Nanopore environmental DNA sequencing of catch water for estimating species composition in demersal bottom trawl fisheries

Bycatch and discards, representing unwanted catches, undermine sustainable fisheries and hinder the conservation of vulnerable and endangered species. To effectively monitor bycatch and enhance the effectiveness of management measures while promoting sustainable fishing practices, reliable data is essential. Here, we explore the use of Nanopore metabarcoding to analyze the catch composition in demersal bottom fisheries. We collected eDNA samples directly from an onboard catch holding tank ( catch water ) for 10 fishing hauls from a fishing vessel operating in the Skagerrak (North-East Atlantic). The approach involved sequencing a combination of long ( ~ 2 kb) and short ( ~ 170–313 bp) mitochondrial amplicons and was validated by analyzing a fishery-related mock community sample and fishing haul replicates. Overall, the detection rate accuracy was 95% for landed species, and replicates obtained from the same fishing haul showed consistent results, validating the robustness of this approach. The detection rate accuracy for all caught species observed on board (including the non-landed fraction) was 81%. Undetected species were always limited to species in low abundance, but may also be attributed to problems with identifying closely related species due to the impact of sequencing errors and limited diagnostic variation in the genetic regions used. In the future, such biases may be reduced by using additional markers to increase species discrimination power and applying newly available technological advantages in flow cell chemistry to improve sequencing ac - curacy. In conclusion, this study demonstrates the effectiveness of Nanopore eDNA sequencing of catch water for estimating species composition in demersal bottom trawl fisheries, including catches of non-commercial and threatened and vulnerable species, without disrupting fishing activities. Incorporating eDNA analysis of catch water may therefore help facilitate effective monitoring, leading to better-informed


| INTRODUC TI ON
Bycatch and discards are critical global issues that undermine sustainable fisheries, jeopardize the conservation of vulnerable and endangered species, and impede effective fisheries management (Hall et al., 2000;Pérez Roda et al., 2019).Discards of commercial fish species, in particular, present challenges for stock assessment by introducing uncertainty into the estimation of fishing mortality (Guillen et al., 2018;Zeller et al., 2018).To address the discard issue in Europe, the European Union landing obligation was fully implemented in 2019, mandating that all species with a quota, or minimum landing size, in the fishing area must be recorded and landed (with some exceptions), as defined in Article 15 of the basic regulation EU 1380/2013 underpinning the Common Fisheries Policy (CFP) (EU, 2013).Once landed, bycatch species are counted against the quotas of the species.Despite the phased implementation of the EU landing obligation aimed to incentivize fishers to adapt fishing patterns or modify their vessels (e.g., Cosgrove et al., 2019;Kennelly & Broadhurst, 2021), monitoring fishing activities and collecting data on bycatch and discards remain challenging due to the complex nature of marine ecosystems and the difficulties in exercising direct control (Condie et al., 2014;Guillen et al., 2018).Indeed, the success of the landing obligation can only be achieved through effective monitoring and comprehensive data collection (Guillen et al., 2018).
As recording and landing bycatch species require additional time and space on board, the handling costs associated with such activities can be expensive for fishers.Several methods have been used to enhance data collection on bycatch and discards, including self-sampling programs (e.g., Figueiredo et al., 2020), at-sea monitoring programs (involving onboard scientific observers) (Ewell et al., 2020), and electronic monitoring (consisting of onboard videos) (Emery et al., 2018).These programs are crucial for data collection, yet they can be time-consuming, expensive, disruptive to regular fishing operations, and often limited to commercial species (James et al., 2018).
Among different fishing gears, bottom trawling is known to produce high levels of discards (Kennelly & Broadhurst, 2021) and have a significant impact on the seafloor and the benthic community (Hiddink et al., 2019;Rijnsdorp et al., 2020).On average, annual discards have been estimated at 9.1 million tons worldwide, and bottom trawling accounts for 45.5% of total annual discards (Pérez Roda et al., 2019).Among target species, fisheries targeting demersal fish and crustaceans have the highest discard rates (Pérez Roda et al., 2019).For instance, it was estimated that when using conventional trawling methods to catch 1 kg of Norway lobster, approximately 4.5 kg of discards (equivalent to 4.1 kg of fish) are generated, with the bycatch of fish significantly decreasing when using sorting grids (Ziegler & Valentinsson, 2008).
Here, we tested the possibility of using environmental DNA (eDNA) to estimate the whole catch and bycatch composition of commercial bottom trawling targeting Norway lobster (Nephrops norvegicus) and roundfish (such as Gadus morhua and other gadoids) in the Skagerrak, North-East (NE) Atlantic Ocean.eDNA analysis involves obtaining genetic material directly from the environment (e.g., seawater, air, soil) and has proven to be a cost-effective and non-invasive tool for monitoring and conservation, including in marine ecosystems (Fediajevaite et al., 2021;Gold et al., 2021;Guardiola et al., 2016;Holman et al., 2019;Kirtane et al., 2021;McClenaghan et al., 2020;Miya, 2022;Stat et al., 2017).The method offers a promising tool to estimate catch composition without disrupting fishing operations (Afzali et al., 2021;Kirtane et al., 2021;Stoeckle et al., 2021;Thomsen et al., 2016).It has been successfully used to estimate the species catch composition of samples collected in various fisheries-related environments, including meltwater from industrial and artisanal fishing vessels (Willette et al., 2021), silage and fish blocks (Hansen et al., 2020), water draining from trawling nets (Russo et al., 2021), and from a custom-made device used in bottom trawling operations (Maiello et al., 2022(Maiello et al., , 2023)).
Furthermore, eDNA has recently emerged as a promising tool for assessing the abundance of Atlantic mackerel bycatch in Atlantic herring pelagic fisheries (Urban et al., 2022), as well as analyzing catch mixtures of European sprat and Atlantic herring (Urban et al., 2023(Urban et al., , 2024)).Finally, comparisons between eDNA in seawater and bottom trawl catches revealed significant correlations, demonstrating the overlap in species detection in different environments, including subarctic deep-sea habitats (Thomsen et al., 2016), coastal areas (Stoeckle et al., 2021), demersal fish communities (Afzali et al., 2021), and oceanic waters (Salter et al., 2019).Its advantages, including practicality, cost and time efficiency, and reproducibility, along with its high resolution for species identification, make eDNA an excellent choice for monitoring (Deiner et al., 2021) and wider applications in fisheries sciences (Jacobsen et al., 2018).
Here, we explored the potential of eDNA to estimate whole catch composition, including bycatch and discards of bottom trawling fisheries, including species not subject to the landing fisheries management, biodiversity conservation efforts, and the implementation of relevant legislation such as the EU landing obligation.

K E Y W O R D S
bottom trawling, bycatch and discards, eDNA, landing obligation, marine fish species, metabarcoding, monitoring, Nanopore sequencing obligation, in a non-invasive manner.Samples were collected directly on board the vessel from a tank in which the entire catch was briefly kept immediately after onboarding and before sorting activities, hereafter referred to as catch water.Sequencing was conducted with the Nanopore MinION Mk1C portable and realtime sequencing machine to develop a rapid and feasible method that could be used, for example, at landing facilities.Nanopore sequencing has been successfully used for species identification in field-based laboratories (Maestri et al., 2019;Parker et al., 2017;Truelove et al., 2019;Watsa et al., 2020), thus highlighting the potential for its application in fisheries-related environments (Jacobsen et al., 2018).We evaluate the practicality and reliability of eDNA as a tool for monitoring and estimating catch composition, particularly in fisheries with high discards, such as bottom trawling, and to contribute to data collection in the context of fisheries management and conservation, especially regarding vulnerable and endangered species.The primary aims of our study are to assess the abilities of eDNA analysis of catch water for the following purposes: (1) detecting the presence of landed species, (2) providing insights into unwanted catches (bycatch and discards), (3) monitoring elusive and vulnerable species within the catch, and (4) understanding the relationship between fishing depth, target fishery, and catch composition.

| Sample collection of catch water
Samples were collected on board a commercial bottom otter trawler fishing in the Skagerrak (ICES subdivision 20) from a catch holding tank in January 2021.The sampling included 10 fishing hauls, targeting either Norway lobster (five hauls) or roundfish (five hauls), such as Atlantic cod (Gadus morhua).Once the trawling net (mesh size 90 mm) was taken on board, the entire catch from the haul was released into a tank before processing activities were initiated (i.e., sorting of the catch).As the catch was transferred into the emptied tank, residual seawater associated with the catch and net, along with additional seawater used for briefly rinsing the net and operating area, accumulated at the bottom of the tank.This water, referred to as catch water, was collected for the project.Sampling was conducted within minutes after the transfer of the catch to the tank to avoid disrupting fish processing operations on board.Following the processing of the catch, which involved sorting and subsequent emptying of the tank, both the tank and the catch processing station were cleaned with fresh water to prepare for the next haul.For each of the 10 fishing hauls, three catch water samples were collected from the F I G U R E 1 On the left side, the sampling locations in Skagerrak are shown.Red circles indicate fishing hauls targeting Nephrops, while yellow triangles represent fishing hauls targeting roundfish species.Three catch water sample replicates were analyzed for the fishing hauls indicated by an asterisk.On the right side, the top section compares species detected in the eDNA and observed on board for Nephrops and roundfish samples.In the bottom section, the number of species identified in the eDNA that were also observed on board for each fishing haul is shown (eDNA refers to the detection of all four markers pooled).Species are colour-coded based on their lifestyle (FishBase), as indicated in the legend.
tank in 50 mL centrifuge sterile tubes.Based on the target species, we refer to these samples as either Nephrops or roundfish samples (Figure 1).Notably, Nephrops were fished during the day at shallow depth (average mean depth = 80.37 m, max = 172.78m, min = 54.01 m), whereas roundfish were fished at deeper depth (average mean depth = 237.74 m, max = 305.89 m, min = 183.38 m) at night (Table S1).The samples were stored at −20°C until further processing.In addition to collecting water samples, we documented the species compositions in each catch by collecting a list of species observed during the initial 10 min of catch sorting activities for each haul (Table S2).However, species observations were mostly limited to species under the landing obligation.
Additionally, we have data on the weight of landed species per fishing haul (Table S2).Here, we refer to the combination of these data as observed onboard species.

| Catch water DNA extraction
DNA extraction was performed in a dedicated clean laboratory facility at DTU Aqua (Technical University of Denmark, Silkeborg, Denmark).Although we collected three catch water samples for each fishing haul, eDNA analysis was conducted on only one sample per haul, except for a subset of four hauls.In these four hauls (two for each target fishery), all three catch water samples were analyzed to assess reproducibility.After thawing the samples at room temperature for 4 h, a total of 20 mL of catch water was pressed through a 0.22 μm Sterivex filter (SVGPL10RC; Sigma-Aldrich, St. Louis, MO, USA) using a sterile 60 mL disposable syringe (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) to retain DNA and cells.
We followed a modified version of Spens et al. (2017) extraction protocol for DNA extraction from filters using the DNeasy blood and tissue kit (Qiagen, Hilden, Germany).The final elution volume was adjusted to 60 μL of Buffer AE, and samples were stored at −20°C.Each extraction batch included a blank sample to test for exogenous DNA contamination from reagents and the laboratory.
Surfaces and laboratory equipment were cleaned with bleach solution, distilled water, and 70% ethanol before and after sample processing in each batch.DNA concentration was measured for all samples using a Qubit fluorometer and the Qubit dsDNA high sensitivity kit (ThermoFisher Scientific, Waltham, MA, USA).

| Genetic markers
Primers were chosen to target teleost fish species, elasmobranchs and Norway lobster.We used four different primer sets to amplify short and long regions of the mitochondrion genome.These included: MiFish-U (targeting 170 bp ca. of the 12S rRNA gene in bony fish (Miya et al., 2015)); Fish_12S-16S-ONT (here referred to as fish-2 kb, targeting a 2 kb region of the 12S-16S rRNA genes in fish and elasmobranchs (Doorenspleet et al., 2021)); Leray-XT (amplifying ~313 bp of the cytochrome c oxidase subunit I gene, COI, in metazoan (Wangensteen et al., 2018)); metazoan-2 kb (targeting a 2.4 kb region spanning the 12S-16S rRNA genes in bony fish, elasmobranchs, and hagfish).The metazoan-2 kb primer set was made using a combination of primers available in the literature: the forward primer was designed by Machida et al. (2012) while the reverse was developed by Kelly et al. (2016).The final PCR reactions were performed in 20 μL reaction volumes with 4 μL of DNA.Each PCR run consisted of 40 cycles.
For every eDNA extract, three reactions were performed, and each PCR batch included a negative control.Details on the primer sets and PCR amplification can be found in Supplementary Material 1.

| Library preparation
After PCR amplification, 2 μL of amplicons were run on a 2% agarose gel (0.7% for long region markers).When visualizing successful amplification, the three PCR replicates per eDNA extraction sample were pooled.No positive amplification of the PCR-negative controls or extraction-negative controls was observed; therefore, they were not sequenced.Pooled DNA extracts were cleaned using Agencourt AMPure XP magnetic beads in accordance with the manufacturer's specifications (PCR product:1.8xbead ratio).However, it is more optimal to adjust the bead ratio based on amplicon lengths.DNA concentration was then measured using the Qubit dsDNA high sensitivity kit.Next, the purified DNA fragments (MiFish, COI, fish-2 kb, and metazoan-2 kb) were mixed in equimolar ratios for each of the 18 samples individually, and 100 fmol of the pooled amplicon mix was used for the library preparation.A library was prepared for each sample separately using the Nanopore Amplicons by Ligation (SQK-LSK110) sequencing kit, according to the manufacturer's instructions.Sequencing was performed on a Flongle flow cell (FLO-FLG001) separately for each of the 18 samples (including 10 fishing hauls and eight sample replicates).Following the protocol recommendation, 20 fmol of the final library was loaded onto the flow cell, and the sequencing run started, enabling live basecalling using Guppy with the high accuracy (HAC) model.The sequencing runs were set at a maximum of 24 h and lasted, on average, 22.9 h.The basecalled reads were then exported for further analysis.

| Bioinformatics analysis of Nanopore reads
We used a custom pipeline for pre-processing the reads.The preprocessing pipeline involved quality and length filtering of the reads using NanoFilt (quality scores≥10) (De Coster et al., 2018) and primer and adapter removal using Cutadapt 2.4 (Martin, 2011), with a maximum error rate of 0.2.Only reads for which both forward and reverse primers were located were trimmed and retained, while untrimmed reads were discarded.In this way, we ensured that only reads from the correct amplicon regions were used in the subsequent analyses.Analyses were carried out separately for each of the used primer sets, leading to four markerspecific fastq files for each sample.
For taxonomic assignment, we used MetONTIIME (version 1), a metabarcoding pipeline for analyzing ONT data in the QIIME2 framework (Matoute et al., 2024).The MetONTIIME analysis started from the trimmed fastq files and was run using BLAST as a taxonomic classifier with the following parameters: maximum number of hits 3, minimum alignment identity threshold of 0.94, and minimum query coverage of 0.8 and 0.2 for short and long regions, respectively.If there was a discrepancy in the identity, the reads were classified automatically only at a higher taxonomic level.
We used a local reference database comprising COI, 12S, and 16S rRNA genes, along with whole mitogenome sequences of Norway lobster and North Sea fish species, including agnathans and elasmobranchs.Initially, we downloaded a list of 196 North Sea fish species from FishBase, which includes teleosts, agnathans, and elasmobranchs.Subsequently, we built the database by merging the MIDORI2_UNIQ_NUC_GB252 CO1, lrRNA (mitochondrial 16S ribosomal RNA), and srRNA (mitochondrial 12S ribosomal RNA) databases (Leray et al., 2022) and filtering it for North Sea fish species.Additionally, whole mitogenome sequences of North Sea fish species and Norway lobster were downloaded from NCBI and incorporated into the previously built Midori database.
Alongside, the COI sequencing reads were also analyzed using the MIDORI2_UNIQ_NUC_GB252 CO1 database, not restricted to North Sea species.The output of the taxonomic assignment was then imported into R version 4.2.2 for statistical analyses (R Core Team, 2022).

| Validation of primers and Nanopore sequencing using a mock community sample
A mock community was analyzed to evaluate the use of the different primer sets, Nanopore sequencing, and the bioinformatics pipeline.
The mock community contained 12 marine species (Appendix S1).
DNA extracts from tissue samples were pooled by concentration.
PCR amplifications were carried out as described for the catch water samples, with the only difference being that 2 μL of final DNA was used instead of 4 μL to account for the higher concentration of the samples.The library preparation followed the same protocol as described for the eDNA samples; however, two different libraries were prepared for long and short region amplicons and sequenced on different flow cells.Results were analyzed following the same bioinformatics pipeline as described for the eDNA samples.
Based on the mock analysis results, a primer-specific abundance threshold was chosen to avoid false positives (i.e., species not included in the mock sample but found among the DNA sequences) resulting from Nanopore sequencing errors and erroneous taxonomic assignments.We calculated the maximum relative abundance of a false positive for each marker in the mock sample, i.e., the number of reads assigned to an erroneous species divided by the total number of reads assigned at the species level.The ratios were subsequently used as filters in the eDNA samples to discard species with abundance below or equal to this threshold.

| Species filtering
The outputs of the taxonomic assignment were imported into R version 4.2.2 (R Core Team, 2022).At this point, primer-specific taxonomic assignments were available for each sample.Our analysis was restricted to the species level, as our aim was to test the feasibility of using eDNA to estimate species composition.Accordingly, results exclusively assigned at higher taxonomic levels due to discrepancies in taxonomic annotation were filtered out.Additionally, we filtered for species identified with at least three reads.Subsequently, primerspecific abundance thresholds were applied to filter out false positives based on the results of the mock community (Appendix S1).The relative abundance of each species was calculated by dividing the number of reads assigned to a certain species by the total number of reads assigned at species level for each sample primer-specific dataset.Species detected with a relative abundance below or equal to the primerspecific abundance thresholds for false positives were discarded.
Further, closely related fish species (identity ≥98%) were manually identified for the MiFish and long region markers designed in the 12S and 16S rRNA genes, due to low interspecies variation within these genes compared to COI.As observed in the mock data (Figure S1), closely related species, such as Lophius piscatorius and L.budegassa, were detected in the sequencing reads, although the latter was not included in the mock samples (analogously for Clupea harengus and Sprattus sprattus).Interestingly, the proportion of reads of the closely related false positive species was always negligible compared to the species added in the mock (range 0.09%-1.5%,mean = 0.7%).We chose 1.5% as a threshold based on the results of the mock community analyses.Hence, for each sample, species in a closely related group with less than 1.5% of the group reads were discarded from further analyses.In four samples, the fish-2 kb marker did not amplify.However, we observed that the addition of fish-2 kb resulted in an increase of 0.85 species per sample on average (SD 1.09).For this reason, the primer set was kept in the data analyses.

| Statistical analysis
After filtering the results of the taxonomic assignment, we aggregated the relative abundance of detected species in each sample across primer sets, converting the dataset into a phyloseq object.Alpha and beta diversity were calculated using phyloseq (version 1.42) (McMurdie & Holmes, 2013), with species richness and the Jaccard distance, respectively.Principal coordinate analysis (PCoA) plots of beta diversity were created using the Jaccard distance of dissimilarity matrix (based on presence or absence).To test for differences in the species detected among samples and evaluate the effects of sampling replicates (fishing hauls), target fishery (Nephrops vs. roundfish), and mean depths of the hauls, permutational multivariate analyses of variance (PERMANOVA) were performed.PERMANOVAs were based on the Jaccard distances and conducted using the adonis2 function in vegan (version 2.6-4) with 9999 permutations (Oksanen et al., 2022).To assess the beta diversity in a semi-quantitative way, PCoA plots were generated separately for each of the four markers using Bray-Curtis dissimilarity matrices based on species composition and relative abundance.Additionally, for landed species, correlations between relative read abundance and landed weight in kilograms across genetic markers were examined using linear regressions in R.

| Landed species
A total of 10,045 kg was landed from the 10 fishing hauls, including eight bony fish species (4 families), thorny skate (Amblyrajaradiata, Rajidae), and Norway lobster.On average, each haul yielded around 280 kg (ranging from 121 to 410 kg), with individual species weights within the haul ranging from 1 to 303 kg (Table S2).

| Mock analysis
All the species in the mock sample were successfully detected using the four primer combinations.COI and MiFish (short region markers) successfully detected all metazoan and teleost species, respectively, while the long region markers had some limitations in species detection (Figure S1).When the MIDORI2 COI reference database (not restricted to North Sea species) was used, COI read misclassification was observed (Figure S1).

| Catch water sequencing results
Nanopore sequencing generated 239,539-1,704,608 sequences per sample (average = 631,164; SD = ±388,556) (Table S3).After filtering and discarding species with less than three reads, species with relative abundance below the primer-specific threshold, and closely related species with reads below 1.5% of the species group, the average was 119,454 (SD = ±89,520).
Using the North Sea database-comprising Norway lobster, myxine, elasmobranchs, and fish species found in the North Seaacross the 18 samples and four primer sets, 52 unique species were detected, representing 46 genera, 28 families, 17 orders, and four classes (Actinopterygii, Myxine, Chondrichthyes, and Malacostraca).
COI found more species compared to other markers across samples and taxonomic levels, aligning most closely with the diversity observed on board (Figure 2).
Combining the results across primer sets, the number of species detected in the eDNA ranged from 19 to 34 (mean = 25.2,SD = ±4.45)per sample.The roundfish samples contained on average 22.1 species (SD = ±2.3),while Nephrops samples on average contained 28.3 species (SD = ±3.8).Accordingly, the number of species detected in Nephrops samples is significantly higher compared to roundfish fishery samples (Wilcoxon rank-sum test: N = 18, W = 75.5,p = 0.002).
However, no significant difference was found between the target fisheries in the number of species reported on board (Wilcoxon rank-sum test: N = 10, W = 19, p = 0.198).It is worth noting that there is a bias in the onboard observations, as they were mostly restricted to species subject to the landing obligation.Therefore, they are by definition forced to be more similar, as it removes differences in unrecorded species between catches.Species detection comparisons between primer sets are available in Appendix S1.

| eDNA detection of species observed on board
The eDNA method detected an average of 81.89% of the species reported by visual observations (including species landed) across all samples (SD = ±11.19).Notably, in Nephrops samples, eDNA detected 84.75% on average (SD = ±12.98) of the onboard observed species, while in roundfish samples, the detection rate was 79.03% on average (SD = ±8.89)(Figure 3).Nephrops samples showed a F I G U R E 2 Taxa detected across different taxonomic levels by the four markers.From left to right: COI, fish-2 kb, metazoan-2 kb, MiFish, and species observation on board.
Interestingly, species reported on board but not detected in eDNA analysis were associated with low numbers of individuals (1-4 individuals, mean = 1.43,SD = ±0.85).When restricting the analysis to landed species, the detection rate was 95.32%.The eDNA assay detected all the landed species with the exception of 2 kg of Melanogrammus aeglefinus (haddock) in haul 11 and 15 kg of Norway lobster in haul 3 (sample replicates 2 and 3).Although few haddock reads were detected with the metazoan-2 kb marker in haul 11, their high sequence similarity with cod and Merlangius merlangus (whiting) reads, combined with their low abundance (falling the 1.5% abundance threshold), led to their exclusion during the initial analysis.Also, MiFish does not differentiate between haddock and whiting (100% identical).Furthermore, eDNA amplification for haul 11 proved challenging, resulting in multiple PCR attempts and minimal product amplification, which potentially negatively affected species detection efficiency.

| Replicates
Field sample replicates were analyzed in four hauls (Nephrops hauls 5 and 9; roundfish hauls 3 and 4).Alpha diversity among sampling replicates was found to be highly similar for most replicates (Figure S6).
Venn diagrams showed that the majority of species were consistently detected across all three sample replicates within each fishing haul (Figure S7).However, incorporating sample replicates resulted in an average increase of 3.41 additional species detected in the eDNA (SD = ±2.27).For instance, Norway lobster in haul 3 (15 kg landed) was detected only in one out of three sample replicates.The PCoA plot, based on the Jaccard distances, showed that replicates from the same fishing haul clustered together (Figure S6).Furthermore, the PERMANOVA analysis indicated a significant difference among fishing hauls (p = 0.0001), explaining 87% of the variation.This finding suggests that intra-variability within each haul is lower compared to the inter-variability among samples from different fishing hauls.
Therefore, there is a significant effect of the fishing haul, and the diversity observed across replicates within the same haul is consistent.

| Fisheries-related diversity
For subsequent analyses, we combined the species found across sample replicates within the same fishing haul into consensus samples.We analyzed the species composition of the samples to determine whether significant differences exist between the Nephrops and roundfish targeted fisheries.The PCoA ordination plot, based on the Jaccard distances of dissimilarity, revealed a clear clustering of samples into two groups along the first axis (Figure 4).The PERMANOVA analysis confirmed a significant difference between the two target fisheries, Nephrops and roundfish, accounting for 31% of the observed variance (PERMANOVA p-value <0.05).Moreover, fishing depth had a stronger effect, explaining a larger portion of the variance (40%, PERMANOVA p-value = 0.006).This is shown by Nephrops sample haul 2, which clustered on Axis 1 with the roundfish samples in the PCoA.The mean fishing depth of haul 2 is deeper (−172 m) compared to the average shallower depth of Nephrops fisheries samples (−57.2 m), excluding haul 2.Moreover, the same pattern emerged when plotting the PCoA of species diversity observed on board (Figure 4).Haul 2, once again, clustered with the roundfish samples.These findings support the association between species composition and the depth at which the fishery was conducted.
Whereas, we observed distinct clustering patterns in the PCoA plots generated using Bray-Curtis distances for each marker separately (Figure S8).Specifically, we consistently observed the clustering of sample replicates within each fishing haul across all markers, indicating the robustness of the analysis.However, the overall clustering pattern across markers showed less consistency, suggesting variability in species composition among different markers.

| Associations of detected species with target fisheries and fishing depth
We found associations between the species detected in eDNA, the target fisheries, and the depth of fishing hauls.Specifically, Lesueurigobius friesii was consistently detected in all four shallowwater samples collected during the Nephrops fishery (Figure 5).In contrast, Phycisblennoides was consistently detected in the deeperwater samples, including four roundfish samples and the deeperwater Nephrops haul 2. Likewise, for elasmobranchs, Etmopterus spinax was detected in five out of the six deeper water hauls and in Nephrops haul 9, although very few reads.Chimera monstrosa was detected in the eDNA of roundfish hauls 11 and 12, as well as in the eDNA of deeper-water Nephrops haul 2. In contrast, Scyliorhinus canicula was detected in three of five Nephrops hauls, and Amblyraja radiata was observed in both Nephrops and roundfish samples, indicating its presence as bycatch in both fisheries.

| Quantitative analyses
We investigated correlations between the relative abundance of reads and landed weight (in kilograms) across genetic markers.

F I G U R E 3
Comparison of species detection between eDNA and observation on board, where eDNA refers to the pooled detection of all four markers.(a) Percentage of species observed on board that were also detected in the eDNA (green) or undetected in the eDNA (blue).(b) Percentage of species that were detected only in the eDNA (magenta), both observed on board and detected in the eDNA (green), and observed on board only (blue).Species numbers are provided in Figure S5.In this study, we demonstrate the effectiveness of eDNA sequencing of catch water for estimating the species composition of demersal fisheries.This innovative approach involves the analysis of samples collected directly from an onboard tank used to briefly hold the entire catch before sorting activities.Our findings indicate that by analyzing eDNA in catch water, data can be collected on bycatch and discards, including species not subject to the landing obligation, all while maintaining cost-effectiveness and, importantly, without causing disruptions to fishing activities.

| Genetic markers
Using a four-primer combination was essential for effectively capturing fishery-related biodiversity in both the mock community and eDNA samples.These primers did, however, perform differently.
Our study did not specifically aim to compare short and long region primers; however, we found important differences that highlight the need for further investigations, especially as long-read sequencing methods like Nanopore sequencing become increasingly accessible for eDNA analysis (Toxqui Rodríguez et al., 2023).In general, we observed that short region markers like COI and MiFish generally outperformed long region (~2 kb) markers in terms of species detected.It is possible that this may be attributed to the sequencing of eDNA from external sources, which has already been degraded and therefore was not amplified using the long region markers (Holman et al., 2021).External eDNA might originate from the sea water itself, DNA from the stomach contents of the caught species (e.g., from prey), and residual eDNA contaminants from prior catches in the tank.However, long-region markers did also fail to detect some species in the mock sample, despite successfully amplifying them in individual PCRs (e.g., fish-2 kb and Atlantic cod).This suggests that at least some of the differences may be ascribed to amplification and primer biases during PCR reactions, which are known to impact species detection, especially when dealing with mixed genetic material (e.g., mock communities and eDNA samples) (Aylagas et al., 2016).
These biases could affect species richness estimations and lead to false negatives (Toxqui Rodríguez et al., 2023).Incomplete reference databases for the long region markers could be another reason for the fewer species identified.Other markers, such as cytochrome b, are promising, as they are typically included in population genetic studies and may have better coverage in the databases (Toxqui Rodríguez et al., 2023).To address such primer-related issues, the use of a combination of markers is often recommended (Clarke et al., 2017;Kelly et al., 2017;Wangensteen et al., 2018).Further investigation, ideally using mock communities (e.g., Baetscher et al., 2023;Holman et al., 2021), is needed to create a long region marker portfolio for metabarcoding studies, as demonstrated by Toxqui Rodríguez et al. (2023).Reliable long region markers would have several advantages over shorter markers, such as reducing false positives from latently degraded eDNA, increasing species discrimination power, and reducing the impact of sequencing errors for genetic species identification (Parker et al., 2017).

| Species detection in catch water
Overall, our eDNA approach showed a 95.32% detection success for landed species-covering fish, elasmobranchs, and Norway lobster.
We consistently detected all landed species, with minimal exceptions-Norway lobster in two out of three catch water samples from fishing haul 3 and haddock in fishing haul 11.In the case of Norway lobster, only one out of the three sample replicates detected the species in fishing haul 3 (Figure 5).This highlights the importance of using sample replicates to reduce the risk of false negatives and enhance elusive species detection (Beng & Corlett, 2020;Hestetun et al., 2021;Stauffer et al., 2021;Xing et al., 2022).Moreover, the difficulty in detecting Norway lobster may be because crustaceans release less DNA than other taxa (Komai et al., 2019) and because of the relatively small proportion of Norway lobster (15 kg) compared to other species in the catch, each exceeding 30 kg.Similarly, only a small proportion of haddock (2 kg) was landed in comparison to other species (37-102 kg) in haul 11.The limited abundance of haddock likely produced a false negative result, compounded by the low species identification resolution of haddock and whiting with the genetic markers used and Nanopore sequencing error rates (Wang et al., 2021).
By combining data on landed species with observations made on board, our eDNA assay successfully detected 81% of these species.
This detection rate is consistent with previous studies employing eDNA analysis to estimate the catch composition of trawling fisheries, reporting detection rates ranging from 70% to 90% of species recorded on board (Maiello et al., 2022(Maiello et al., , 2023;;Russo et al., 2021).
Our detection rate is also in line with previous studies that compared eDNA metabarcoding and trawling surveys, showing similar percentages of species detected (70-87%) (Afzali et al., 2021;Stoeckle et al., 2021).
It is important to note that the species that remained undetected through our eDNA assay were consistently observed on board in small numbers (1-4 individuals) during the initial 10 min of catch sorting, indicating their generally low abundance within the catch.
Consequently, their limited DNA contributions to the catch water reduced the probability of detection or resulted in minimal contributions to the sequenced reads.These factors may have led to their exclusion during filtering or prevented their assignment to species level due to their high similarity with other species (e.g., haddock and whiting).Moreover, inconsistent identification of rare species is a common feature of biological surveys (Furlan et al., 2016).Including more sample replicates (Stauffer et al., 2021), filtering a larger volume of water, increasing sequencing depth, or using qPCR to detect specific species (Xia et al., 2021) may help in detecting some of these elusive and less abundant taxa (Ohnesorge et al., 2023).
We found species exclusively detected through eDNA analysis but not reported as observed on board in the catch.The presence of | 11 of 16 eDNA-only species is commonly reported in eDNA studies, often resulting in a more comprehensive species list compared to traditional methods (Afzali et al., 2021;Gibson et al., 2023;Gold et al., 2021;Maiello et al., 2023).Similarly, Stoeckle et al. (2021) found that eDNA analysis retrieved a higher number of species per month compared to trawling.It is worth noting that, unlike the findings in Maiello et al. (2023), we did not observe a prevalence of pelagic species among our eDNA-only species (5%).This may be due to differences in eDNA sampling methods, as their sampling method also included the surrounding pelagic waters, which could have affected the observed species composition.Instead, our eDNA-only species mainly consisted of demersal, bathydemersal, and benthopelagic species (95%) commonly found as bycatch in bottom trawling.It is possible that some of these eDNA-only species were indeed present in the catch but were not reported, as our onboard observations were limited to the initial 10 min of catch sorting and focused mainly on species subject to the landing obligation.Moreover, we found supporting evidence for the presence of some eDNA-only species in images taken during catch sorting (i.e., Cyclopteruslumpus in fishing haul 2).However, species identification from pictures is imperfect, especially when distinguishing between closely related species.
There could be other sources of the DNA detected in the catch water.Some of the DNA may originate from species present in the water column, although this is expected to be relatively diluted compared to species present in the tank.For instance, seagulls are consistently present during fishing operations, especially during onboarding activities; however, we found 1 seagull eDNA read (Larus spp.) in three out of 18 samples.This indicates a minor contribution of eDNA from the water column compared to species in the catch.
Another potential confounding source of DNA could be from the stomach contents of fishes.When caught in the trawling net, fish may experience physical damage and stress associated with pressure changes.Additionally, there is a possibility that some DNA traces may originate from previous catches.However, we believe that the signal from species in the current catch would largely exceed any remnants from previous catches in terms of abundance.
Nevertheless, these potential problematic sources of DNA should be further investigated.To avoid problems associated with confounding DNA traces, environmental RNA could be a promising approach for species identification (Giroux et al., 2022;Veilleux et al., 2021).RNA degrades more rapidly and can provide a "real-time" picture of the tank's contents, thereby avoiding signals from other potential confounding sources of DNA, such as residuals from previous catches, eDNA in the water column, and DNA released from the stomach contents of damaged individuals within the catch.
Future studies are essential to explore the impacts of external potential sources of eDNA in catch water, potentially through visually monitoring the entire catch, analyzing seawater blanks (e.g., Russo et al., 2021), or assessing eDNA decay in artificial tanks (e.g., Holman et al., 2021).This will provide insights into the contribution of eDNA in the water column, previous catches, and the relative abundance of species present in the catch.The collection of seawater blanks has been discussed by Stoeckle et al. (2021), who reported uncertainties related to determining appropriate field blanks for eDNA studies conducted during trawling.
One feasible way forward could involve integrating eDNA analysis of catch water with existing methodologies for data collection on unwanted catches.Tools for monitoring and surveillance of unwanted catches were summarized by James et al. (2018) and included electronic monitoring, onboard observers, and self-sampling, where data are directly collected by fishers.While their review did not mention eDNA as a potential tool, over the past few years, there has been growing exploration of eDNA applications for data collection on unwanted catches and monitoring purposes (Aglieri et al., 2023;Albonetti et al., 2023;Maiello et al., 2022Maiello et al., , 2023;;Russo et al., 2021).
Collecting data on unwanted catches by combining eDNA with electronic monitoring programs (involving onboard videos) or atsea monitoring programs (onboard observers) could provide a holistic approach to address the limitations of each method when used in isolation.The primary limitation of electronic monitoring often lies in species identification for highly similar species and large-volume catches (Emery et al., 2018;Ewell et al., 2020).
In this regard, eDNA could offer higher species resolution, while electronic monitoring can provide information on the total species observed on board (including length information), allowing the development of a quantitative assay.Implementing eDNA in such cases would be relatively straightforward, involving the collection of a water sample during routine fishing activities.This integrated approach could significantly enhance the accuracy of species detection and monitoring efforts, essential for data collection on unwanted catches.

| Quantitative analysis
This study primarily aimed to assess the feasibility of using eDNA for species identification in demersal fisheries through catch water analysis.However, we also explored the potential for quantitative analyses and found significant positive correlations between eDNA read abundance and the weight of landed species.This suggests the potential for quantitative assessments using the catch water method.However, it is important to note that developing a robust quantitative assay that generates precise measures requires further research and calibration experiments (Hansen et al., 2020;Kirtane et al., 2021;Stoeckle et al., 2022;Urban et al., 2022;Yates et al., 2019).These steps are essential to account for various factors influencing eDNA shedding (Hansen et al., 2020) and persistence in the environment (Holman et al., 2021;Yates et al., 2021).The development of a quantitative eDNA assay tailored to the complexities of fisheries catches represents a promising area for future research.This holds true for both demersal (Afzali et al., 2021;Kirtane et al., 2021;Maiello et al., 2023;Russo et al., 2021;Stoeckle et al., 2021) and pelagic fisheries (Urban et al., 2022(Urban et al., , 2023(Urban et al., , 2024)), offering more precise insights into catch composition and enhancing the utility of eDNA analysis in fisheries management.

| Ecological insights from eDNA data
Our eDNA analysis revealed greater species richness in Nephrops fisheries samples, consistent with the generally high bycatch rates reported in Nephrops trawling fisheries (Avsar et al., 2023).eDNA is powerful in providing a more comprehensive overview of the species caught in these fisheries, including the presence of species that might be easily unnoticed using traditional methods (i.e., species not subject to the landing obligation) and species that cannot be easily identified.
Despite the relatively small sample size, our study demonstrates that eDNA analysis can effectively discriminate between catches from different fisheries, such as Nephrops and roundfish demersal fisheries.
Additionally, we detected differences in species diversity between deeper and shallower fishery samples, indicating a depth-related effect on catch composition (Afzali et al., 2021).These findings highlight the potential of eDNA analysis to uncover ecological patterns that can inform fisheries management strategies, in line with previous studies (Afzali et al., 2021;Aglieri et al., 2023;Maiello et al., 2023).
Furthermore, eDNA findings confirm an association of species with shallower and deeper water fishery samples, as expected by their biology.We observed eDNA detection of elasmobranch species that aligned with the known distributional range of these species.The eDNA analysis also enabled the detection of vulnerable and nearthreatened species and provided insights into their association with the target fisheries.Notably, E. spinax and C. monstrosa are categorized as vulnerable according to the IUCN Red List (Finucci, 2020;Finucci et al., 2021).Additionally, we detected several near-threatened species, including Hippoglossus hippoglossus, Rajaclavata, and Cyclopteruslumpus.This emphasizes the importance of monitoring their bycatch and implementing conservation measures for more sustainable fisheries.eDNA enhances our ability for taxonomic identification at the species level, which is especially important for taxa typically reported at higher taxonomic groupings, such as sharks and rays.This in turn contributes to higher data availability and resolution for conservation efforts and monitoring (Albonetti et al., 2023).With such information, best practices and adaptive fishing patterns can be promoted to reduce and prevent certain bycatch, ultimately minimizing fishingrelated mortality and safeguarding vulnerable and endangered populations.For instance, in pelagic longline fisheries, data obtained from electronic monitoring programs provided the basis to inform management strategies, such as modifying the time of day and fishing depth, aimed at diminishing the mortality of threatened seabirds as bycatch species (Gilman et al., 2023).Similarly, eDNA analysis can provide insights into the spatio-temporal patterns of bycatch species in demersal fisheries, promoting more effective evidence for bycatch avoidance and aiding in the monitoring of mitigation measures.

| Limitations and perspectives
While our study has demonstrated the potential of eDNA sequencing to estimate catch composition in demersal fisheries, there are several limitations and areas for future research.Unlike Illumina sequencing, where established bioinformatics pipelines exist for eDNA metabarcoding (e.g., Boyer et al., 2016;Xing et al., 2022), the same is not true for Nanopore reads.Recent studies have adopted various strategies, including clustering and aligning reads to construct consensus sequences (Doorenspleet et al., 2021) or direct assignment of the reads to a taxonomic database (Srivathsan et al., 2022).We expect advancements in Nanopore sequencing technology and more variable genetic markers (e.g., COI) to hold the potential to improve species identification and simplify data analysis.Addressing these limitations will contribute to the refinement and broader applicability of Nanopore eDNA sequencing in fisheries analysis.Long-read eDNA metabarcoding, made possible by Nanopore sequencing, could provide significant advantages when the objective is to detect species truly present in the sample (as in our case where species in the catch are directly releasing genetic materials into the tank) rather than detecting species in the environment (more fragmented eDNA).Targeted monitoring could benefit from a long-region marker portfolio for metabarcoding analysis.
Additionally, such efforts will enhance the accuracy and reliability of eDNA based assessments of biodiversity and ecosystem health, ultimately promoting fisheries management and conservation practices.

| CON CLUS IONS
We demonstrate the effectiveness of Nanopore eDNA sequencing for estimating species composition in demersal fisheries by analyzing catch water collected on board a fishing vessel.The high detection rates achieved and consistency in species composition highlight the robustness and versatility of this approach.
Challenges in detecting low-abundance species emphasize the need for sample replicates to enhance detection probabilities of relatively low-abundant species in the catch.Our research contributes to advancing the field of eDNA analysis in fisheries management, showcasing its potential for non-invasive, cost-effective, and comprehensive species monitoring.Our results show that eDNA analysis of catch water and Nanopore sequencing for estimating catch and bycatch composition in trawling fisheries is possible.The method offers highly cost-effective and practical opportunities to investigate spatiotemporal bycatch patterns and, importantly, can be readily integrated into self-sampling programs.
However, careful consideration of genetic markers and improvement of sequencing accuracy, as well as sampling replicates, are critical to improving the accuracy and reliability of species detection, particularly of rare species in the catch.eDNA analysis has the potential to provide reliable, high-resolution evidence for fisheries management, including decision-making, conservation actions, and the implementation of relevant legislation such as the EU landing obligation.

AUTH O R CO NTR I B UTI O N S
EEN, DB, PU, MWJ, and BKH contributed to the design of the study.
PU and BKH carried out fieldwork.SM contributed to the design of 26374943, 2024, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/edn3.555 by Danish Technical Knowledge, Wiley Online Library on [15/05/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

F
Principal Coordinate Analysis (PCoA) based on the Jaccard distances of dissimilarity using (a) species observed on board, including landed species, and (b) species detected in the eDNA (referring to the detection of all four markers pooled).For samples with replicates, a consensus was used.Samples are coloured based on mean depth, with Nephrops samples represented by circles and roundfish samples by triangles.F I G U R E 5Full overview of species observed on board, including landed species, and those detected by eDNA (all four markers pooled).The numbers of sequencing reads are depicted by colours (log-transformed).Numbers within the cells indicate the weight (kilograms) of landed species, and species reported as observed on board are indicated by asterisks (*).Landed species were always detected in the eDNA analyses, except Nephrops norvegicus in two out of three replicates in fishing haul 3, and Melanogrammus aeglefinus was not detected in haul 11.