CHEN Jianwei, CHEN Zhi, LIU Shanshan, , GUO Wenjie, LI Di,MINAMOTO Toshifumi, and GAO Tianxiang

1) BGI-Qingdao, BGI-Shenzhen, Qingdao 266555, China

2) College of Fisheries and Life Science, Hainan Tropical Ocean University, Sanya 572022, China

3) China National GeneBank, BGI-Shenzhen, Shenzhen 518120, China

4) Qingdao Underwater World, Qingdao 266003, China

5) Graduate School of Human Development and Environment, Kobe University, Kobe 657-8501, Japan

6) Fisheries College, Zhejiang Ocean University, Zhoushan 316022, China

Abstract Environmental DNA (eDNA) metabarcoding has emerged as a potentially powerful tool to monitor invasive fish species.As an alternative (or complementary) tool for biodiversity monitoring, eDNA metabarcoding had been used to detect species in aquariums, which represents an important transit avenue for introducing non-indigenous species with high population densities. In this study, eDNA metabarcoding as well as morphological characterization were used to reveal the diversity of non-indigenous species in a large aquarium at Qingdao Underwater World. Environmental DNA metabarcoding of 14 water samples at five locations from the Big Water Tank detected 24 non-indigenous species and four putative non-indigenous operational taxonomic units (OTUs).In contrast, only 20 non-indigenous species were observed by morphological characterization. Some species undetected by morphological characterization, such as Oreochromis niloticus (Linnaeus, 1758), are highly adaptable to various environments and/or have invaded preferred regions where they threaten native aquatic species. eDNA metabarcoding also detected seven local fishes that were not identified by morphological characterization. However, analysis of OTU diversity among stations and sample replications revealed that eDNA varied within and/or between stations. Increasing sampling effort as well as negative controls are required to increase the detection rate of species and to eliminate false-positive OTUs.

Key words eDNA metabarcoding; fish diversity; non-indigenous; sampling effort; aquarium

1 Introduction

Invasive species are a growing source of ecological and economic harm worldwide (Pimentelet al., 2005; Mainaliet al., 2015; Gallardoet al., 2016; Crowleyet al., 2017).Non-indigenous organisms are transported worldwideviainternational and domestic trade at an ever-increasing rate,making the introduction of new species relatively easy (Havelet al., 2015; Macisaac, 2017). Invasive species in aquatic systems strongly impact native flora and fauna, human resources and economy and entire aquatic ecosystems(Havelet al., 2015; Gallardoet al., 2016; Simmonset al.,2016). The ornamental fish trade is the centerpiece of a rapidly growing aquarium industry. Thus, aquarium trade represents an important way of introducing non-indigenous aquatic species, and has been linked to hundreds and thousands of species invading natural ecosystems around the world because of eggs and larvae discharged into the environment along drainage systems (Mendozaet al., 2015;Perdikariset al., 2016; Vodovskyet al., 2017).

Early detection of invasive species is critically important for effective management. Once an invasive species has spread through aquarium channels, its management becomes more difficult and the costs of control or eradication increase exponentially (Miyaet al., 2015; Maistrelloet al., 2016; Epanchin-Niell, 2017; Trebitzet al., 2017).However, early eradication is difficultly because traditional survey methods (net, hook and line fishing, and electrofishing) tend to miss low density populations and cryptic taxa as well as eggs or larvae or other small stages of organisms/microorganisms. Therefore, monitoring techniques that allow for the early detection of invaders are needed to help protect native species and ecosystems. Environmental DNA (eDNA) is the DNA extracted from an environmental sample (e.g., soil, air, sediment, or water) without isolating a target organism (Ficetolaet al., 2008; Dejeanet al., 2011). Detecting eDNA is a new and rapidly growing monitoring tool for biodiversity research and management of organisms in ecosystems (Lodgeet al.,2012; Reeset al., 2014; Oldset al., 2016). It has been applied to detect and monitor invasive aquatic species frequently and conveniently (Jerdeet al., 2011; Dejeanet al.,2012; Goldberget al., 2013; Uchiiet al., 2016; Larsonet al., 2017). The monitoring scope of eDNA research in aquatic systems has been increasingly applied from fish and amphibians to more diverse taxa including mammals,gastropods and bivalves (Ficetolaet al., 2008; Footeet al.,2012; Thomsenet al., 2012; Eganet al., 2013; Goldberget al., 2013). While most eDNA detection results complement traditional approaches on aspects of sensitivity,cost, accuracy, and environmental friendliness, some limitations, or different detection results have been shown using eDNA for species monitoring. For example, the discharge of eDNA showed a contrasting transport tendency in larger and smaller systems (Deiner and Altermatt, 2014;Pilliodet al., 2014; Jerdeet al., 2016; Wilcoxet al., 2016).Therefore, this revolutionary approach still needs testing and improvements in various systems.

Previous biological studies on marine aquarium invasive species based on traditional methods have focused on some countries, such as Brazil, the Caribbean Islands, the European Union, Kenya, Southeast Asia, and the United States, but comparatively little is known regarding risk assessments in China, let alone using the eDNA method(Chan and Sadovy, 1998; Caladoet al., 2003; Gaspariniet al., 2005; Rhyneet al., 2012; Murray and Watson, 2014).Qingdao Underwater World locates in Shandong Province,adjacent to Yellow Sea. It is one of the oldest and largest aquariums in China. Import and export of aquarium fish occurs every year at Qingdao Underwater World. Although fish trade is under strict control by related departments to prevent the potential spread of non-indigenous organisms,some unseen species could escape regulation and become invasive species. Due to the exchange, purchase, or exhibition between a loose union consisting of about 28 large and dozens of small aquariums all over China, a species could break through multiple invasion-prevention measures in the natural environment and accelerate invasive spread. Purchases and exchanges of species within this union are very frequent. Qingdao Underwater World is just one link in this chain that could potentially transfer an invasive species into the north Yellow Sea. Invasion risk assessment is meaningful not only for species with considerable suitability and which could survive in a target area but also for the potential and indirect harm (Noonburg and Byers, 2005; Becket al., 2017). All non-indigenous species should be taken seriously, so many high density non-indigenous species living in Qingdao Underwater World can be transported to other aquariums all over China. At present, invasive species monitoring by the eDNA method has never been carried out in such an aquarium. In this study, eDNA metabarcoding was used to reveal the non-indigenous fish diversity at Qingdao Underwater World. The objectives of this study are: 1) to detect the presence of unseen aquatic species outside the range of monitoring that are not on the introduction list and 2) to evaluate the feasibility of this burgeoning tool in the ‘Big Water Tank’ or other aquatic system.

2 Materials and Methods

2.1 Species Identification with Morphological Characteristics

Between September 2017 and March 2018, a list of aquarium fish species and bait fish species in the Big Water Tank of Qingdao Underwater World were obtained from the caretakers who strictly distinguish and identify every individual fish and bait fish species with morphological characteristics when fish and bait fish were imported by or exported from Qingdao Underwater World. The Fishes(http://researcharchive.calacademy.org/research/ichthyology/catalog/fishcatmain.asp) and FishBase Catalog (http://www.FishBase.org/search.php) were used to assign scientific names to these fish species. The native fish species around north Yellow Sea including Qingdao were also summarized to exclude possible indigenous species and eDNA contamination caused by bait fish and exchange of water.

2.2 Target Tank

The Big Water Tank at Qingdao Underwater World was the target tank. The water volume in this tank is vast and appropriate. Species inhabiting this tank have visible body sizes and are identified with the highest reliability compared to those in other tanks. A long physical distance separated the Big Water Tank from the others, forming an ideal anti-eDNA pollution interspace. The water supply system for this tank was independent and continuously pumped fresh seawater into the tank from the outer slope (60 m offshore). The water was directed into the Big Water Tank after filtration. The false-positive eDNA metabarcoding was thus maintained at a very low possibility.

2.3 Water Sampling and DNA Extraction

Water sampling and filtration were undertaken on January 11, 2018. Before sampling, species in this tank were subjected an 8-day starvation to reduce bait pollution and purify water quality. In total, five sampling locations were set along the suspended passage (Fig.1). Samples were collected from the surface using a bucket and divided into three plastic cups, 2 L each site. To minimize cross-contamination, the buckets, filter funnels and measuring cups were bleached with 0.1% sodium hypochlorite for about 10 min before sampling and washed more than twice with surface water at each sampling location and were then placed in water for several minutes before sampling (Miyaet al., 2015; Minamotoet al., 2017; Yamamotoet al.,2017). All water samples were filtered through 0.45 μm pore size WCN filters (Whatman Cellulose Nitrate Membrane Filters, 7184-004). Filtration equipment (Combisart 3-branch stainless steel manifold, GM-0.33A pump) was used to accelerate the experiment. A total of 1500 mL water each sample was filtered and stored at ?80℃. Subsequently, three artificial seawater samples collected from Qingdao Underwater World and filtered were used as negative controls.

The filtered samples were subjected to DNA extraction using the DNeasy Blood and Tissue Kit (Qiagen, Hilden,Germany) according to the manufacturer’s instructions which were described previously (Miyaet al., 2015; Minamotoet al., 2017; Yamamotoet al., 2017). A Qubit Fluorometer and gel electrophoresis were used to determine DNA quantity.

Fig.1 Seawater sampling locations, five in total and all at surface. The passage hung in the air and did not affect water exchange. Size of the Big Water Tank was about 40 m(length) × 25 m (width) × 7 m (depth).

2.4 Library Construction and Sequencing

Metabarcoding was performed for each DNA extract with the universal PCR primers targeting the 12S rDNA region of mitochondrion (MiFish-U/E-F: GTCGGTAAA WCTCGTGCCAGC, MiFish-U/E-R: CATAGTGGGGTA TCTAATCCYAGTTTG) (Miyaet al., 2015). The following PCR procedure was used: 94℃ for 5 min, followed by 10 cycles of 94℃ for 30 s, 60℃ for 30 s, and 72℃ for 30 s, with a final extension at 72℃ for 10 min. After amplification, the primer with the BGISEQ-500 adapter was used for the next two amplifications of 20 cycles using the same PCR procedure, and the PCR products with target bands were obtained. After equal mixing of all PCR products, 2% agarose gel electrophoresis and gel cutting were performed to generate the single-stranded circular DNA library (Zouet al., 2020). Three negative control libraries were constructed simultaneously to evidence the effective amplification. All the amplified libraries were subsequently used as input to prepare the DNA Nano Ball and were sequenced in the paired-end model with 150 bp length reads on the BGISEQ-500 platform. The data are available in CNGB Nucleotide Sequence Archive(CNSA: https://db.cngb.org/cnsa; accession number CNP 0000391).

2.5 Bioinformatic Processing

The reads with adapters or at low-quality (＞20% base at Q-value ≤ 20 or ＞5% N bases) were removed using SOAPnuke (v1.5.6). The paired-end reads were combined into tags using FLASH (v1.2.11) and the primer base was removed. All tags were clustered to the OTU using USEARCH (v7.0.1090) with a 97% threshold, and chimeras were filtered using UCHIME (v4.2.40). The OTU abundance table was summarized with tags mapped to each OTU representative sequence. When the OTU was detected in the negative controls or the OTU cumulative absolute abundance in all samples was ＜ 100 tags, it was removed from the subsequent analysis.

The OTU sequences were aligned against the non-redundant nucleotide (nt) database (NCBI, https://www.ncbi.nlm.nih.gov/), and the Mitochondrial Genome Database of Fish (MitoFish, http://mitofish.aori.u-tokyo.ac.jp/) using BLAST+ (Version 2.2.26, best match according to BlastN,Evalue ≤ 1e?5). The OTU sequences were assigned to a species if there was ≥ 99% sequence identity across the entire length. After the OTU assignments, potential contamination was removed with the threshold correction method (Yamamotoet al., 2017). If a species assignment was not recorded, the taxonomic resolution was collapsed,and the indigenousor non-indigenous attributes ofthese OTUs were made putative.

Sequences for the phylogenetic tree construction were aligned using DNASTAR software (DNASTAR Inc., Madison, WI, USA). MEGA 5.0 was used to construct the neighbor-joining (NJ) tree under the Kimura 2-parameter(K2P) model. A Venn diagram was drawn with the ‘Venn-Diagram’ package in R (v3.1.1). A principal component analysis (PCA) was performed using the ‘a(chǎn)de4’ in R (v3.1.1)based on OTU abundance information. A beta (β-) diversity analysis was calculated using QIIME software (v1.80)and the Bray-Curtis distance was determined.

3 Results

3.1 List of Aquarium Fish Species in the Big Water Tank

After verification and exclusion, 25 fish species representing 15 families were present in the Big Water Tank(Fig.2, Table1). The number of elasmobranchs (13 species)was slightly more than that of bony fish (12 species). Most of these species were non-indigenous species (Fig.2), purchased from the South China Sea and south of the East China Sea, even from Southeast Asia (see the origin listed in Table 1). All 25 species showed a tendency to be warmwater species and mainly lived in the sea area south of the Taiwan Strait. Some species, such as the sand tiger sharkCarcharias taurus, bowmouth guitarfishRhina ancylostoma, spotted eagle rayAetobatus narinari, milkfishChanos chanos, and mottled spinefootSiganus fuscescenshave been marked as possibly native to the Qingdao Sea area by FishBase; it was difficult to find them in the north Yellow Sea. Nevertheless, they were temporarily classified as indigenous species (possible) in this study. There were 20 non-indigenous species in the Big Water Tank after combining the point and native distributions in Fig.2.

Fig.2 The point distribution of 22 species identified by morphological characterization in the Big Water Tank (referred from FishBase). Relative probabilities of occurrence are marked in four classes (four color). Maps of Dasyatis akajei(Müller and Henle, 1841), Dasyatis bennettii (Müller and Henle, 1841), and Lutjanus stellatus Akazaki, 1983 were unavailable. The number in this figure was same as that in Table 1.

3.2 High-Throughput Sequencing Statistics and Contamination Controls

We obtained an average of 102747 sequence tags per sample with a sequencing depth from 602 to 167755 and realized a high read utilization ratio (shown in Table 2).Sample Tank-5.3 with a shallow sequencing depth was removed. Thus, 14 samples were used for further accumulation curves (Fig.3), OTU annotation, and diversity analysis.

Table 1 List of fish species in the Big Water Tank based on morphological identification

Table 2 Statistics of the sequencing reads for 15 eDNA samples

Fig.3 Environmental DNA (eDNA) species accumulation curves for the five sampling sites at Qingdao Underwater World.

In total, 156 OTUs were detected, consisting of fishes,mammals, poultry, and bacteria. To remove contaminants,we removed the OTUs that occurred in the negative controls (most were indigenous fish species that might have come from bait fish, other OTUs belonging to mammals and bacteria), and OTUs that accumulated fewer than 100 tags. After filtering the contamination, 44 OTUs (all fishes) remained and were identified at different taxonomic levels (Table 3). Of the 44 OTUs, 35 were assigned to an accurate species level representing 36 species with high coverage (≥ 98%) and identity values (≥ 99%) (OTU-66 represented two species after revision of the visual census).The abundance rates of the 44 OTUs were highly variable. For instance, 96668 reads for OTU-6Gnathanodon speciosuswere recorded, whereas only 90 reads were observed for OTU-137.In general, there were far fewer

elasmobranch reads [90–14393 (2045 ± 3812)] than those of teleosts [257–183914 (16216 ± 38978)].

(continued)

3.3 Comparison of eDNA Metabarcoding and Traditional Identification

Out of 25 target species recorded by morphological characterization, 23 species were also detected by MiFish-U metabarcoding (Fig.4a). Considering thatGinglymostoma cirratummay have been falsely identified (see Discussion), MiFish-U metabarcoding detected 92% (=23/25) or 96% (=24/25) of the visually observed fishes. Furthermore,although eDNA metabarcoding did not detectTrachinotus baillonii(reference sequences: LC277971, LC146221 and LC458153) in the Big Water Tank, it has detected many other species. In total, 36 were fully determined by eDNA metabarcoding, which was 1.44 times higher (36/25) than that by the traditional (Fig.4a).

Fig.4 Species detection statistics using the traditional and eDNA metabarcoding methods. a) Species and undetermined operational taxonomic units (OTUs) detected by the traditional and eDNA metabarcoding methods. b) Summary information of indigenous or non-indigenous species and/or OTU detection by eDNA metabarcoding and a visual census.

Fig.5 Phylogenetic tree based on a neighbor-joining (NJ) analysis. Numbers above the branches indicate NJ bootstrap percentages. Only bootstrap values ＞ 80% are shown on the NJ tree.

In addition to 23 target species in the Big Water Tank,13 non-target species from other sources including other tanks at Qingdao Underwater World or possible bait were detected by metabarcoding. After verification and exclusion using a local fish list, six species were determined to be non-indigenous species (Fig.4b). Therefore, the total number of non-indigenous OTUs assigned to species level was 23, representing 24 species.

In addition to 35 OTUs assigned to 36 accurate species,nine OTUs, which were ambiguous at the species level,were also found. The nature of these nine undetermined OTUs detected by eDNA metabarcoding is debatable due to their imprecise taxonomic names. There are multiple possibilities for their origins, particularly for teleost because of the enormous number of possible groups. However, the OTUs of elasmobranches are deducible. After exclusion, the Big Water Tank was determined as the origin of four elasmobranch OTUs (OTU-14, -61, -91, and-137), which agreed with the observation that most of the species in the Big Water Tank were elasmobranchs. A phylogenetic analysis suggested that the four elasmobranch OTUs differed from all of those confirmed in the tank,but they mingled with the ‘ray, skates, and sawfish’ group(Fig.5). The genetic distance between OTU-14, -61, -91,and -137 and OTU-18, -35, -36, and -66 ranged from 0.0354 to 0.2156 (0.1146 ± 0.0474). Short branches as well as moderate genetic distances among these eight OTUs implied a possible similarity in morphology. Individuals of the former four OTUs might be hidden in the latter four groups and were introduced into Qingdao Underwater World.It speculated that the OTU-14, -61, -91, and -137 were nonindigenous organisms which might have the same origin of the fishes in the Big Water Tank (non-indigenous OTUs)(Fig.4b).

3.4 OTU Diversity Among Stations and Sample Replicates

The number of fish OTUs varied among the sampling stations. The highest number of OTUs was detected in Tank-1, Tank-3, and Tank-4 (44 OTUs, representing 45 species) (Fig.6a). A high number of species was also detected in Tank-2 (43 OTUs). OTUs varied only slightly among these four stations. The minimum number of fish species was detected in Tank-5 (34 OTUs), which was obviously lower than those of the other four sampling stations. The mean number of OTUs detected by each sample amplification was 39.3, which increased to 42.55 and 43.75 for two and three sample replications, respectively(Fig.6b). Increasing replicates result in detecting more OTUs.

Fig.6 Relationship between the number of detected species and sampling locations (a) and number of PCR replications (b).

Similarity among the samples was examined using a principal component analysis (PCA) (Fig.7a). Tank-5 was the most isolated and shallow location; thus, it had a distinct OTU assemblage from the other locations. It was evident from the library sequences that samples from this site were always separated from others in total reads and number of OTUs (Fig.1a and Fig.6). In contrast, locations 1 (Tank-1) and 2 (Tank-2) displayed similar biodiversity,followed by location 4 (Tank-4). The Bray-Curtis-based heatmap (Fig.7b) revealed a similar pattern of primary clustering of the community structure. Variation was detected not only between locations, but also within groups. Based on theβ-diversity measures among all replicates in Fig.7,replicates from the same location were occasionally more different than replicates between locations. The straightline distance between Tank-3 and Tank-5 ranged from 4.71 to 11.29 (6.95 ± 3.04), while the distance between Tank-1 and Tank-2 was 0.36–5.11 (3.64 ± 1.59). The latter was much smaller than the former (one-way ANOVA:F= 11.81,P=0.0056). As in the PCA analysis, the Bray-Curtis-based heatmap cluster results also revealed that Tank-4.3 and Tank-2.3 were more similar with each other than to the respective replications. Difference within groups might be larger than those between groups.

Fig.7 Ordination diagrams of the PCA analysis and heatmap based on the Bray-Curtis distance of the eDNA samples.

4 Discussion

eDNA metabarcoding applied to Qingdao Underwater World revealed a higher species diversity than that of morphological characterization. The non-visual census OTUs such as OTU-14, -61, -91, and -137 might be non-indigenous organisms and might be at risk of species invasion.Even if the population density of some non-indigenous fishes only detected by eDNA is relatively low, their survival within northern China is likely (Elvira and Almodovar, 2001; Ellis, 2006). For example, eDNA results detected the existence ofOreochromis niloticus, an Africa fish that has strong adaptability to various environments,including euryhaline, low dissolved oxygen, and turbid water (Petersonet al., 2005; Zambranoet al., 2006; Weyl,2008). With the increase of temperature in summer, this species may be appropriate for short-term survival in natural habitats. In total, eDNA metabarcoding identified 24 traded aquarium fish species and at least four OTUs belonging to non-indigenous organisms, which was 1.2 or 1.4 times higher than that detected by the visual census(20 species). A major advantage of a morphological characterization is that when a species is captured at a given site at a given time, its presence is unequivocal (Arduraet al., 2015; DiBattistaet al., 2017). However, the disadvantage is that a morphological characterization requires morphological identification of the species with precision in the part of the aquarium world where fish diversity is still not well described and there is a high level of endemism (Steinkeet al., 2009; Collinset al., 2012; Bamaniyaet al., 2016). Despite the preciseness of taxonomic keys, assigning individuals with certainty to a given species can be arduous when they are juveniles or larvae or when different species share very close traits (Collinset al.,2012; Yamamotoet al., 2017; Fernandezet al., 2018).Dasyatis akajeiandDasyatis bennettiwere mis-identified by our staff before the study period. The Pacific seaweed pipefish,Syngnathus schlegeli(OTU-37) is a small fish with a quiet character and a body like seagrass, which makes it easy to be concealed and be overlooked by a morphological characterization (Chenet al., 2017). The eventuality of cryptic species present in the Big Water Tank cannot be ruled out. Cryptic phenomena are widespread among metazoan taxa and biogeographical regions and, as a consequence, could potentially reduce species richness when morphological characterization/morphological criteria are used alone (Decruet al., 2016; Hinloet al., 2017; Statet al.,2019). As a consequence, errors in a morphological characterization are likely to be difficult to evaluate precisely.

eDNA metabarcoding is an effective way to reveal species that are indistinguishable morphologically and/or disdetected or leak detected. The preliminary results we obtained for ‘unseen’ individuals/taxa at Qingdao Underwater World have led us to reach similar conclusions to recent eDNA metabarcoding studies deemed to overmatch traditional methods (Yamamotoet al., 2017; Cilleroset al.,2019; Jerdeet al., 2019; Statet al., 2019). Although some species, such as OTU-29Decapterus maruadsiand OTU-42Epinephelus coeruleopunctatus, have been recorded in other tanks, which could represent eDNA pollution and/or tank-to-tank water exchange, we were unable to detect other fish species living with these two species. The first possibility is that the two species hid in the Big Water Tank. This conclusion is more credible for OTU-8, -28,-34, and -50, which have never been recorded at Qingdao Underwater World. OTU-34Siganus guttatuswas closely related to OTU-109Siganus fuscescensrecorded in the Big Water Tank, but it may have been mis-identified as the latter. A similar inference was applicable to the combination of OTU-50Nebrius ferrugineusandGinglymostoma cirratumandOTU-8Trachurus japonicasand OTU-29Decapterus maruadsi.The status of the nurse sharkGinglymostoma cirratumidentified by morphological characterization is controversial. While this species together with many other aquatic organisms (not in the Big Water Tank) were purchased from the US or South Africa, it was not detected by eDNA metabarcoding. This species might be a mis-identification of OTU-50, as both are from the family Ginglymostomatidae and this family only has four species (FishBase, https://www.FishBase.de/Summary/Family Summary.php?ID=493).

Unclassified taxa (OTUs) also reflected the ‘unseen’risk of species invasion. Because the sequence variation of close species is too small to distinguish (for instance,OTU-15, -120, and -160) or the species is unknown and the taxonomic assignment could be at a higher level (for instance, OTU-61 and -91) (Yamamotoet al., 2017; Gilletet al., 2018), even if the sequence has a high BLAST identity to the references, it still may have no clear information at all taxonomic levels. After contaminating controls and the species cut-off, majority of unclassified taxa (OTUs)in this study belonged to the second case. However, species surviving in the Big Water Tank are mainly sharks, rays,skates, and sawfish, which allowed us to make the conclusion that OTU-14, -61, -91, and -137 could not have come from another tank or local bait and are non-indigenous OTUs. There are very few elasmobranchs in the north Yellow Sea in past the decade due to overfishing, and the possibility that these OTUs wereSphyrna lewini,Platyrhina tangi, orOkamejei kenojeiwas excluded. We have never seen any other elasmobranchs in fish markets. All elasmobranchs at Qingdao Underwater World have been recorded by staff, including OTU-50Nebrius ferrugineus, OTU-91Rhinopterasp. orDasyatissp. 1, OTU-137Himanturasp., OTU-61Dasyatissp. 2, orPteroplatytrygonsp. OTU-14Dasyatissp. KAUM and their possible candidates did come from other tanks or as local bait because of their conspicuous appearance and easily discoverable body sizes.These unseen OTUs escaped regulation and may be putative invaders.

Although a more efficient method is needed to detect species in an aquarium system than a morphological characterization, eDNA metabarcoding studies are prone to false-positive and false-negative errors (Taberletet al., 2012;Gilletet al., 2018). Caution during sampling and molecular experiments needs to be undertaken and various controls performed to monitor contamination (Jerdeet al., 2011;Goldberget al., 2013). To strengthen our results, the survey sampling strategy was designed to have numerous replicates within a small area (sampling at five different locations and with triplicates at each site). The results between locations showed some discrepancy caused by contingent and unknown factors. eDNA composition varied in the small aquatic system (samples from Tank-5 displayed well). Detection performance in this study was also affected by PCR replication number, which is similar with other eDNA studies (Yamamotoet al., 2017). The level of sampling effort required is important to increase the detection rate of a species. The choice to sample less frequently may have caused false-negative errors. However,contaminants and false-positives will be inevitable with increasing sampling effort. Yamamotoet al. (2017) employed a 1.5% cut-off value by setting a negative control percentage of 9.62% (negative control number: 30; total eDNA sample number: 282). In this study, the percentage was 17.65% (negative control number: 3; total eDNA sample number: 14) and the consequential cut-off value was 1% (represents for 100 sequenced tags). Thus, more than 100 OTUs were successfully removed. Negative control in space, in time, and in the experiment was necessary to counteract the false-positive results and to increase the OTU cut-off value.

5 Conclusions

Based on the eDNA approach and the morphological characterization method, this study revealed the high potential for species invasion in one of the oldest and largest aquariums in China. eDNA metabarcoding could be useful to anticipate the risk assessment and the species monitoring needed for future prevention of non-indigenous organisms that would affect unknown ecosystems in preferred regions. Particularly, the survey highlights the necessity to adopt eDNA metabarcoding when performing invasion monitoring in aquariums. The use of eDNA metabarcoding allowed a more sensitive and efficient biodiversity monitoring method than a morphological characterization, although some limitations (false-positives and false-negatives) were encountered.

Acknowledgements

This study was supported by the National Key R&D Program of China (Nos. 2018YFD0900301, 2019YFD09 01301) and the National Natural Science Foundation of China (No. 41776171).