Sedimentary eDNA provides different information on timescale and fish species composition compared with aqueous eDNA

eDNA provides different information on timescale and fish species composition compared with aqueous eDNA Abstract Aqueous environmental DNA (eDNA) analysis has been applied to the monitoring of various ecosystems and taxa, and the characteristics of aqueous eDNA have been previously studied. In contrast, although sedimentary eDNA has been used to restore past information, the characteristics of sedimentary eDNA are not well understood. In this study, we compared the properties of sedimentary and aqueous eDNA of macro-organisms. First, to clarify the preservation ability of sediments, we compared the difference in decay rates

Most eDNA studies on macro-organisms analyzed eDNA in water samples (Goldberg, Strickler, & Fremier, 2018;Ishige et al., 2017;Minamoto et al., 2012;Miya et al., 2015;Ushio et al., 2018;Valentini et al., 2016); however, several recent studies have targeted eDNA in underwater sediments (Buxton, Groombridge, & Griffiths, 2018;Shaw et al., 2016;Turner, Uy, & Everhart, 2015;Wei, Nakajima, & Tobino, 2018). To date, comparison of the basic properties between sample types (i.e., water/sediment) has been performed only for fish eDNA. According to Turner et al. (2015), fish eDNA derived from sediment samples contains a higher copy number per unit weight than that derived from water samples in both experimental ponds and a natural river. For micro-organisms, underwater sediments accumulate information of the surrounding ecosystem. It is considered that some of aqueous eDNA-bearing particles, such as feces, are too large to remain suspended in water (Maggi, 2013) and consequently precipitate onto the sediments . Such precipitation of aqueous eDNA-bearing particles forms a part of the mechanism of sedimentary eDNA accumulation. It has been reported that animal feces contain viable epithelial cells (10 −1 -10 6 cells/g) and large amounts of mtDNA (10 −1 -10 7 copies/g) (reviewed in Caldwell, Payment, & Villemur, 2011) and that feces from aquatic macrofauna rapidly sinks (Robison & Bailey, 1981;Wotton & Malmqvist, 2001).
Additionally, several studies have reported that sediments reduce biologically driven DNA decay by adsorbing both DNases and DNA molecules and in low-oxygen environment such as deeper sediment (Levy-Booth et al., 2007;Pietramellara et al., 2009). This delay in DNA decay caused by the binding of DNA and DNases to sediment particles would also be observed in shallow sediments (Shogren et al., 2017). Furthermore, in aquatic sediments, it was reported that the chemical DNA decay rate appeared to be low (Corinaldesi, Barucca, Luna, &Dell'anno, 2011), andOgram, Mathot, Harsh, Boyle, andPettigrew (1994) reported that the decay of DNA absorbed to soil particles was minimal. In fact, Turner et al. (2015) discovered that the detectable period of sedimentary eDNA was longer than that of aqueous eDNA. Moreover, it has been reported that virus DNA infecting common carp showed higher concentration in sediment than in the water column , and that fish eDNA did not rise to the surface in the water column (Kamoroff & Goldberg, 2018). Another study reported that the concentration of bigheaded carp eDNA was higher per g of sediment than per ml of water (8-to 1846-fold) (Turner et al., 2015). Therefore, it is expected that sedimentary eDNA concentration will be higher than aqueous eDNA concentration, and it can be hypothesized that the difference in concentration is caused by differences in decay rates. However, a direct comparison of decay rates between sedimentary and aqueous eDNA in controlled experimental conditions and the subsequent comparison of information contained in sedimentary and aqueous eDNA in a natural lentic environment have not been reported.
Although it has been reported that eDNA in biofilms (part of the surface sedimentation) decays to nondetectable levels within 2 days (Seymour et al., 2018), knowledge on the persistence of sedimentary eDNA is limited. If the decay rate varies between sedimentary eDNA and aqueous eDNA, it may follow that information on the biota obtained from both types of samples would be qualitatively different.
In a previous study that compared the detected fish species between sedimentary and aqueous eDNA, eDNA metabarcoding for fish was performed using 1 L water samples and 0.25 g sedimentary samples from rivers (Shaw et al., 2016). The species detected from sedimentary eDNA were fully encompassed within those from aqueous eDNA. In contrast, a different study reported that the fish species detected in 10 g of sediment only partially overlapped with those from 0.9 L water samples (Siegenthaler et al., 2018). Therefore, in simple comparisons between sediment and water, such as the present study, it may be necessary to consider the weight of the sediment sample in order to test the qualitative difference between sediment and 1 L of water. In addition, it is reported that eDNA retention is influenced by sediment substrate (Shogren et al., 2017) and bacterial abundance (Tsuji, Ushio, Sakurai, Minamoto, & Yamanaka, 2017). Shogren et al. (2017) showed that the dynamics of aqueous eDNA, such as transport, retention, and resuspension, were influenced by sediment substrates (in particular, particle size).
Therefore, the dynamics of sedimentary eDNA may be influenced by sediment substrates because it is assumed that DNA adsorption is affected by changes in surface area with particle size.
In recent years, studies that use ancient DNA from sediment cores have increased (Parducci et al., 2017). Much of that research has been targeted to species for which the body remains in the sample, such as bacteria (Domaizon et al., 2013), other microbes (Hou et al., 2014), and plants (Pansu et al., 2015), or species for which part of the body, such as bones, are preserved (Wooller, Gaglioti, Fulton, Lopez, & Shapiro, 2015). More recently, studies have been developed to detect macro-organisms using extracellular DNA remaining in sediments (i.e., sedimentary eDNA), for example, for fish (Nelson-Chorney et al., 2019;Stager, Sporn, Johnson, & Regalado, 2015). These studies showed that fish sedimentary eDNA can be detected in sediments from ~140 years ago. As such, revealing past information by analyzing the extracellular DNA of macro-organisms in sediments has been realized. The abovementioned studies on fish are important not only for ecology but also for the fishery industry because they could enable predictions of future trends by observing the past. However, although it is known that DNA molecules remain for a long time in low-oxygen environments, such as deeper sediments as mentioned above, there is little information on the decay rate of fish sedimentary eDNA on the surface before deposition in the anoxic deep layer. As basic information of sedimentary eDNA is lacking, we measured the decay rate of sedimentary eDNA and compared with that of aqueous eDNA.
Our findings regarding decay rates will provide a part of mechanism on holding long time of sedimentary eDNA.
In this study, to elucidate the characteristics of sedimentary eDNA compared with aqueous eDNA, we first compared the decay rates of sedimentary and aqueous eDNA under the condition of a random sediment substrate type. Next, we collected paired sediment and water samples from a natural lake and compared fish eDNA concentration between sample types. Finally, the fish species detected by eDNA metabarcoding were compared. Our study partially clarified the characteristics of sedimentary eDNA. Specifically, we clarified quantitatively that detection was possible over a longer period of time when the decay rate of sedimentary eDNA was slower and found that sedimentary eDNA works in a complimentary manner with aqueous eDNA in biomonitoring.

| eDNA sampling
Sediments and water were sampled from a biotope (an artificial pond made of concrete with a volume of approximately 4,000 L) in the Tsurukabuto Second Campus of Kobe University, Japan (34.734°N, 135.234°E), to compare the decay rates of eDNA in the water and sediment samples. Two fish species (Hemigrammocypris rasborella and Oryzias latipes) inhabited the biotope, and half of the water area in the biotope was covered with emergent plant species. The sediment in biotope mainly consists of organic matter and mud. Nine bulk sedimentary samples of approximately 45 g each were collected from surface sediments in 50-ml tubes (includes random particle size), and after thoroughly stirring, 3 g was transferred into each of nine 15-ml tubes per one bulk sample (81 sediment samples in total; Figure 1). Nine water samples were collected using 5-L plastic tanks (sample series ID: 1-9; Figure 1) and mixed well, and then, 250 ml of the sample was transferred into each of nine 250-ml bottles per one bulk sample (81 water samples in total; Figure 1). In addition, reverse osmosis membrane water (5 L) was divided into nine subsamples to serve as negative controls (NCs). To test for decay and avoid any large temperature fluctuations, all samples and controls were retained for different time periods (0.5 day (12 hr), 1 day (24 hr), 2 days (48 hr), 3 days (72 hr), 7 days (168 hr), 14 days (336 hr), 21 days (504 hr), and 28 days (672 hr)) in boxes prior to filtration for a maximum of 28 days (672 hr). Furthermore, to monitor only the water temperature fluctuation, a 250-ml bottle with 250 ml of water containing a temperature logger (HOBO pendant logger, HOBO) was placed in each box. Temperature range was 15-19.6°C except for the first 48 hr before stabilizing the water temperature ( Figure S1).
One of the nine water samples per single bulk sample (i.e., 250 ml of water) was filtered with a glass fiber filter with nominal pore size of 0.7 μm (GF/F; GE Healthcare Life Science) immediately after the sampling, and this was defined as the Time 0 sample. Subsequently, filtration was performed at 12, 24, 48, 72, 168, 336, 504, and 672 hr. Filtration of the control series was performed at the same times to evaluate any possible DNA contamination during filtration. Sediment samples were stored frozen at −25°C at the same timing as filtration. To prevent cross-contamination among samples, all tools used were decontaminated with chlorine bleach (0.1% effective chlorine concentration).

| eDNA extraction
eDNA on the filters (i.e., eDNA from water samples) was extracted using the Salivette (Sarstedt) and DNeasy Blood & Tissue Kit (Qiagen,) methods and stored at −25°C according to the methods described by Minamoto, Hayami, Sakata, and Imamura (2019).
DNA was extracted from sediment samples by combining alkaline DNA extraction (Kouduka et al., 2012) with ethanol precipitation and a fecal-soil DNA extraction kit (PowerSoil DNA Isolation Kit, MO Bio Laboratories). Briefly, 6 ml of 0.33 M sodium hydroxide solution and 3 ml of Tris-EDTA buffer (pH 6.7) were added to the sediment sample, and it was thoroughly mixed by vortex and then incubated at 94°C for 50 min. The sample was cooled down to room temperature and centrifuged at 5,000 × g for 30 s, and then, 7.5 ml of supernatant was transferred to a new 50-ml tube and neutralized with the same volume of Tris-HCl (1 M, pH 6.7). Next, 1.5 ml of 3 M sodium acetate solution (pH 5.2) and 30 ml of absolute ethanol were added to the mixture and placed in a freezer (−25°C) for more than 1 hr. Cooled samples were centrifuged at 5,350 g for 20 min, and the supernatant was discarded. The pellet was transferred to a Power Bead Tube (PowerSoil DNA Isolation Kit). To retrieve any residual DNA, the remaining precipitate in the 50-ml tube was dissolved with 100 μl of ultrapure water and transferred to the same Power Bead Tube. Subsequently, DNA extraction was performed according to the "Experienced User Protocol 3 to 22" of the PowerSoil DNA Isolation Kit. eDNA extraction was performed in a separate room from PCR operations to prevent contamination.

| Real-time quantitative PCR (qPCR)
The amount of eDNA of H. rasborella was quantified by TaqMan real-time quantitative PCR (qPCR) targeting the cytochrome b region of H. rasborella using previously developed primers and a probe (Fukuoka, Takahara, Matsumoto, Ushimaru, & Minamoto, 2016; Table 1), and the specificity of the primers/probe set was confirmed by Fukuoka et al. (2016) through specificity tests using DNA of closely related species. Real-time qPCRs were carried out in triplicate using extracted eDNA from each sample as template.
Each reaction (20 μl final volume) contained 900 nM primers and 125 nM TaqMan probe in 1× TaqMan Gene Expression Master Mix (Life Technologies) and 2 μl eDNA. The real-time PCR conditions were as follows: 2 min at 50°C, 10 min at 95°C, and 55 cycles of 15 s at 95°C and 60 s at 60°C. To obtain calibration curves, a dilution series of standards (3 × 10 1 -3 × 10 4 copies in each reaction) were simultaneously quantified: The standard was linearized plasmids that contained synthesized artificial DNA fragments of the target cytochrome b gene sequence of H. rasborella. Ultrapure water was used instead of DNA in three reaction mixtures as nontemplate negative controls.

| eDNA sampling and extraction
On 15 July 2015, we sampled water and sediments at four points at the Lake Iba shore (Figure 2), one of the lakes adjacent to Lake Biwa, Shiga Prefecture, Japan (surface: 0.5 km 2 , average depth: 1.5 m; Figure 2; Table S1). Approximately 45 g of sediment, scooped F I G U R E 1 Workflow of sampling and subsampling from the surface of the lake bottom, and 1 L of surface water were collected at these points (lake shore). Sediment and water samples were transported to the laboratory on ice, and water samples were immediately filtered at the laboratory. We used two glass fiber filters with nominal pore size of 0.7 μm (GF/F) to filter a 1-L water sample because filters sometimes clogged, and it was impossible to process 1 L through a single filter. Two filters were pooled into a single tube and preserved at −25°C. Sediment samples were stored frozen at −25°C until DNA extraction. eDNA was extracted by the methods described in "eDNA extraction" section and stored at −25°C. All tools used were decontaminated with chlorine bleach (> 0.1% effective chlorine concentration).

| Comparison of eDNA concentration between sediment and water samples by qPCR
To compare eDNA concentration between sediment and water samples, qPCRs were performed targeting the cytochrome b region of the common carp (Cyprinus carpio), bluegill sunfish (Lepomis macrochirus), and largemouth bass (Micropterus salmoides), using previously developed primers and probes (Table 1). All qPCRs were performed under the conditions described above ("Real-time quantitative PCR" section) except for primers and probes. Further, to test PCR inhibition, each qPCR solution was spiked with 2,000 copies of lambda phage DNA as an internal positive control (IPC). This test was carried out

| Detection of fish species by eDNA metabarcoding
To compare the fish-faunal information retained in eDNA between the sediment and water samples, eDNA metabarcoding was per-

| Bioinformatics
We performed data preprocessing and analysis of MiSeq raw reads using USEARCH v10.0.240 (Edgar, 2010) according to the following steps. (a) Paired-end reads (reads 1 and 2) were merged using the command "fastq_mergepairs" with a default setting. During this process, low-quality tail reads with a cutoff threshold set at a quality (Phred) score of 2, too short reads (<100 bp) after tail trimming, and those paired reads with too many differences (>5 positions) in the aligned region (ca. 65 bp) were discarded; (b) Primer sequences were removed from those merged reads using the command "fastx_truncate"; (c).
Quality filtering using the "fastq_filter" command was performed to remove low-quality reads with an expected error rate (Edgar & Flyvbjerg, 2015) of >1% and too short reads of <100 bp; (d) The preprocessed reads were dereplicated using the "fastx_uniques" command, and all singletons, doubletons, and tripletons were removed from the subsequent analysis following the recommendation of the author of the program. (e) The dereplicated reads were denoized using the "unoise3" command to generate amplicon sequence variants (ASVs). We removed ASVs with all putatively chimeric, erroneous sequences (Edgar, 2016), and those with less than 10 reads; 6) Finally ASVs were subjected to taxonomic assignments to species names using the "usearch_global" command with a sequence identity of >98.5% (two nucleotide differences allowed) with the reference sequences and a query coverage of ≥90%. Finally, species reads that were detected in both sample and positive control were regarded as possible contamination if the number of reads in the sample was less than that in the positive control, and these species reads were discarded.

| Statistical analysis
In the qPCR results, the DNA concentration was calculated as the average of the three replicates. When a negative detection was obtained in any of the replicates, the DNA concentration of that replicate was assigned as zero (Ellison, English, Burns, & Keer, 2006).
To compare eDNA concentration between sediments and water, eDNA concentrations were converted to copy numbers per unit weight and then log-transformed. These concentration comparisons implicitly assume the equivalence of 1 ml and 1 g of water. To analyze the decay rates of both sample types, we fit a linear mixed model using the function LMER in R package LME4 (Bates, Mächler, Bolker, & Walker, 2015). In this model, eDNA copy numbers served as a response variable. The time point, the sample type (water/sediment), and their interaction were set as explanatory variables, and the sample series ID was set as a random effect to consider differences in decay rates between sample series. Additionally, the decay rate was calculated based on the untransformed data assuming exponential decay. To compare sedimentary and aqueous eDNA concentrations for field samples, linear mixed modeling was performed.
In this model, eDNA copy numbers served as a response variable.
The sample type (water/sediment) was set as an explanatory variable, and the fish species was set as a random effect. To compare the number of detected species between sample types, a paired t test was performed. In addition, to compare fish species composition between sampling types, nonmetric multidimensional scaling (NMDS) was performed with "Jaccard methods" and 10,000 permutations, and the PERMANOVA analysis was performed with "Jaccard methods" and 10,000 permutations using the "adonis" function. In this analysis, abundance information was not included. All analyses were performed using the software R ver. 3.5.1 (R Core Team, 2018).

| Comparison of decay rates of eDNA for biotope samples
The R 2 values of calibration curves were >0.976 in all runs. The slopes ranged between −3.428 and −4.077, the intercepts ranged between 42.648 and 45.455, and the PCR efficiency ranged between 75.908% and 95.741%.
The slopes of the regression lines based on all eDNA concentrations at each time point differed significantly between sample types (p < .05; Figure 3;

| Comparison of eDNA concentration between sediment and water samples
The 0.3-3.8 copies/ml for bluegill sunfish; Figure 4). For three of the four sites, the eDNA concentration of largemouth bass was higher per g of sediment than per ml of water (sediment: 84.9-570.1 copies/g; water 0.5-12.7 copies/ml; Figure 4); however, sedimentary eDNA was not detected at one of the four sites. The results of the LMM analysis revealed that sedimentary eDNA concentration was significantly higher than aqueous eDNA one (p < .05; Table 3, Figure 4).  clusters were assigned to fish taxa (>98.5% identity) and the other clusters were assigned to mammals, reptiles, and bacteria. After possible contaminant sequences were removed, the number of clusters subjected to the following analyses was reduced to 90.

| Detection of fish species by eDNA metabarcoding
Only T. japonicus sequences were detected in the positive controls.
A total of 22 fish species were identified by eDNA metabarcoding (Table 4). The detected number of species was saturated for the number of reads in all samples ( Figure S4). Fourteen fish species were detected from sediment samples, and 20 fish species were detected from water samples. In the paired t test, the number of detected species was higher in that water sample than in the sediment sample ( Figure 5; p < .01). On the contrary, the fish species composition detected by eDNA metabarcoding was not statistically different between the sediment and water samples (PERMANOVA: p = .59; Figure S5). However, although there was no statistical significance, some fish species were detected only in sediment or water samples ( Figure 5; Table 4; Figure S5).

| D ISCUSS I ON
The present study revealed the decay rate of sedimentary eDNA in surface sediments compared with aqueous eDNA. Long-term retention in low-oxygen environments, such as deeper sediment, is well known, but our finding that the decay rate in surface sediment is also very slow reveals part of the mechanism underlying the long holding time of sedimentary eDNA. The sedimentary eDNA concentration was higher than the aqueous eDNA concentration both in an artificial biotope and a natural lake, as reported in a previous study (Turner et al., 2015). Further, the fish species identified from eDNA metabarcoding were different between sediment and water samples. These results Note: In this model, eDNA copy numbers served as a response variable. The time point, the sample type (sediment and water are included as 0 and 1, respectively), and their interaction were set as explanatory variables. Asterisks show the significant effects of each parameter (***p < .001). Note: In this model, eDNA copy number served as the response variable and the sample type (sediment and water are included as 0 and 1, respectively) was set as the explanatory variable.

TA B L E 2
Asterisks show the significant effects of each parameter (*p < .05).

TA B L E 3
Summary results of linear mixed models for field samples indicate that sedimentary eDNA has a slower decay rate than aqueous eDNA (approximately 1/57; Figure 3) and suggest that sedimentary eDNA may complement aqueous eDNA for revealing previously undetectable species ( Figure 5). For example, fish that migrate seasonally may not be detected in water samples when they are not present, but they may be detected in sediment samples.
The decay rate of eDNA in sediment samples was much lower than that in water (sedimentary eDNA = 0.00033/hr, and aqueous eDNA = 0.019/hr; Figure 3). In the present study, we clarified the decay rate of sedimentary eDNA, which was rarely reported in previous studies. This decay rate (0.019/hr) of aqueous eDNA is comparable with that of natural environments reported in previous studies (0.0097-0.101: Sassoubre, Yamahara, Gardner, Block, & Boehm, 2016;Sansom & Sassoubre, 2017). Therefore, although our experiments were conducted in a controlled environment, the decay rate of sedimentary eDNA can be assumed to be within the range of rates found in the natural environment. However, previous studies reported that aqueous eDNA was degraded rapidly by the effects of water state, temperature, sunlight (UV), and pH (Andruszkiewicz, Sassoubre, & Boehm, 2017;Eichmiller, Best, & Sorensen, 2016;Strickler, Fremier, & Goldberg, 2015;. Sedimentary eDNA might be protected from such decay by adsorption to soil particles, and this may explain its slow-decay rate compared with aquatic eDNA. For example, a previous study indicated that the DNA adsorbed to soil particles was protected from decay by nucleases (Ogram et al., 1994). Our finding that the decay of sedimentary eDNA was slower than that of aqueous eDNA supports the findings of previous studies on the dynamics of DNA molecules in sediments (Corinaldesi et al., 2011;Levy-Booth et al., 2007;Pietramellara et al., 2009). A previous study found that carp eDNA remained detectable up to 132 days in sediment after removing carp fish (Turner et al., 2015), whereas the detectable time of aqueous eDNA was reported to be several days to several weeks Dejean et al., 2011;Goldberg et al., 2013;Pilliod, Goldberg, Arkle, & Waits, 2014;Thomsen, Kielgast, Iversen, Møller, et al., 2012).
Therefore, the sedimentary eDNA would persist longer than aqueous eDNA, and deriving historical information on macro-organisms by collecting eDNA from sediment core samples is possible as reported previously (Bálint et al., 2018;Parducci, Suyama, Lascoux, & Bennett, 2005;Stager et al., 2015). Moreover, combining the decay rate of sedimentary eDNA as shown in our study and the initial concentration of sedimentary eDNA in sediment cores will be useful for revealing past  The qPCR results of field samples showed that sedimentary eDNA concentration was higher than that of aqueous eDNA for the same sample weight (Figure 4). This result matches that of a previous study (Turner et al., 2015), and the concentration of eDNA in sediment would be generally higher than that in water.
The number of fish species detected by eDNA metabarcoding was significantly higher in water sample than in sediment sample (paired t test: p < .01; Figure 5). This may be because of aqueous eDNA, indicating a wider spatial scale, but in this comparison, the volume of sample was not considered (sediment: 3 g, water: 1 L).
Therefore, water sample may be advantageous because a large amount of sample can be used. On the contrary, the fish species composition detected by eDNA metabarcoding was not statistically different between sediment and water samples ( Figure S5).
However, some species were detected only in sediment or water samples ( Figure 5; aqueous eDNA reflects a wider spatial scale. This result was consistent with a previous study that used 10 g sediment sample (Siegenthaler et al., 2018). In another previous study, which used low sample weight (0.25 g), the species detected from aqueous eDNA subsumed those from sedimentary eDNA (Shaw et al., 2016). This inconsistency could be caused by differences in the eDNA timescale reflected, or differences in spatial scale, in addition to differences in decay rate between sediment and water samples. For example, aqueous eDNA would be expected to reflect a wide spatial scale because it diffuses in the water (Dunker et al., 2016), whereas it would be more difficult for sedimentary eDNA to diffuse. Overall, our result suggested that information on fish species composition obtained from 3 g of sediment samples and 1 L of water samples were comparable, and this result is similar to that of a previous study (10 g and 2 L; Siegenthaler et al., 2018). Sediment samples of 3-10 g are preferred over less samples (0.25 g; Shaw et al., 2016), because greater sample sizes may further improve the results of sedimentary eDNA analysis. Meanwhile, in metabarcoding assays, the estimation of abundance was prevented by PCR bias (Bista et al., 2018).
Therefore, the estimation of abundance by using eDNA metabarcoding needs careful interpretation. Wei et al. (2018) observed two phases in the sedimentary eDNA decay process: a fast-decay phase (until 72 hr after removal of individuals), followed by a slow-decay phase (until 480 hr). However, we F I G U R E 5 Number of fish species detected by eDNA metabarcoding. Blue and red show results from water and sediment samples, respectively Site1 S ite2

Site3 S ite4
Water Sediment assumed that we observed only a single decay phase in the present study ( Figures S2 and S3). It appears that the fast-decay phase did not occur in our study. The slow-decay speed may have been due to the low temperature early in the experiment (until approximately 72 hr; Fig. S1) (Eichmiller et al., 2016;Strickler et al., 2015).

| CON CLUS ION
In this study, we demonstrated sedimentary eDNA characteristics in comparison with aqueous eDNA under a controlled and natural lentic condition. The decay rate of sedimentary eDNA was low even for surface sediment, and the sedimentary eDNA concentration was higher than that of aqueous eDNA for the same sample weight.
Additionally, the composition obtained by metabarcoding was not statistically different between sediment or water samples. However,

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

E TH I C A L S TATEM ENT
No animal experiments were performed in this study. All experiments were performed according to the current law of Japan.

DATA AVA I L A B I L I T Y S TAT E M E N T
The raw data were deposited to Dryad (https ://doi.org/10.5061/ dryad.mgqnk 98wd).