Positive effect of filtration additives for increasing environmental DNA detections in summer and winter oceanic waters

Abstract

Environmental DNA (eDNA) is a convenient biodiversity monitoring tool, and many studies have focused on river and coastal areas, while deep‐water oceanic surveys are rare. However, using eDNA techniques to learn about oceanic communities may also be a useful approach. We tested methods to enhance ocean eDNA collections adding four different materials (sediment, diatomite, zirconia bead, and molecular sieve) in combination with 0.45‐μm pore size filter, also compared 0.22‐μm pore size filter, using five 2‐L volume replicates. The experiments were conducted off the Pacific coast of Hokkaido, Japan, in summer and winter. The winter trial also included a larger water volume (10 L, duplicate) experimental treatment. The added diatomite showed the highest number of detected operational taxonomic units (OTU) in the summer trial, while sediment and beads showed higher numbers in winter. GLM analysis showed that pore size does not affect the number of OTUs, while season and the other four additive materials were significant effects. The added materials in the water likely trapped more eDNA and prevented it from passing through the filters. The 10‐L filtered sample had more OTUs than combining five replicates of 2‐L filtered samples. Comparison among the same trial revealed that DNA composition in the water is variable, and detected species differed even in the same water sample. Species richness and diversity were higher in winter than in summer, and species compositions differed between seasons. The greater numbers of species detected using filtration additives suggest these methods could be useful for eDNA ocean biodiversity surveys.

INTRODUCTION

Environmental DNA (eDNA) is genetic material released from organisms into the environment (Pilliod et al., 2012) as was introduced by Ficetola et al. (2008), and it has been developed as an alternative species monitoring technique in the last decade. Its merits include convenience and noninvasiveness compared with directly sampling organisms, and eDNA has now become a valuable tool for various surveys, including biodiversity assessment and detection of alien, rare, or endangered species (Bohmann et al., 2014; Keskin, 2014; Pfleger et al., 2016; Thomsen et al., 2012).

Environmental DNA analysis can be broadly used by sampling soil, sediment, water and ice, and a wide range of studies on eDNA in aquatic environments have been reported (Thomsen & Willerslev, 2015). However, most studies were conducted in rivers, lakes or coastal areas within 10 km from land (Garlapati et al., 2019; Yates et al., 2019), and oceanic surveys with eDNA are still rare (e.g., Suter et al., 2020; Truelove et al., 2019). Water sampling in the offshore ocean is not easy because it requires a ship or research vessel with proper equipment (e.g., Niskin bottles). In addition, since the spatial dimension is larger and DNA is likely more diluted in the ocean, large volumes of water are necessary to detect DNA in low densities in the water column (Wilcox et al., 2013).

The recent eDNA studies have provided various types of information including the sources for eDNA collection, such as about substrates for attachment or DNA retention. One observation is that previous studies using eDNA techniques found higher species richness in sediment samples than in paired water column samples (Holman et al., 2019). It has also been reported that suspended particulate materials in aquatic environments were a source of eDNA, which was attached to the particles (Díaz et al., 2020). The higher eDNA levels associated with sediment samples may be caused by particulate matter from organisms settling to the bottom or possibly sticking to sediment particles. Another factor in oceanic or clear freshwater environments might be to retain smaller particles containing DNA, and an experiment using different pore size filters was conducted by Jeunen et al. (2019) with polycarbonate and cellulose-nitrate filter membranes to capture DNA. The 0.2-μm filters showed significantly higher DNA yields than the 1.2-μm filters for both types of membranes when 500-ml coastal water samples were filtered.

We speculated that there may be several techniques that might increase the retention of DNA molecules from water samples during filtration, some of which might pass through the filters used in eDNA studies. In the present study, we added four different materials (sediment, diatomite, zirconia bead and molecular sieve) to water samples in separate treatments prior to filtration and tried a fine pore size filter (0.22 μm) to trap and accumulate eDNA. We also changed the sample volume of oceanic waters off Hokkaido of northern Japan to evaluate the effect of amount of water used. Diatomite was used because it has a characteristic of absorbing water (Elden et al., 2010), and it has been used for DNA extraction with its binding affinity to DNA molecules (Kim et al., 2006). Zirconia beads were also found to increase total DNA detections and species richness (Ushio, 2019). The molecular sieve was included as additive material, even though it is generally used for chromatography to entrap gases (Kosslick & Fricke, 2007) because its rough shape could be expected to be a substrate for DNA attachment that is different from the smooth and circular surface of beads. Bessey et al. (2020) investigated how water sample volume affects the detections of species in a fish community by using various water sample volumes from 25 to 2000 ml, and more species were detected as more water was filtered; thus, to examine the factor of sample volume, we compared 2-L and 10-L water samples.

Efforts to improve the detection rate of eDNA have also been made by finding optimal methods of extraction (Amberg et al., 2015; Deiner et al., 2018) and PCR (Jackson et al., 2017; Uchii et al., 2019), but our goal was to find ways to improve the capture rate of eDNA before the laboratory DNA analysis stage. Therefore, our objectives were to compare the fish species compositions in oceanic water eDNA samples that were obtained using different filtration additives during two different seasons and to evaluate the influence of sample volumes and eDNA distribution within the samples on the detections.

MATERIALS AND METHODS Sampling site

The sampling site was located offshore of Hokkaido Island of northern Japan (Figure 1). This location of the western North Pacific Ocean can be affected by both the Tsugaru Warm Current and the Coastal Oyashio Current. Sampling was carried out twice (summer: 25 August 2020, winter: 22 February 2021) at the same sampling station (summer: 41.9528°N, 141.6678°E, winter: 41.9554°N, 141.6583°E), which is about 45.3 km away from land (Cape Esan), over the deep continental slope (water depth ~800 m; Figure 1). Water temperature and salinity were collected by conductivity temperature depth (CTD, SBE-19 plus; SeaBird Electronics Inc.).

FIGURE 1. Map of the environmental DNA (eDNA) sampling site near Hokkaido in the western North Pacific Ocean. The red circle indicates the sampling station

Water sampling

Water sampling for eDNA filtration was conducted during two cruises of Ushio-Maru, the training ship (T/S) of Hokkaido University (summer: crs-471–2020, winter: crs-484–2021). Surface water (approximately 1 m below from the surface) was sampled from a tap of the ship's continuous flow-through seawater system that is connected to a pump drawing water through the system in real time. To avoid bias and obtain water as evenly as possible for the replicate samples, we made a hose branching system as shown in Figure 2. Eight 10-L volume tanks (low-density polyethylene; Kakuri Co.) were fully filled in simultaneously and were then immediately treated with 10 ml of 10% benzalkonium chloride (Taiyo Pharmaceutical Co.) to prevent DNA degradation (Yamanaka et al., 2017).

FIGURE 2. Method of water sampling on the training ship (T/S) Ushio-Maru; (a) diagram and (b) actual photo. Surface water was sampled from a tap of the ship's flow-through seawater system that is connected to a pump drawing water through the system in real time and prefiltered by 5-mm2 mesh. Water was sampled in eight 10-L tanks at the same time using branching hoses

Experimental design and preparation

Six experimental treatments were designed as follows and as shown in Figure 3. Exp1: pore size 0.45 μm (control); Exp2: pore size 0.22 μm; Exp3: add sediments (0.2 g/L) to water; Exp4: add diatomite (0.2 g/L) to water; Exp5: add zirconia beads (0.5 g/L) into filter cartridges; Exp6: add molecular sieves (0.05 g/L) into filter cartridges. Sterivex, the incapsulated filter cartridges (Merck Millipore) were used for both 0.22-μm and 0.45-μm pore size. For the winter trial, one more experimental treatment (Exp7) was tested to determine the effect of water volume on detection rate. Except Exp2, 0.45 μm pore sizes were used for filtration.

FIGURE 3. (a) Sampling design showing each experimental treatment; sample volume, number of replicates, pore size of filter, and type of additive materials. (b) Photographs of filtration onboard and filter cartridges of each experimental treatment

Additive materials were prepared in a clean laboratory. Sediments were taken from a nearby beach (Hakodate Bay, 41.8132°N, 140.7041°E), and diatomite (Ishigaki Product Co.), zirconia beads (0.5 mm, YTZ-0.5; AsOne), and molecular sieves (Tenax TA; GL Science) were purchased. Sediment and diatomite were ground with a mortar and pestle, weighed, heated at 500°C for 2 h to destroy all DNA that could be contained within the materials, and were carried in aluminum foil enclosed in double zipper bags to the ship. Before the cruise, the heat-treated sediment and diatomite were contamination-tested by electrophoresis to make sure that they did not contain any DNA (Figure S1a). Zirconia beads and molecular sieves were directly filled inside the filter cartridges following Ushio (2019) to prevent clogging the inlet of the cartridge during filtration as the particles are relatively big.

Onboard filtration

The water of each individual tank was divided into five 2-L volume replicates for one experimental treatment (Figure 3). Tanks were shaken continuously to ensure homogeneity across the replicate, and replicates were made to determine the evenness of eDNA inside the tank. The same experiment was conducted during summer and winter to replicate the experiment, and this allowed the seasonal differences to be evaluated. However, we added one more experimental treatment in the second trial in which one whole tank (10 L) was filtered as one sample, and duplicate samples were taken (Exp7). All samples were filtered onboard right after water sampling by an EZ-Stream^TM Pump (Merck Millipore). After filtration, cartridges were immersed in 1.6-ml RNAlater Stabilization Solution (Thermo Fisher Scientific) and frozen in −30°C until DNA extraction. All equipment was disposable or cleaned by DNA off (TaKaRa Bio Inc.), and the ship laboratory was cleaned with 10% commercial bleach solution before use. Negative controls (1 L of Milli-Q water) were prepared for each experimental treatment and filtered in the same laboratory right before experiment to check for contamination. It was confirmed that all additive materials were not contaminated before or during onboard filtration by electrophoresis after PCR (Figure S1b) and final sequence data.

Procedure of eDNA metabarcoding

Samples were sent to the IDEA Consultants Inc. for extraction, paired-end library preparation, and next-generation sequencing (MiSeq) following Miya and Sado (2019) with some modification. In brief, total DNA was extracted from the Sterivex filter cartridges using a DNeasy Blood and Tissue Kit (Qiagen) with an AE buffer volume of 100 μl.

A two-step PCR for paired-end library preparation was employed in the MiSeq platform (Illumina). For the first round of PCR (1st PCR), KAPA HiFi HotStart ReadyMix (KAPA Biosystems) was used with a 3:1 ratio mixture of the universal primer (MiFish-U:MiFish-E). The blank (nuclease free water; Nacalai tesque) was prepared during 1st PCR to monitor any contamination. After completion of the 1st PCR, equal volumes of the PCR products from the eight replicates were pooled and purified using a SPRI select (BECKMAN COULTER) following the manufacturer's protocol. Fragment sizes between 300 to 400 bp were selected during purification. The purified products were employed as templates for the second-round PCR (2nd PCR) that was carried out with dual-index primers. Samples were purified once again after the 2nd PCR by the same method.

The concentration of the dual-indexed libraries was measured using a quantitative PCR (QuantStudio3; Thermo Fisher Scientific) and size-confirmed by TapeStation 4200 (Agilent Technologies). The libraries were pooled and diluted to 4 nM and sequenced on the MiSeq platform using a MiSeq Reagent Kit v2 300 cycle (Illumina) following the manufacturer's protocol.

Data processing and taxonomic assignment

Data preprocessing and analysis of MiSeq raw reads were performed by pipeline (MiFish ver. 2.4) using USEARCH 11.0.667. The applied process and assignment methods were described in Ahn et al. (2020). In brief, (1) forward (R1) and reverse (R2) reads were merged by aligning them with the fastq_mergepairs command. During this process, the following reads were discarded: low-quality tail reads with a cutoff threshold set at a quality (Phred) score of 2, reads that were too short (<100 bp) after tail trimming, and paired reads with multiple differences (>5 positions) in the aligned region (ca. 65 bp). (2) Primer sequences were removed from merged reads using the fastx_truncate command. (3) Reads without primer sequences underwent quality filtering using the fastq_filter command to remove low-quality reads with an expected error rate >1% and reads that were too short (<120 bp). (4) Preprocessed reads were dereplicated using the fastx_uniques command, and all singletons, doubletons, and tripletons were removed from subsequent analysis as recommended by Edgar, 2010. (5) Dereplicated reads were denoised using the unoise3 command to generate amplicon sequence variants (ASVs) without any putatively chimeric and erroneous sequences (Callahan et al., 2017). (6) ASVs were subjected to taxonomic assignments of species names (operational taxonomic units; OTUs) using the usearch_global command with sequence identity >98.5% of OTUs to the reference sequences. ASVs with sequence identities of 80%–98.5% were tentatively assigned “U98.5” labels before the corresponding species name with the highest identity (e.g., U98.5_Pagrus_major), and they were subjected to clustering at the 0.985 level using the cluster_smallmem command. In an incomplete reference database, this clustering step enables the detection of multiple OTUs under an identical species name. We annotated such multiple OTUs with “gotu1, 2, 3…” and tabulated all of these outputs (OTUs plus U98.5_OTUs) with read abundances. ASVs with sequence identities <80% (saved as “no_hit”) were excluded from the above taxonomic assignments and downstream analyses because all of them were found to be non-fish organisms. All negative controls and PCR blanks were also analyzed using this pipeline.

Statistical analysis

To examine the effect of pore size of the cartridge and additive materials, a generalized linear model (GLM) with a Poisson distribution was performed. The number of OTUs detected in each experimental treatment was used as a response variable, and experimental condition (pore size, sediment, diatomite, zirconia bead, and molecular sieve) and season (summer, winter) were explanatory variables without interaction between variables. The data set was found to be applicable for Poisson distribution by a dispersion test; z = −0.70542, p-value = 0.7592, dispersion is 0.907749 when alternative hypothesis is “true dispersion is greater than 1 sample estimates” (package of “AER_1.2–9”).

For seasonal comparison, we calculated species richness, diversity by Shannon–Wiener H index, and Bray–Curtis dissimilarity of the fish community by a principal coordinate analysis (PCoA) of β-diversity (packages of “vegan_2.5–7” and “gclus_1.3.2”). To visualize the bias within and across the treatment in the same season, Venn diagrams (packages of “VennDiagram_1.7.1” and “venn_1.10”) were made to show the overlap of species among different samples. All statistical tests and graphics were produced using R version 4.0.5 (R Core Team, 2018) and edited by Adobe Illustrator 2020.

RESULTS Environment of sampling site

Sampling site is located where the Tsugaru Warm Current and the Coastal Oyashio Current meet south of Hokkaido (Figure 1). In summer, the surface layer is covered by the Tsugaru Warm Current (characterized as higher than 6°C, 33.7–34.3 psu by Ohtani, 1971), and the thermocline is present at 20–140 m (Figure S2a). In winter, the Coastal Oyashio Current (characterized as lower than 3°C, 33.0–33.3 psu by Ohtani, 1971) is present in the surface layer, while middle layer around 50–200 m is affected by the Tsugaru Warm Current (Figure S2b). The water temperature and salinity of samples were 21.4°C, 33.2 psu in summer, and 2.9°C, 33.0 psu in winter.

Number of OTU of each experimental treatment

The number of detected operational taxonomic units (OTUs) in the experimental treatments was different between seasons (Table S1). In summer, 12 OTUs were detected in Exp4 out of one experimental tank (10 L), which was the largest number of detections. The average number of OTUs of the five replicates of 2 L was 5. Four OTUs were detected in both Exp1 (control) and Exp2 for 10 L, and averages were 1.6 and 1.4. The total numbers of OTUs of Exp3, Exp5, and Exp6 were 3, 8, and 6 in 10 L and averages were 1.6, 2.6, and 2.2, respectively (Table 1; Figure 4a,c).

TABLE 1 List of the 18 operational taxonomic units (OTUs) that were detected in the summer trial

	Exp1	Exp2	Exp3	Exp4	Exp5	Exp6
Number of OTU in 2 L	1.6	1.4	1.6	5	2.6	2.2
Number of OTU in 10 L	4	4	3	12	8	6
Engraulis japonicus a	o	o	o	o	o	o
Clupea pallasii b				o
Sardinops melanostictus a	o	o	o	o	o	o
Notoscopelus japonicas		o		o
Symbolophorus californiensis b						o
Gadus chalcogrammus b				o
Lophius litulon b					o
Sebastes spp. b				o
Seriola quinqueradiata a	o	o	o	o	o	o
Pagrus major b				o
Oplegnathus fasciatus b					o
Petroscirtes breviceps b					o
Xiphias gladius b					o
Scomber spp.				o	o	o
Thunnus orientalis b				o
Hyperoglyphe japonica b				o
Paralichthys olivaceus b						o
Thamnaconus modestus	o			o

Note:

Circles (o) represent occurrences.

"Number of OTU in 2 L" is the average number of detected OTUs from each replicate of one experimental treatment (n = 5). "Number of OTU in 10 L" is the total number found from the combination of five 2-L filtered replicates.

^aOTU detected in every experimental treatment.

^bOTU detected in single experimental treatment.

FIGURE 4. Number of operational taxonomic units (OTUs) in one filter cartridge in (a) summer and (b) winter (2-L volume, n = 5) of each experimental treatment, and total number of OTU of one experimental treatment (10-L volume) that combined five replicates in (c) summer and (d) winter. For the box-and-whisker plots, thick lines represent medians, boxes represent ranges of lower and upper quartiles, bars represent minimum and maximum values, and circles represent outlier values

In winter, almost the same number of OTUs were detected in Exp3 (23) and Exp5 (24) out of 10 L, which were the first and second largest numbers. The averages were 9.2 and 11.4. The number of OTUs were similar in Exp1 and Exp2 with total numbers of 6 and 7, and the averages were 3.4 and 3.2. The total OTU numbers of Exp4 and Exp6 were 11 and 19, and the averages were 4.8 and 8 (Table 2; Figure 4b,d). In Exp7, which was conducted only in winter and filtered 10 L for one replicate, 11 and 12 OTUs were detected, with 5 OTUs detected in both samples (Figure 5).

TABLE 2 List of the 39 operational taxonomic units (OTUs) that were detected in the winter trial

	Exp1	Exp2	Exp3	Exp4	Exp5	Exp6	Exp7
Number of OTU in 2 L	3.4	3.2	9.2	4.8	11.4	8	-
Number of OTU in 10 L	6	7	23	11	24	19	11, 12
Squalus suckleyi a			o
Synaphobranchus spp.			o		o
Engraulis japonicus a			o
Clupea pallasii					o		o
Sardinops melanostictus			o		o	o	o
Macropinna microstoma a	o
Leuroglossus schmidti	o		o		o	o	o
Lipolagus ochotensis			o	o		o	o
Oncorhynchus spp.			o				o
Stomias affinis a			o
Tactostoma macropus a						o
Diaphus theta			o	o	o	o
Lampanyctus jordani		o			o
Stenobrachius spp.		o	o	o	o	o	o
Coryphaenoides cinereus	o		o				o
Laemonema longipes a			o
Eleginus gracilis a					o
Gadus chalcogrammus b	o	o	o	o	o	o	o
Gadus macrocephalus a					o
Sebastes spp.		o	o	o	o	o	o
Hexagrammos otakii b	o	o	o	o	o	o	o
Pleurogrammus azonus	o		o	o	o	o	o
Enophrys diceraus					o	o	o
Gymnocanthus herzensteini		o			o		o
Hemilepidotus spp.				o	o	o
Myoxocephalus spp.			o	o			o
Hemitripterus villosus a			o
Podothecus spp.					o		o
Liparis ochotensis a						o
Seriola quinqueradiata a			o
Leptoclinus maculatus					o	o
Stichaeus spp.			o		o
Cryptacanthodes bergi						o	o
Scomber spp.			o				o
Thunnus orientalis					o	o
Lepidopsetta mochigarei				o	o	o
Pleuronectidae		o	o	o	o	o	o
Pseudopleuronectes spp.			o		o
Takifugu spp.					o	o

Note:

Circles (o) represent occurrences.

"Number of OTU in 2 L" is the average number of detected OTU from each replicate of one experimental treatment (n = 5). "Number of OTU in 10 L" is the total number found from the combination of five 2-L filtered replicates (Exp1–6) and 10-L filtered duplicate samples (Exp7).

^aOTU detected in single experimental treatment.

^bOTU detected in every experimental treatment.

FIGURE 5. Venn diagram of operational taxonomic units (OTUs) shared between the duplicate samples (10 L) of Exp7. The species names outside the area of overlap indicate the OTUs that were detected only in one sample, and the species names in the middle indicate the OTUs that were detected in both samples. Numbers in parenthesis indicate the total number of OTUs

The result of the GLM showed that pore size had no significant effect (p >0.1) on the number of detected OTUs. However, season (p < 0.001) and the four additive materials (sediment, diatomite, and molecular sieve p < 0.01; zirconia bead p < 0.001) significantly affected the positive detection rate of OTUs (Table 3). There were weak to no relations between total DNA concentration and the number of OTUs (summer: r² = 0.485, winter: r² = 0.103; Table S2) since total DNA contains non-fish species.

TABLE 3 Results of the general linear model using a Poisson distribution to assess the effects of different experimental conditions and season on the number of operational taxonomic units (OTUs)

	Estimate	SE	z value	Pr (>|z|)
(Intercept)	0.2803	0.22408	1.251	0.21097
Season	1.02165	0.13744	7.434	1.06E−13*
Pore size	−0.08338	0.28893	−0.289	0.77289
Sediment	0.77011	0.24191	3.184	0.00146**
Diatomite	0.67294	0.24578	2.738	0.00618**
Zirconia bead	1.02962	0.23299	4.419	9.91E−06*
Molecular sieve	0.71295	0.24415	2.920	0.0035**

*p < 0.001.

**p < 0.01.

Detected OTUs of each experimental treatment

The detected OTUs not only varied among the experimental treatments, but also they were different within the same treatment (Figure 6). Exp1 and Exp2 of the summer trial had no overlapping OTUs in the five replicates, even though they were filtered from water of the same tank. There were OTUs that were only detected in a single replicate for every experimental treatment. Only 3 OTUs (Japanese anchovy Engraulis japonicus, Japanese pilchard Sardinops melanostictus, and Yellowtail Seriola quinqueradiata) were detected in all 6 of the experimental treatments in summer among the 18 total OTUs. In the winter trial, 2 OTUs (walleye pollock Gadus chalcogrammus and Fat greenling Hexagrammos otakii) were present in all the experimental treatments out of the 39 total OTUs (Figure 7). No detection was confirmed in the final sequence data of all negative controls in summer trial, and some freshwater fishes (e.g., Gobio gobio and Nipponocypris temminchii) were detected in negative controls of the winter trial (Exp1, 2, 5, and 6), which seem unlikely to be present so far offshore in the ocean naturally or by human activities (not food fish), thus may have been caused by carry-over from the sequencer.

FIGURE 6. Venn diagrams of operational taxonomic units (OTUs) shared between replicates (2 L) of the same experimental treatment. Upper row: summer trial, lower row: winter trial, gray: Exp1, yellow: Exp2, green: Exp3, light blue: Exp4, dark blue: Exp5, red: Exp6. The numbers of OTUs in the middle of each diagram indicate the numbers that were detected in every replicate sample of each experiment

FIGURE 7. Venn diagram of OTUs shared between experimental treatments in (a) summer and (b) winter. Gray: Exp1, yellow: Exp2, green: Exp3, light blue: Exp4, dark blue: Exp5, red: Exp6, dark gray: Exp7. Numbers in parenthesis indicate the total number of OTUs in 10 L (five replicates of 2-L sample) for Exp1 to Exp6, and 20 L (duplicate of 10-L filtered sample) for Exp7. The numbers OTUs in the middle of each diagram indicate the numbers that were detected in every experimental treatment in each season

Seasonal changes

Both species richness and Shannon–Wiener H species diversity were higher in winter. The fish community calculated by PCoA showed a clear division between seasons (Figure 8). In addition to the community differences, the frequency of appearance of the 8 OTUs that were present in both seasons were different in summer and winter (Table 4). For example, Japanese anchovy were detected from 28 out of 30 samples in summer but were only detected from 2 samples in winter. Similarly, walleye pollock was only detected from 3 samples in summer but were present in all 30 samples in winter.

FIGURE 8. The (a) species richness, (b) Shannon–Wiener H diversity index, and (c) principal coordinate analysis (PCoA) of the β-diversity with Bray–Curtis dissimilarity, in summer and winter. For the box-and-whisker plots, thick lines represent medians, boxes represent ranges of lower and upper quartiles, bars represent minimum and maximum values, and circles represent outlier values

TABLE 4 Frequency of appearance of eight operational taxonomic units (OTUs) that overlapped in both summer and winter, calculated by the number of detected OTU out of the number of total OTUs

Season	Engraulis japonicus	Clupea pallasii	Sardinops melanostictus	Gadus chalcogrammus	Sebastes spp.	Seriola quinqueradiata	Scomber spp.	Thunnus orientalis
Summer	0.93 (28)	0.03 (1)	0.40 (12)	0.10 (3)	0.03 (1)	0.23 (7)	0.17 (5)	0.03 (1)
Winter	0.07 (2)	0.07 (2)	0.13 (4)	1.00 (30)	0.33 (10)	0.10 (3)	0.03 (1)	0.07 (2)

Note:

Numbers of detections out of the 30 total samples are in parenthesis.

DISCUSSION Uneven eDNA distribution in water

Despite the efforts to homogeneously collect water samples, the OTUs were not evenly detected in water from the same tank. Japanese anchovy was not detected from two replicates (one replicate of Exp1 and 2 each) even if it was one of the major species in summer, with high numbers of reads and frequency of appearance (Figure 6, Table 4, Table S1). Also, variation in the detected OTUs was found between every experimental treatment and within the replicates of same treatment.

Uneven distribution of eDNA in aquatic environments has been reported, including that the concentration of eDNA varied in a small rearing tank with aeration (Ahn, 2020). Bessey et al. (2020) indicated that DNA molecules can be patchy within seawater, and this patchiness may cause a false negative result (not detected even when present) when filtered water volume is insufficient but that can be mitigated by a larger water sample volume (~20 L in temperate seawater). However, an increase in water sample volume was not effective in the study by Suter et al. (2020) in which more species were detected from multiple small volume samples (2 L) than in a single large-volume sample (more than 2000 L). This result might be caused by the coarse filter (270 μm) used for the large-volume sample or because DNA might have become degraded during filtration, which took over 9 h. Both studies obtained single larger water volume samples since it is difficult to sample and filter that amount of water. Therefore, the data were not enough to estimate the effect of water volume.

In the present study, duplicate samples were taken for large-volume filtration (Exp7). It took about 2.5 h for the 10-L filtration, while it took less than 10 min to filter 2 L (approximately 1 h for five replicates including changing filter cartridge and filling RNAlater). The number of detected OTUs in the 10-L samples were larger than the total number of OTU that were obtained from the five combined 2-L replicates of for Exp1, which had the same pore size and the same final water volume (Table 2). The number of OTUs were similar (11 and 12) for the duplicate samples, but only 5 OTUs were in common for the duplicate samples (Figure 5). Additionally, the DNA concentration was quite different in the two samples (125.27 nM and 27.05 nM; Table S2), and the relation between concentration and number of OTU was weak because total DNA included non-fish organisms even after target size selection. These findings are in agreement with Ahn et al. (2020), which also found that neither sample volume nor DNA concentration correlated with fish diversity.

This bias of the detected species could be a limitation of using eDNA for biodiversity surveys. However, bias is inevitable even for conventional monitoring methods such as the capture of the actual animals (Fujii et al., 2019; Sard et al., 2019) and for visual observations (Hayami et al., 2020; Valdivia-Carrillo et al., 2021). Taking multiple samples of eDNA with adequate filtration volume can be an easy and cost-effective solution that can lessen the bias to reduce the occurrence of false negative species (present but not detected) as shown in Figure 6.

Efficiency of additive materials

Additive materials significantly affected the detection rate (Table 3). The morphological features of additive materials were varied (Figure S3), but the typical porous shape of diatomite could not be observed because it had been ground and heated to avoid contamination. As the efficiency of the additive material varied by season, the number of detected OTUs seemed to be more affected by DNA distribution bias than by the shape or chemical composition of the additive materials. This might explain the low OTU number of Exp3 in summer. However, the additive material did have a positive effect on capturing the existing DNA according to the GLM results (Table 3). Ushio (2019) used zirconia beads to apply a bead-beating step during extraction to maximize efficiency within the encapsulated structure of the filter cartridges. Instead, we focused on using the beads as possible sites of DNA attachment because it was found that DNA could be extracted off of microplastic from the ocean (Debeljak et al., 2017).

Even though there was a report that water turbidity has negative effect on eDNA detection rate (Wittwer et al., 2018), it may mostly be a problem of decreasing the filtration volume. However, filter cartridges in the present study did not have the limitation of clogging, and the filtration times with additives were the same as in Exp1 (without additive material). In addition, Wittwer et al. (2018) used water from a stream in which sediment and suspended materials caused water turbidity that may include PCR inhibitors such as humic acid (Uchii et al., 2019). Our additive materials were heat-treated (sediment and diatomite) or newly opened before use (bead and molecular sieve), which are free from those inhibitors. Considering that we found no disadvantages and found increased numbers of OTUs, adding material during filtration to capture eDNA can be an easy and effective method to enhance detection rates.

Contrary to the additive materials, pore size of the filters did not affect the detection rate (Table 3) because the number of OTUs of Exp1 (0.45 μm) and Exp2 (0.22 μm) had no significant differences in both seasons (Tables 1 and 2). A previous study reported that fine pore size showed better performance for DNA yield with the same amount of water volume (Jeunen et al., 2019). However, DNA yield was higher with coarse pore size filter when water was filtered until clogging since the coarse filter had the capacity to filter a greater volume of water per filter. Therefore, the efficiency of fine pore membrane filters remains unclear in relation to filtration volume. Also, Jeunen et al. (2019) did not compare the species richness of different pore sizes, and richness does not seem to have a strong relation with DNA yield as mentioned above. Deiner et al. (2018) and Bessey et al. (2020) also tested different pore sizes and concluded that pore size had no influence on the estimated richness of detected OTUs, which is in accordance with our result.

Characteristics of detected species

A total of 49 OTUs (summer: 18, winter: 39, both: 8) were detected in the surface water at the sampling station within the 62 samples (Table S1). The composition of the OTUs appeared to reflect the geographic location where the species community is mixed by both the Coastal Oyashio and Tsugaru Warm Currents (Figure 1). For example, yellowtail (Yamamoto et al., 2007) and Pacific bluefin tuna Thunnus orientalis (Fujioka et al., 2015) migrate from the south with the Tsugaru Warm Current and Kuroshio Current while walleye pollock comes from the Okhotsk Sea with the Coastal Oyashio Current (Tanaka et al., 2017).

Results of the GLM showed that season significantly affected the number of OTUs (Table 3), and the species richness and diversity were higher in winter (Figure 8a,b). This might be caused by the species that migrate to Hokkaido (e.g., walleye pollock, salmon Onchorhynchus spp). It also appears that eDNA from the bottom may have moved to the surface by upwelling since the water column is vertically mixed in winter, or benthic/demersal species moved upward as surface water temperature decreased (Figure S2). Indeed, the proportion of benthic-demersal fish increased to 51.3% in winter compared with 16.7% in summer (classified by Nelson, 2006). Another possible reason is that the eDNA decay rate would be faster as temperature increased, and the eDNA would exist longer after being released in a colder environment (Eichmiller et al., 2016; Strickler et al., 2015), and the surface water at our sampling site differed by about 18°C between the two seasons.

The fish community was clearly divided by season (Figure 8c). It is not just because only 8 OTUs were overlapped, but also those OTUs showed seasonal differences. For example, the number of reads of Japanese anchovy was much larger (Table S1) and appeared more frequently (Table 4) in summer than winter, even though they were distributed in the study area during the entire year (Nakabo, 2013). This is because the summer sampling date was during the spawning season of the Japanese anchovy in the study area (May to September). High concentrations of eDNA were detected after spawning events due to the release of genetic material into the environment (Takeuchi et al., 2019; Tillotson et al., 2018). The frequency of appearance of yellowtail was high in summer and low in winter as well. Yellowtail are known to migrate northward in summer and then southward in winter (Yamamoto et al., 2007). However, some catches of yellowtail in winter have been reported recently (The catch quantity data of Hokkaido, http://www.pref.hokkaido.lg.jp/sr/sum/03kanrig/sui-toukei/suitoukei.htm) that are probably caused by global warming. Catch data have also showed that walleye pollock has increased in the winter season (October to February), which is consistent with the frequency results (Table 4). Metabarcoding research has also detected marine mammals such as the Dall's Porpoise in the study area, and their major prey are these abundant species of Japanese anchovy, Japanese pilchard, and walleye pollock (Ohizumi et al., 2000) that occurred both summer and winter.

CONCLUSIONS

Our results demonstrated the effectiveness of additive materials (sediment, diatomite, zirconia bead, and molecular sieve) during filtration to enhance detection rates of eDNA in oceanic waters. Even if the pattern differed by season, all materials significantly increased the number of OTUs detected, probably because the eDNA in water attached to the additive materials and remained within the filter cartridge. In addition, the detected OTUs varied among the 30 replicates (five replicates of six experimental treatments) of the same sampling station due to the uneven distribution of eDNA within the water samples. Therefore, taking multiple samples is needed to obtain reliable data even if additives are used, and this combination of approaches may be an effective method for ocean eDNA surveys. Filtration additives can also be tested for use in studies in clear-water freshwater environments to increase the efficiency of species detections.

ACKNOWLEDGEMENTS

We are grateful to captain Kamei and the crew members of the T/S Ushio-Maru (Cruises: crs-471-2020 and crs-484-2021). Also, we appreciate Dr. Takahiro Iida for checking the species list and Prof. Tetsuya Takatsu for valuable comments. This study is supported by Balance de Ocean program of Hokkaido University and was funded by the Hirose Foundation and Maekawa Houonkai Foundation.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

AUTHOR CONTRIBUTIONS

HA, YA, SS, and AO conceptualized this study. Onboard water sampling and filtration were conducted by HA and YA. HA did data curation, analysis, and writing the first draft. The funds were acquired by HA. All authors gave final approval of the version to be published.

DATA AVAILABILITY STATEMENT

All relevant data are included in the Appendix S1. (Species detected in this study are listed in Table S1. Concentration of total DNA and number of OTU were listed in Table S2.) All FASTQ files have been submitted to DDBJ (Accession No.: DRA012157).