 3
figure 3

Concentration of eDNA at each sampling station at a, b: 0 h; c, d: 1 h; e, f: 2 h; g, h: 4 h; i, j: 8 h; k, l: 24 h and m, n: 48 h after setting the cage of fish. Left and right column graphs represent the west and east transect lines, respectively. Open circles represent the average data from PCR triplicates of each water sample (n = 3 in each station and time point)

Full size image
Table 1 The detectability (%/h) of eDNA between 0 and 1 h after removing the cage
Full size table
Fig. 4
figure 4

Concentration of eDNA at each sampling station at a, b: 0 h; c, d: 1 h and e, f: 2 h or later after removing the cage. a, c, e: west; b, d, f: east. Open circles represent the average data from PCR triplicates of each water sample

Full size image

Mean ± SD water temperature was 21.4 ± 0.4 °C and salinity was 31.7 ± 0.8. The flow velocity during the field experiment was 5 ± 4 (mean ± SD, range: 0–25) cm/s, with directions mainly east or southeast except for time points of 0–4 h after setting the cage (Fig. S1a, b). The direction of current for 1 h before each sampling time point was compared with the total eDNA concentration along the transects (Fig. 3); there were only two out of nine time points where more eDNA was detected along the transect in the direction of currents (24 h after setting the cage and 1 h after removing the cage), whereas an opposite trend was observed in six out of the others. The tidal range was 19 cm during the survey period according to data from the Maizuru Tidal Station of the Japan Meteorological Agency (Fig. S1c).

eDNA concentration decreased exponentially with distance from the source, and was significantly different among sampling stations to the east and west (p < 0.05, nested ANOVA; Fig. 5). In the west, the concentration of eDNA was highest at St. 1, followed by W10 and W30. eDNA concentration at W100 was significantly lower than that at St. 1 and W10. eDNA concentration was not significantly different between W100 and W1000 (Fig. 5a). In the east, the concentration of eDNA was significantly higher at St. 1 and E10 than at E100 or E1000. eDNA concentration was not significantly different between E30 and E1000 (Fig. 5b).

Fig. 5
figure 5

Spatial change of eDNA concentration for all the time points along the northwest (a) and northeast (b) transects after setting the cage and northwest (c) and northeast (d) transects after removing the cage. Open circles represent eDNA concentration at each time point 0–48 h after setting the cage, and those at 0 h and 1 h after removing the cage. Different lowercase letters represent significant differences in ANOVA (p < 0.05) followed by Tukey’s HSD test

Full size image

The change in detectability over distance after removing the cage was as follows: St. 1–W10, 3.2%; St. 1–W30, 0.7%; St. 1–E10, 21.2%; St. 1–E30, 2.4%. Because eDNA was not detected at 2 h after the removal of the cage (Fig. 6), we assumed that eDNA transferred from adjacent stations was negligible for the above estimation.

Fig. 6
figure 6

Temporal change of eDNA concentration accumulated for all the sampling stations after setting (a, b) or removing the cage (c, d). Different shadings represent different sampling stations as shown in the inserted legends. Different lowercase letters represent significant differences in nested ANOVA (p < 0.05) followed by Tukey’s HSD test

Full size image

eDNA concentration was significantly different among time points after setting and removing the cage (p < 0.05, nested ANOVA). The concentration of eDNA was significantly higher at 0 h after setting the cage than at other time points (Tukey’s HSD test, p < 0.05; Fig. 6a, b). eDNA concentration at 0 h after removing the cage was also significantly higher than at 1 and 2 h, in both the east and west transects (p < 0.05, Tukey’s HSD test; Fig. 6c, d).

In the comparison of depth layers, eDNA was detected in 8 of the 27 samples collected at 29 h, 32 h and 37 h after setting the cage, among which 5 samples were from surface water (Fig. 7). A significant difference in eDNA concentration was identified between depth layers at 37 h, with the highest value observed in surface water (Fig. 7c).

Fig. 7
figure 7

eDNA concentration collected from three different layers of the water column in St. 1 at a 29 h, b 32 h and c 37 h after setting the cage. Open circles represent average data from PCR triplicates of each water sample (n = 3). Different lowercase letters represent significant differences in ANOVA (p < 0.05) followed by Tukey’s HSD test

Full size image

Discussion

This field experiment evaluated the spatiotemporal detection range of eDNA using qPCR. We found that eDNA was detectable at stations mostly within 30 m from the source (among 57 amplified samples, 45, or 79%, were obtained within 30 m from the cage; Figs. 3, 4). Our study also indicates that the target DNA was detectable during the time the cage was present and up to 1 h after removing the cage, but not after 2 h had elapsed (Fig. 6). Previous studies conducted in rivers reported longer dispersion distances. For example, Jane et al. (2015) found that eDNA of brook trout Salvelinus fontinalis in a cage could be detected at 240 m downstream from its source. Deiner and Altermatt (2014) reported that eDNA of daphnia Daphnia longispina could be detected at 12.3 km and DNA from the pelecypod Unio tumidus could be found at 9.1 km downstream. The limited dispersion of eDNA in our study may be partly because the marine costal study area has a more limited current speed with complicated flow direction, and involves a larger open area than relatively small rivers.

The dispersion area of eDNA in the sea demonstrated in the present study was within the range suggested by previous studies (Fig. 5). A previous study comparing eDNA concentration and echo intensity in sonar in Maizuru Bay suggested that eDNA reflects the source within a distance of 10–150 m (Yamamoto et al. 2016), although in their study, sampling stations were set at 400-m intervals and the estimation of dispersion and degradation area was based on a statistical model. Meanwhile, nearly 100% of eDNA shed by shiner perch Cymatogaster aggregate, a nearshore fish, could be detected within 100 m from the shore (O’Donnell et al. 2017).

We did not find any explainable relationship between the pattern of current direction and eDNA dispersion (Fig. 3, Fig. S1). The tidal amplitude is generally small in the Sea of Japan, and tidal currents are very weak. Therefore, the current we measured may not have been a major factor in the dispersion of eDNA. The distance between the location of current flow meters and the cage (~ 300 m from NW/NE and St. 1) may also have hindered monitoring of the currents affecting dispersion. Future studies should be conducted in an area where the current is strong to focus on the relationship between current and eDNA dispersion.

The decrease in eDNA in the sea over time could be due to not only dispersion but also degradation. The decrease in eDNA in the sea demonstrated in the present study (Table 1) was much faster than the degradation reported in previous tank experiments. Degradation rates of eDNA for freshwater fishes in tanks are reported as 10.5%/h in common carp Cyprinus carpio (Barnes et al. 2014) and 15.9%/h in bluegill sunfish Lepomis macrochirus (Maruyama et al. 2014). The degradation rates of eDNA of marine fishes in tanks are 4.6%/h in European flounder Platichthys flesus, 1.5%/h in three-spined stickleback Gasterosteus aculeatus (Thomsen et al. 2012a), and 5.5–10.1%/h in northern anchovy Engraulis mordax, Pacific sardine Sardinops sagax, and Pacific chub mackerel Scomber japonicus (Sassoubre et al. 2016). In the present study, eDNA was not detected at any stations more than 1 h after removing the cage. The detectability of eDNA between 0 and 1 h after removing the cage was 0–36.5%/h. Although previous studies of eDNA degradation in a tank used a model of power decline, the degradation rate of eDNA in the sea occurred more quickly, and did not fit this model. eDNA retention time in the sea may be shorter owing to the combination of biological decay and physical dispersion. Additionally, the degradation of eDNA could have been accelerated by high water temperature (21.4 °C) because warm temperature hastens degradation of eDNA due to microbial activity (Strickler et al. 2015; Lance et al. 2017; Tsuji et al. 2017).

It is also important to consider that the emission of DNA might not occur constantly but intermittently. Our dataset showed a number of cases where eDNA was not detected at the nearest station to the cage (Fig. 3). This may be partly due to dispersion caused by flow and the limited probability of catching the DNA fragments during water sampling. However, fish were constantly swimming and forming a group during the experiment. The detected eDNA pattern may also indicate that eDNA was emitted intermittently, with a fluctuating rate of release. A large amount of eDNA was detected immediately after setting and removing the cage, most likely because of handling (Fig. 6). Less eDNA was detected at 1–8 h after setting the cage. A large quantity of eDNA was also detected at 24 h and 48 h, suggesting that a large emission of eDNA could occur temporarily. Fish activity generally increases when prey such as crustaceans are present (Masuda et al. 2012). Feeding behavior contributes to an increased release rate of eDNA in carp, suggesting that feces and intestinal tissues are major sources of eDNA (Klymus et al. 2015). Feeding, defecation, and response to a potential predator were not temporally consistent and thus might have affected the fluctuation of DNA emission.

There were some cases where eDNA was detected at a relatively long distance from the source, with the most distant observation occurring at W1000 immediately after setting the cage (Fig. 1). The summertime current in Maizuru Bay is generally weak, with an average velocity of 5–20 cm/s at the surface (Miwa and Ikeno 2007). The measured hourly velocity immediately after setting the cage was 0.4–13.5 cm/s. Such a slow current can transport eDNA only about 14–480 m in 1 h, making it difficult to explain the rapid transport observed. None of the negative controls in either filtration or PCR (45 in total) showed any amplification, suggesting that contamination after the sampling process is unlikely. Considering that the fish seem to have emitted a substantial amount of eDNA when being introduced in the cage, the potential risk of contamination may have been high at this time point. Alternatively, striped jack is distributed as food for human consumption in this region, and so we cannot exclude the possibility that eDNA in this sample may have come from drainage discarded from houses or vessels. Jo et al. (2017) recently developed a method using two sets of primers with long and short DNA fragments to distinguish between eDNA emitted from live fish and eDNA that is relatively old and degraded. We suggest that long fragments of eDNA would likely be present only near the cage, as opposed to the short fragments we targeted in this study; this would require further investigation to resolve.

In examining eDNA at different depths, we found that eDNA was most frequently present at the surface (Fig. 7). This implies that eDNA tended to stay in the layer where it was released, or tended to be buoyant in surface water, and justifies sampling from sea surface water in our experiment. Moreover, this result coincides with a previous finding that the vertical distribution of eDNA reflects the layer in which organisms live, based on a study by Minamoto et al. (2017) wherein eDNA of the jellyfish Chrysaora pacifica was more abundant near the bottom where this species tends to accumulate. On the other hand, we did detect eDNA in the middle and bottom layers, implying that some eDNA may have sunk after emission. eDNA at the surface might be detectable relatively soon after emission, while eDNA may be detected near the bottom at distant points after transport and gradual settlement. In a lake, eDNA of bigheaded Asian carp Hypophthalmichthys spp. was detected in greater quantities in sediment than in surface water (Turner et al. 2015), suggesting that settled eDNA could be preserved in sediment. On the other hand, eDNA of African jewelfish Hemichromis letourneuxi in an experimental pond was present in greater quantities near the surface than at the bottom (Moyer et al. 2014). These findings indicate that vertical distribution of eDNA depends on both the depth of source organisms and the behavior of eDNA, with the latter dependent on its chemical condition and physical factors in the environment.

Although the use of eDNA as a tool for marine studies still has many obstacles, the present investigation implies that the area of eDNA dispersion in the sea is as narrow as 30 m and the duration of persistence could be as short as 1 h when emitted from a limited source. Collectively, these results suggest that eDNA reflects presence and abundance of organisms not in a large scale of time and space, but within a few hours and tens of meters. Further study should elucidate how biomass and other factors influence the distance of dispersion, degradation, and concentration of eDNA.