Fish is the largest and most diverse class of vertebrates, with many species having important ecological and economic value (Nelson et al., 2016). However, climate change (Barbarossa et al., 2021), habitat loss (Light & Marchetti, 2007), overfishing (Yan et al., 2021), environmental pollution (Lima-Junior et al., 2006), and invasive species present great threats to global fish diversity (Leveque et al., 2008). Fish resource and diversity conservation are reliant on the accurate monitoring of fish communities to obtain a full and systematic understanding of species composition, abundance, and population structure (Rees et al., 2014). However, traditional capture-based fish surveys (e.g., netting, cage trapping, and electrofishing) not only require manpower and material resources but it is also difficult to identify closely related species and individuals in the early life stages (Lyal & Weitzman, 2004; Piggott et al., 2021).
Species monitoring based on environmental DNA (eDNA) is revolutionizing the way the biological composition of ecosystems is understood and has proven to be an efficient, cost-effective, and noninvasive method for biomonitoring (Pereira et al., 2021; Taberlet et al., 2012). Biomonitoring based on eDNA has achieved great success in biodiversity surveys at large scales (e.g., river basins, oceans), revealing the species richness at a fraction of the cost of traditional capture-based surveys. It is also capable of discovering hidden biodiversity and the presence of potential invasive species and has therefore been increasingly integrated into aquatic biomonitoring programs globally (Klymus et al., 2017; Valdivia-Carrillo et al., 2021; Zhang et al., 2022). For fish with the highest species richness, eDNA-based surveys will not disturb the ecosystem and provide not only comprehensive species diversity information but also an abundance of genetic evolution information for the organisms studied (Weitemier et al., 2021).
As an emerging technology, eDNA-based biomonitoring needs to be further verified in terms of its stability and reliability. The selection of target sequences (i.e., barcodes) and polymerase chain reaction (PCR) primers is critical. In targeted species-specific detection, the efficiency of specific primers amplification will directly affect the rate of false-negative/positive detection (Alberdi et al., 2018). In community surveys, universal primers should have specificity, coverage, and species discrimination to ensure the comprehensive and accurate species identification of environmental/mock samples (Hu et al., 2022; Shu et al., 2021). In addition, due to the read length limitation of high-throughput sequencing and the highly degrading nature of eDNA, small amplified fragments (typically <200 bp) may be more efficient for species detection (Bylemans et al., 2018; Freeland, 2017). Several primers have been developed for fish surveys, and each of them has successfully described local fish diversity. Previous studies based on in silico PCR and in vitro experiments also evaluated the ability of different primers to identify fish species and suggested that the most suitable primer for fish amplification is the 12S region (Zhang, Zhao, & Yao, 2020). However, there are huge biogeographic differences in fish communities, and it is difficult to find a pair of primers that can be applied to all areas or ecosystems (Guo et al., 2019; Rojas et al., 2019). In particular, universal primers will lead to a large number of non-fish amplifications, such as human sequences, resulting in serious data wastage and an increase in monitoring cost (McColl-Gausden et al., 2021; Sales et al., 2021). The amplification effectiveness and coverage of primers can vary greatly depending on the biodiversity, complexity, and taxonomic composition of the studied ecosystem. It is necessary to design PCR primers for the composition of fish in the specific ecosystem to accurately reveal the native biodiversity.
We designed new universal primers for fish monitoring by analyzing the whole mitochondrial genome of Chinese freshwater fish. The performance of the primers was compared using an in silico PCR, followed by an in vitro metabarcoding analysis using eDNA from waterbodies in the Yangtze River (Jiangsu section). The specific objectives were to (1) design primers based on the sequences of Chinese freshwater fish, (2) evaluate the identification performance of primers for the global fish community, and (3) evaluate the non-fish amplification of primers by an in vitro metabarcoding analysis.
MATERIALS AND METHODS In silicoTo construct a global vertebrate database (GVD) for the in silico PCR, we first downloaded all barcode sequences from the National Center for Biotechnology Information (NCBI) Genbank with the keywords 12S, 16S, and COI on April 15, 2022. The taxID of each sequence was obtained and taxonomic information was collected with the “taxize” package in R. Subsequently, sequences with lengths of <500 or longer than 20,000 bp were removed. Repeated sequences of the same species were also removed from the database. Finally, the database format was standardized using the OBITools package (Boyer et al., 2016). Fish sequences not distributed in China were removed from the GVD, and the sequences of Chinese native fish that we sequenced were inserted to create the Chinese Fish Database. The amplification coverage and taxonomic resolution of the primers were evaluated by the ecoPCR program (Ficetola et al., 2010) with the following settings: the amplification length range was 30–400 bp, less than three mismatches were maintained between each pair of primers and the target sequence, and there were no mismatches in the last two nucleotides at the 3′-end of the primers (Bellemain et al., 2010; Epp et al., 2012). All sequences that could be successfully amplified were retained to assess the taxonomic coverage and specificity of the primers. The vsearch --cluster_size was used to cluster all amplified sequences to operational taxonomic units (OTUs) with different similarities (Yang et al. 2017a). The OTUs were annotated based on the Basic Local Alignment Search Tool (BLAST) against the GenBank nucleotide database and GVD (Yang & Zhang, 2020). A species was considered unambiguously assigned if its sequence was more similar to a sequence in the reference database than to a certain threshold (from 0.99 to 0.95).
Water samplingTo assess the ability of the primers in vitro, we performed a metabarcoding analysis using water eDNA collected and extracted from the Yangtze River in Jiangsu, which is the largest river in China and the third largest in the world. Historical specimen records described a total of 161 fish species distributed in the Jiangsu area of the Yangtze River, belonging to 13 orders, 33 families, and 107 genera (Jia et al., 2022). Cypriniformes was the largest taxonomic group at the order level, with 72 species. Perciformes (51 species) and Siluriformes (13 species) were the next largest fish orders (Yang et al., 2020). We collected water samples from 22 sections of the Yangtze River in Jiangsu from November 2021. At each sampling site, 3 × 1 L water was collected from 5 to 10 cm below the surface using disposable 1 L plastic bottles. Water samples were filtered through a 0.45 μm mixed cellulose membrane (MCE) filter (47 mm diameter) on-site and then stored in individual sterile tubes at −20°C prior to DNA extraction. All equipment and utensils were cleaned with a 10% bleach solution prior to sample processing and between samples. At each site, 1 L double distilled water (ddH2O) was also filtered as a blank control.
The extraction ofThe eDNA was extracted using a water DNA kit (ENG, China), according to the manufacturer's instructions. Negative controls consisting of the extraction reagents were included for every set of extractions. The pellets in the final step were rehydrated with 100 μl ddH2O. The purity of the eDNA was confirmed by a NanoDrop spectrophotometer (Thermo Fisher Scientific) and quantified by a Qubit™ dsDNA HS Assay Kits (Invitrogen). A unique 12-nucleotide sequence tag was added to the 5′-ends of the forward primers in each PCR to enable the identification of different PCR products during the sequencing data analysis. The PCR was performed with a final volume of 25 μl, which was made up of 1 μl of 10 μM primers, 1 μl eDNA template, 19.9 μl ultrapure water, 2.5 μl 10 × PCR High Fidelity PCR buffer, 0.5 μl dNTP mix (10 mM), and 0.1 μl Taq DNA polymerase (Invitrogen). The PCR reaction procedures included an initial denaturation step at 95°C for 10 min; followed by 35 cycles of 95°C for 10 s, annealing temperature (Ta) for 20 s, and 72°C for 20 s, with a final elongation step at 72°C for 5 min. Three PCR replicates were performed for each sample to minimize any PCR bias. Negative controls without an eDNA template were included in all experiments. The PCR product was checked by 1.5% agarose gel electrophoresis and purified using an E-Z 96 Cycle Pure Kit (Omega). All purified PCR products were quantified using the 2100 bioanalyzer (Agilent Technologies), and the results were normalized and pooled together. Finally, the quantified amplicon library was diluted to a final concentration of 100 pM and sequenced using Ion S5™ (Thermo Life Technologies) according to the manufacturer's protocol.
BioinformaticsLow-quality reads with an average Q score below 20, and incorrect, ambiguous, or excessively short sequences were removed from the raw data following the QIIME pipeline (Yang et al. 2017a). FASTQ files were converted into the FASTA format and the corresponding quality files using MOTHUR software (Schloss et al., 2009). Sequences with no mismatches in tags and a maximum of two mismatches in primers were identified using the “split_libraries.py” script with “-w 50 -s 20 -M 2′′ and were retained for further analysis (Zhang et al. 2020b). Unique sequences were identified (dereplication) using the “derep_fulllength” command in vsearch. Sequencing errors and PCR chimeras were detected and removed by the “unoise3” function in usearch to produce amplicon sequence variants (ASVs) (Rognes et al., 2016). In silico PCR, operational taxonomic units (OTUs) were clustered with sequence similarity (cutoff 97%) following the UPARSE (USEARCH 8) pipeline. The OTUs/ASVs were taxonomically assigned using BLAST against the GenBank nucleotide database with the following criteria: (1) if the query sequence matched a single species with a max score and similarity >97%, the species was assigned; (2) if the query sequence matched a single species with the max score and similarity <97%, the genus of the species was assigned; and (3) if the query matched more than one species with the max score, a species was assigned based on its known distribution and the lowest taxonomic level (Zhang, Zhao, & Yao, 2020). Any species that did not occur in the Yangtze River was excluded to avoid the misidentification of taxa due to insufficient local species records. Amplicon sequence variants with the same taxonomic assignments were combined.
Statistical analysisOnly ASVs with relative abundances >0.01% were kept as high confidence taxa. The reads of ASVs that were simultaneously present in at least two replicates were averaged as the actually detected reads of each ASV at one site. A Pearson correlation was used to detect the same species with different primers. The difference in primer detection was measured by the consistency of the same species detected at different sites. To evaluate the effects of sequencing depth on the number of detected fish taxa, rarefaction curves were constructed for each primer set with increasing sequencing depth (i.e., the number of sequences reads) using the RARECURVE function in the r package vegan and iNEXT (Dixon, 2003; Hsieh et al., 2016). All data statistics and plots were based on the vegan and ggplot2 packages in R 4.1.2.
RESULTS Design of the new primers based on chinese freshwater fishSequence alignments of the 235 Chinese fish mitochondrial genome showed that the conserved sites were mainly distributed in the 12S–16S gene region (Figure 1a). A total of 17 fish specific primers were designed according to the conservatism of the sequence, with the amplicon length in the range of 100 to 300 bp (Table 1). Metafish11 had the shortest amplicons (~75 bp), while Metafish8 had the longest amplicons (~250 bp) (Figure 2a). Metafish3 could amplify the most Chinese fish sequences, followed by Metafish4, Metafish2, and Metafish1. Metafish1 could amplify the most Chinese fish species, followed by Metafish11, Metafish6, and Metafish7 (Figure 2b,C). Teleo2 and Metafish1 were both located in the 12S region. The last three bases of the 3′-end of the forward and reverse primers of Teleo2 are identical in vertebrate taxa (except amphibians). In contrast, the last three bases of the 3′-end of Metafish1 differ across vertebrate taxa (Figure 1b,C), especially for the forward primer.
Enlarge this image.FIGURE 1. Conservation site of the Chinese freshwater fish mitochondrial genomes. A: Red dots represent areas suitable for fish specific primer design. B: Vertebrate sequence differences in the forward primer of Metafish1 and Teleo2. C: Vertebrate sequence differences in the reverse primer of Metafish1 and Teleo2. N means the possibility of a different base. Bird, amphibian, and mammal sequences from the GVD.
TABLE 1 Summary of 17 fish metabarcoding primer sets designed based on Chinese fish mitochondrial genome and 5 common fish primers from the literature.
Enlarge this image.FIGURE 2. Assessment of the primers based on the Chinese freshwater fish. A: Length distribution of the amplicons of primers. B: Amplified fish sequence (left) and fish species (right). C: Clustering of operational taxonomic units (OTUs) with different similarity thresholds.
The GVD contained 17,406 taxa with 27,079 sequences, including 6834 fish (9004 sequences), 3734 amphibians (7395 sequences), 2240 birds (2683 sequences), 2582 mammals (4687 sequences), and 2016 reptiles (3310 sequences). The sequence lengths in the GVD were mainly concentrated in two regions of 1000 bp (12 S–16S genes) and 15,000 bp (mitochondrial genome) (Figure 3a,b). The COI and 16Sar primers had the largest amplicon length (>500 bp). The Teleo primers had the shortest amplicon length (~70 bp). The amplicon length of Mifish, Teleo02, and Metafish1 was ~170 bp (Figure 3c). Metafish1 had the highest amplification coverage, amplifying 64.91% of vertebrate sequences. Teleo2 and Teleo had the next highest amplification coverage, amplifying 55.2% and 52.45% of sequences, respectively. Only 1520 sequences could be amplified by COI, accounting for 0.56% of the total vertebrate sequences. All primers amplified both fish and non-fish taxa, but the proportion of fish taxa varied considerably from 34.4% (Metafish1) to 72.3% (MiFish) (Figure 3d,e). Except for the COI primer, fish accounted for the highest proportion of sequences amplified. Teleo, Teleo2, metafish1, and Mifish amplified 6523, 6122, 6160, and 5640 fish sequences, respectively. The Teleo primers amplified 4210 amphibians, and mammalian sequences and accounted for a large proportion of the amplicons of Metafish1 (15.1%) and Teleo2 (14.7%). Metafish1 amplicon clustered more OTUs (14,062) under 99% similarity, followed by Teleo2 (11,855) and Teleo (11,439). The order of successful species identification was Metafish1 >Teleo2 >Teleo >16Sar >MiFish >COI (Figure 3f). The primers that identified most species were Teleo02, followed by Metafish1 and Teleo. All primers identified 5986 fish species, among which 35% species could be amplified and identified by four primers (Metafish1, Teleo2, Teleo, and MiFish), and 88 species could only be identified by Metafish1 (Figure 3g).
Enlarge this image.FIGURE 3. Assessment of primers based on the global vertebrate database (GVD). A: Sequence length. B: Sequences and species composition. C: Amplicon length. D: Fish sequences that were successfully amplified. E: Fish species that were successfully identified. F: Clustering of operational taxonomic units (OTUs) with different similarity thresholds. G: Identification of fish species by different 12S primers.
A total of 572,251 high-quality sequences were obtained from Metafish1 sequencing on 66 Yangtze samples (Figure 4a), among which fish sequences accounted for 276,568 (48.32%) and mammal sequences accounted for 281,220 (49.14%) (Figure 4b). Teleo2 obtained 947,564 sequences in total, including 173,838 fish sequences (18.34%) and 728,447 mammal sequences (76.88%) (Figure 4c). Metafish1 sequences were clustered into 96 ASVs, including 79 ASVs in fish and 10 ASVs in mammals (Figure 4d). Teleo2 sequences were clustered into 98 ASVs, including 71 fish ASVs and 17 mammal ASVs. The fish detected by both the Metafish1 and Teleo2 primers were mainly Cypriniformes, accounting for 90% and 91% of the total fish sequences, respectively. A total of 66 fish species were detected by Metafish1 and 57 fish species were detected by Teleo2. A total of 48 fish species were detected by both primers. At the genus level, 57 genera were detected by Metafish1 and 50 by Teleo2, and 45 genera were detected by both primers (Figure 4e). At the same sequencing depth, the number of fish ASVs detected by Metafish1 was higher than the number detected by the Teleo2 primer. At a sequencing depth of 140,000, the Metafish1 primer was able to detect 90% of fish ASVs. The Teleo2 primer could detect 90% of fish ASVs when the sequencing depth reached 550,000 (Figure 4f).
Enlarge this image.FIGURE 4. Taxonomic distributions of sequences amplified using the Metafish1 and Teleo2 primers from Yangtze River samples. A: Sampling sites in the Yangtze River (Jiangsu section). B: Sequences and amplicon sequence variant (ASV) composition of the total sequences. C: Sequence composition of fish sequences. D: ASV composition of fish sequences. E: Differences in the species and genera detected by primers. F: Fish ASVs were detected at different sequencing depths.
A total of 446,708 fish sequences were obtained by high-throughput sequencing, of which 93.2% of sequences could be amplified by both Metafish1 and Teleo2, 26,097 sequences (5.8%) could only be amplified by Metafish1, and 4409 sequences (1.0%) could only be amplified by Teleo2 (Figure 5a). The sequence number per species amplified by Metafish1 was significantly positively correlated with that amplified by Teleo2. There were 32 fish species with more than 1000 sequences. A site-to-site comparison showed that the proportion of mammals amplified by Teleo2 was significantly higher than that amplified by Metafish1 (Figure 5b). The distribution of common fish (reads >1000) amplified by the two primers was basically consistent (Figure 5c). For both Metafish1 and Teleo2, the sequence numbers of Carassius auratus, Cyprinus carpio, Hemiculter leucisculus, and Hypophthalmichthys nobilis were high. At the same site, 60%–90% (mean 77%) of fish species could be detected by both Metafish1 and Teleo2 (Figure 5d). The relative abundance was positively correlated with the consistency of the species detected by Metafish1 and Teleo2 (Figure 5e).
Enlarge this image.FIGURE 5. Comparison of the Metafish1 and Teleo2 primers in fish detection. A: Correlation analysis of sequences per species between Metafish1 and Teleo2. The Venn diagram shows the shared sequences between Metafish1 and Teleo2. B: Comparison of mammal relative abundances detected at each site. C: Comparison of fish reads at each site. D: Consistency of fish detection. E: Correlation between the species relative abundance and consistency ratio.
Biomonitoring based on eDNA relies on PCR amplification to enrich and amplify the target species DNA (Yang et al. 2017a). Understanding the characteristics and bias of PCR primers is critical for the interpretation of high-throughput sequencing data. Although previous studies have evaluated a large number of fish primers and provided recommendations for the selection of primers for fish metabarcoding, there are large geographical variations in fish communities and little is known about the impact of primers on regional biodiversity surveys (Zhang, Zhao, & Yao, 2020). In addition, fish primers can amplify not only fish sequences but also a large number of human sequences, which leads to low monitoring efficiency and high costs (Zhang et al. 2020a). In the current study, a more efficient fish primer was designed based on Chinese freshwater fish, to reduce nonspecific amplification and improve the accuracy of regional fish community monitoring.
The nThe high taxonomic specificity of fish communities is one of the crucial criteria to be considered when selecting amplification primers. Inadequate specificity inevitably leads to the overamplification of nontarget species, which in turn causes a swamping of the fish species and wastage of the sequencing throughput. An evaluation of fish metabarcoding primers by Zhang et al. showed that 12S was an efficient marker gene, all seven 12S primers detected more than 45 fish taxa, and the top six primers with the largest number of recovered fish taxa were located in the 12S region (Zhang, Zhao, & Yao, 2020). However, due to the conservation of vertebrates in the 12S region, an in silico PCR showed that almost all vertebrates, including mammals, birds, reptiles, and amphibians, can be amplified by the 12S primer. Although non-fish amplification provides valuable information about the multitrophic levels of an ecosystem, it is undesirable in studies focusing only on fish (Port et al., 2016). Unfortunately, in the metabarcoding of actual environmental samples, only human DNA is amplified in large quantities, while there are few other non-fish vertebrate (e.g., amphibian and bird) sequences, which does not provide much valuable information. Surface waters inevitably receive various discharges that contain human DNA, and it is therefore common for environmental samples to amplify large amounts of human sequences (Barnes et al., 2014; Kelly et al., 2014; Miya et al., 2015). Although blocking the oligonucleotide modification of primers can reduce the nontarget amplification, it also inhibits the binding of modified primers to certain desired sequences (Hege & Simon, 2008; Valentini et al., 2016). It is therefore recommended to increase the sequencing throughput to compensate for nontarget amplification rather than adding blocking oligos (Zhang, Zhao, & Yao, 2020). The current study showed that minor changes to the primers were more effective than increasing the sequencing throughput. In particular, about 75% of the sequences of Teleo2 amplification products were from humans, and in some sites human DNA sequences accounted for more than 90% of all PCR products. Compensating for the detection of fish diversity by increasing sequencing depth is costly.
The 3′-end base mismatch of primers was the main reason for the decrease in nonspecific amplification. In PCR reactions, DNA polymerase is mainly bound to the 3′-end of the primer, and therefore the mismatch of the 3′-end of the primer will greatly reduce the amplification efficiency (Day et al., 1999; Schoenbrunner et al., 2017). The 3′-end of Teleo2, especially the last three bases, is completely consistent in vertebrates. Therefore, all vertebrate sequences will be indiscriminately amplified in the PCR process, resulting in a large number of non-fish sequences, especially human sequences, in the sequencing results (Rejali et al., 2018). Unlike in Teleo2, the 3′-end of the Metafish1 forward primers was significantly different from that of humans and birds. This will greatly reduce the amplification efficiency of human DNA and increase the proportion of fish sequences in metabarcoding monitoring.
Consistency in fish monitoringSmall changes in the primers significantly altered the qualitative and quantitative monitoring of rare species but had little effect on common species. Despite being the classic barcode gene, COI appeared to be a less common target of fish metabarcoding primers than the other mitochondrial genes (Hebert et al., 2003). Previous research has also shown that the 16S and COI primers generally recovered fewer fish taxa than the 12S primers (Zhang et al. 2020a; Zhang, Zhao, & Yao, 2020). Both Metafish1 and Teleo2 targeted the mitochondrial 12S rRNA gene, and the amplified fragments were highly overlapped. The monitoring results of the two primers for dominant species were basically consistent. At the same sequencing depth, Metafish1 found significantly more fish species than Teleo2. This was largely because most of the sequences amplified by Teleo2 were human sequences. More than 93% of the sequences (64% of the species) were detected by both Metafish1 and Teleo2, indicating that the differences between the two primers were mainly reflected in rare species with a very low abundance in the environment. On average, more than 75% of the fish distributions were consistent when only the species detected were considered. However, if the relative abundance was considered, the Rho value of the correlation between the two pairs of primers was only 0.25. Although the target fragments of the two primers coincided, there was a clear species preference in the actual amplification. In addition, the primer bias was amplified as the number of species in the environment decreased.
Comparison between in silicoThe results of in vitro metabarcoding may be different from those of in silico PCR. The in silico PCR analyses indicated that all classes of vertebrate were successfully amplified, including amphibians, birds, reptiles, and mammals. However, in the in vitro metabarcoding, there were very few vertebrates that were amplified in large quantities, except human sequences. There was less amplification of amphibians, birds, and reptiles mainly because the DNA of these organisms in the actual water environment was much lower than that of fish (Collins et al., 2019). Although Teleo2 identified more fish species than Metafish1 in the in silico PCR, the in vitro metabarcoding tests produced the opposite result, with Metafish1 amplifying and identifying more fish species. There were many reasons for the inconsistency between in silico and in vitro assessments, including differences in the taxonomic composition of the reference database versus the actual biomes of the ecosystem, differences in the simulation conditions and primer binding thermodynamics in real PCR, and the quality and quantity of eDNA templates (Zhang, Zhao, & Yao, 2020). The reliability of the sequences in the database is also one of the reasons for the difference in evaluation (Yang, Zhang, Zhang, et al., 2017). We also suggest the use of in silico PCR for the preliminary assessment of primer specificity and taxonomic coverage, and in vitro metabarcoding evaluation is essential for the final selection of appropriate primers. However, it should be noted that species composition and complexity can vary considerably across geographic areas and across ecosystems, and primer performance in one study may not be completely transferable to another.
CONCLUSIONSThe mitochondrial 12S region is the most suitable metabarcoding gene marker for freshwater fish in China and elsewhere. The minor change at the 3′-end of the primer can greatly reduce the amplification of non-fish sequences and improve the effectiveness of eDNA metabarcoding. Although in silico PCR can be used for the initial assessment of primer coverage and specificity, an in vitro PCR should be performed to determine the primer performance in an actual metabarcoding study. Because different primers may display a different taxonomic amplification bias, multiple primers can be combined to increase the taxonomic coverage and species detection probability. Biomonitoring based on eDNA depends on the integrity and sequence quality of the reference database; therefore, the construction of high-quality reference databases of local species should be the focus of future studies.
ACKNOWLEDGEMENTSWe thank the National Natural Science Foundation of China for their support (41807482 and U1901220). X.Z. was supported by the Fundamental Research Funds for the Central Universities. This work was also supported by the Jiangsu Funding Program for Excellent Postdoctoral Talent.
CONFLICT OF INTERESTThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
DATA AVAILABILITY STATEMENTHigh throughput sequencing FASTQ files underlying this article are deposited in NCBI Bioproject ID PRJNA913702.
© 2023. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Zhang, Lijuan 1 ; Mu, Yawen 2 ; Zhang, Xiaowei 3
1 State Key Laboratory of Pollution Control & Resource, School of the Environment, Nanjing University, Nanjing, China
2 State Key Laboratory of Pollution Control & Resource, School of the Environment, Nanjing University, Nanjing, China; Jiangsu Provincial Environmental Monitoring Center, Nanjing, China
3 State Key Laboratory of Pollution Control & Resource, School of the Environment, Nanjing University, Nanjing, China; Jiangsu Province Ecology and Environment Protection Key Laboratory of Chemical Safety and Health Risk, Nanjing, China