http://orcid.org/0000-0002-8280-4795
Figures
Abstract
The difficulty of censusing marine animal populations hampers effective ocean management. Analyzing water for DNA traces shed by organisms may aid assessment. Here we tested aquatic environmental DNA (eDNA) as an indicator of fish presence in the lower Hudson River estuary. A checklist of local marine fish and their relative abundance was prepared by compiling 12 traditional surveys conducted between 1988–2015. To improve eDNA identification success, 31 specimens representing 18 marine fish species were sequenced for two mitochondrial gene regions, boosting coverage of the 12S eDNA target sequence to 80% of local taxa. We collected 76 one-liter shoreline surface water samples at two contrasting estuary locations over six months beginning in January 2016. eDNA was amplified with vertebrate-specific 12S primers. Bioinformatic analysis of amplified DNA, using a reference library of GenBank and our newly generated 12S sequences, detected most (81%) locally abundant or common species and relatively few (23%) uncommon taxa, and corresponded to seasonal presence and habitat preference as determined by traditional surveys. Approximately 2% of fish reads were commonly consumed species that are rare or absent in local waters, consistent with wastewater input. Freshwater species were rarely detected despite Hudson River inflow. These results support further exploration and suggest eDNA will facilitate fine-scale geographic and temporal mapping of marine fish populations at relatively low cost.
Citation: Stoeckle MY, Soboleva L, Charlop-Powers Z (2017) Aquatic environmental DNA detects seasonal fish abundance and habitat preference in an urban estuary. PLoS ONE 12(4): e0175186. https://doi.org/10.1371/journal.pone.0175186
Editor: Hideyuki Doi, University of Hyogo, JAPAN
Received: January 13, 2017; Accepted: March 21, 2017; Published: April 12, 2017
Copyright: © 2017 Stoeckle et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: New 12S and COI fish sequences are deposited in GenBank with following accession nos. (12S, KX686069-KX686099; COI, KX688296-KX688324). Original fastq files with metadata are deposited in NCBI Sequence Read Archive (NCBI BioProject ID PRJNA358446).
Funding: This work was supported by The Rockefeller University-Monmouth University Marine Science Policy Initiative (MYS) and the National Institutes of Health Grant AI110029 (ZCP). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Effective ocean management depends on knowledge of the diversity, distribution, and abundance of marine life. Because censusing marine life requires costly specialized equipment, skilled personnel, and time, sampling is rarely dense or frequent. As compared to surveying sessile species such as shellfish, monitoring fish and other nekton is particularly challenging because they move—in response to daylight, temperature, and season; to evade capture or predation; and in relation to other short and long-term factors.
Analyzing water for DNA traces shed by organisms—environmental DNA (eDNA)—may help people to learn affordably, quickly, and frequently about the presence and abundance of known forms of marine life, especially fish [1–3]. eDNA sensitivity, accuracy, distribution, and duration have been carefully studied in a number of controlled and natural freshwater settings [4–11]. The utility of eDNA in marine environments, vastly larger and more complex, is just beginning to be explored [12–16].
Here we tested the sensitivity and specificity of aquatic eDNA for fish detection in the lower Hudson River estuary surrounding New York City, the most heavily urbanized estuary in North America (Fig 1) [17]. This complex ecosystem receives daily freshwater inflow from the Hudson River and ocean tidal inflows from Long Island Sound and New York Bight. Although water quality has improved dramatically over the past few decades, fecal contamination from wastewater remains ubiquitous [18,19]. The estuary is essential habitat for anadromous fish that arrive from the ocean in the spring to breed, including American shad, blueback herring, striped bass, and the threatened Atlantic sturgeon, and the catadromous American eel, that returns to the Hudson after breeding in the mid-Atlantic [20].
Major waterways are labeled, and New York City limits highlighted in dark gray with boroughs indicated [Bronx, (BX), Manhattan (MN), Queens (QU), Brooklyn (BK), Staten Island (SI)]. Sampling sites are marked by dots. Figure prepared using U.S.G.S. topographic maps as templates.
The lower Hudson River estuary offers eDNA assessment advantages. First, multiple seining surveys document local fish populations. Second, large seasonal changes in fish abundance test eDNA temporal specificity. Third, daily freshwater and saltwater inflows, which might carry eDNA from non-resident species, challenge geographic localization of eDNA. Contrasting environments within the estuary query eDNA distribution at a finer scale. Finally, most resident fish have mitochondrial sequences in GenBank, enabling identification of amplified DNA sequences.
Aquatic eDNA is often applied to detect rare or well-hidden taxa not easily found by traditional methods, using species-specific primers or exhaustive amplification [21–25]. Here we aimed to assess all resident fish species—or at least all of the more common ones—utilizing a metabarcoding protocol. We analyzed samples collected at two contrasting estuary locations over six months. The overall hypothesis was that eDNA is an indicator of fish presence and abundance. More specifically, we hypothesized that local marine fish would be detected and freshwater and open ocean species would not; that detection would reflect local abundance as determined by a compiled checklist; that eDNA detections would differ by season consistent with the springtime movement of fish populations into regional inshore waters and estuaries; and that there would be differences among the sites related to what is known about fish habitat preferences. Our findings are largely consistent with these hypotheses. eDNA detected abundant and common estuary species more often than uncommon ones, rarely found freshwater species despite Hudson River inflow, differed by season consistent with the springtime movement of fish populations into regional inshore waters and estuaries, and differed by site consistent with fish habitat preferences. In addition, we found eDNAs of commonly consumed fish, which may reflect wastewater contamination in the estuary. We discuss limitations and aspects needing further study, and potential of this technology for marine fish assessment.
Results
A checklist of 85 fish species was compiled from 12 local surveys [26–28], with species categorized according to the number of surveys in which they were present: abundant (9–12 surveys; 14 species), common (5–8 surveys; 28 species), or uncommon (1–4 surveys; 43 species) (S1 Table). We collected 76 one-liter shoreline surface water samples over six months beginning in January 2016. The two collection locations were a marine (high flow, near-ocean salinity, low turbidity) site on the East River and a more typical estuarine (low flow, brackish, high turbidity) site on the Hudson River (Figs 1 and 2). eDNA from water samples was amplified with vertebrate-specific 12S mtDNA primers [29] and sequenced on an Illumina MiSeq. Bioinformatic analysis of MiSeq fastq files detected eDNA from 51 fish species, most of which (82%) matched local marine taxa (S2 and S3 Tables). Consistent with eDNA detectability as reflecting local abundance, the assay identified most of the abundant or common checklist species and relatively few of the uncommon ones (81% vs. 23%, p = 2 x 10−6, Fisher’s exact test) (S2 Table). In addition, abundant or common species were found in more samples than were uncommon fish (average detections per species, 10 vs. 1, p<0.01, Mann-Whitney U test). Overall, nearly all reads (93%) matched locally abundant or common fish species. At both sites there was a strong seasonal increase in fish eDNA detection, consistent with the taxonomically widespread movement of fish populations into regional inshore waters and estuaries in the spring (Fig 3) [30–32]. Species seasonally detected by eDNA known to exhibit regional springtime population increases include Atlantic menhaden (Brevoortia tyrannus), Atlantic silverside (Menidia menidia), river herrings [alewife (Alosa pseudoharengus), American shad (A. sapidissima), blueback herring (A. aestivalis)], bay anchovy (Anchoa mitchilli), black sea bass (Centropristis striata), bluefish (Pomatomus saltatrix), cunner (Tautogolabrus adspersus), striped bass (Morone saxatilis), scup (Stenotomus chrysops), tautog (Tautoga onitis), and weakfish (Cynoscion regalis).
Collection equipment (bucket, rope) visible in upper right panel.
Top: black indicates eDNA presence; each column represents an individual sample, arranged by collection date, with collection site and month shown; each row represents a unique amplified fish sequence with identification at left, arranged by decreasing frequency of detection and number of reads. (M) indicates matching multiple local species; incomplete identifications are shown with genus or family names. Collection dates, scientific names, and identification details are in S3 and S4 Tables. Bottom: number of fish species detected per sample.
This study employed a single amplification protocol. In pilot experiments, we found reproducibility was correlated with number of reads—more reads, more likely to be detected on repeat amplification (overall reproducibility 64%; for reads <1000, 1000 to 10,000, and >10,000, 41%, 63%, and 87%, respectively, all comparisons, p<0.001, Fisher’s exact test) (S5 Table). 12s eDNAs matching species with declining or threatened populations were observed, including American eel (Anguilla rostrata), Atlantic sturgeon (Acipenser oxyrhinchus), and winter flounder (Pseudopleuronectes americanus) (Fig 3). Local time series seining data was available for a few species; in these cases the timing of eDNA seasonal increases (by presence/absence and read number) paralleled increasing fish numbers in the estuary (Fig 4). Most species were more common at East River site or were similarly distributed; the two species more often detected at Hudson River site [naked goby (Gobiosoma bosc), hogchoker (Trinectes maculatus)] are estuary specialists (Fig 3) [30,33]. An interesting comparison is naked goby, largely detected in Hudson, and its congener seaboard goby (Gobiosoma ginsburgi), which was in East River samples only.
Scales of eDNA reads and no. individuals differ between species. Number of individuals from 2013 Long River Survey trawls conducted between March 13 and July 27 at the southern tip of Manhattan, approximately 10 km by water from both study sites [28]. For bay anchovy and hogchoker, graph represents number of yearling and older individuals; for striped bass, graph represents number of eggs collected in the entire estuary as a proxy for migratory adult individuals, as the survey seining equipment does not capture mature striped bass.
One unexpected result was amplification of eDNA matching locally rare or absent fish species, comprising nine (18%) of the 51 unique fish sequences (S2 Table). One or more exotics were detected in 7% of samples; in total they comprised 2% of fish reads. About half were menu or aquarium species including Atlantic salmon (Salmo salar), European sea bass or “branzino” (Dicentrarchus labrax), Pacific red snapper (Lutjanus peru), and common guppy (Poecilia reticulata). One sample amplified a sequence matching (100% identity) Pacific sand lance (Ammodytes hexapterus), and differing (98% identity) from the resident congener Atlantic sand lance (A. americanus). The remaining exotics were freshwater species resident in the Hudson River watershed—common carp (Cyprinus carpio), darter (Ethiostoma sp), channel catfish (Ictalurus punctatus), and white sucker (Catostomus commersonii).
Discussion
Here we show that the detection of fish eDNA correlates with fish abundance as determined by a compiled checklist and differs by season consistent with the widespread movement of fish populations into regional inshore waters and estuaries in the spring. To our knowledge, this is the most extensive time series analysis of marine fish eDNA to date. Together with other studies, these results support eDNA for marine fish assessment and highlight limitations and aspects needing further study.
This study focused primarily on detection, i.e., presence or absence of eDNA. There was not sufficient data on either side of the potential correlation to assess eDNA reads as a general measure of abundance. First, published seining studies to date have not analyzed seasonal fish populations at the temporal and geographic scale of this manuscript. Most are not time series; most do not report number of individuals collected; none have been conducted in the East River. The one time series that reports on abundance of multiple species, the Long River Survey, uses equipment designed to capture ichthyoplankton—eggs and juvenile forms—and not adult fish. We were able to use this data for comparison to eDNA reads in a few species, either because adults are small and so are captured in survey nets (bay anchovy, hogchoker), or because the species’ natural history is well enough known to use eggs as a proxy for migratory adult individuals (striped bass). Second, the single amplification protocol employed in this study is insufficient to quantify differences between species.
Our single amplification protocol reproducibility was dependent on apparent eDNA abundance, as reported in other metabarcoding studies [34]. Overall reproducibility was relatively low, 64%, suggesting that single amplification was likely to miss about one-third of the species in a given sample. However, the composite sensitivity summing all samples should be high. This suggests additional samples or amplifications are unlikely to yield additional taxa at these particular estuary sites, unless there are late seasonal migrations not covered by this time series. Failure to detect was not due to our OTU screening procedure, as no fish species were entirely eliminated from a given MiSeq run by the thresholds applied. Nonetheless, we detected few of the uncommon estuary fishes. This may reflect the highly skewed distribution of fish abundances in the estuary. For example, in the 2013 Long River Survey, 10 of 61 estuary species accounted for 99% of individuals [28]. For scarce taxa, species-specific primers can be more sensitive than metabarcoding [35]. Besides rarity or absence of fish, failure to detect may be due to localization of eDNA within the estuary, primer mismatch, reduced shedding or increased turnover of eDNA, or other unknown factors [36,37].
In about 10% of samples we found DNAs matching fish that are commonly consumed but are locally rare or absent—Atlantic salmon (Salmo salar), Pacific red snapper (Lutjanus peru), and European sea bass (Dicentrarchus labrax). Commonly consumed Salmonidae are reported as an apparent contaminant in some marine eDNA studies [14,16]. We hypothesize that the menu species DNAs in this study originate from processed or raw sewage in local waters. Directly analyzing these sources will help evaluate this conjecture. If confirmed, this could limit study of commonly consumed species in environments with wastewater contamination. Other potential sources of apparently exogenous DNAs include laboratory procedures, commercial reagents, sequencing error, and undocumented haplotypes of local species [38,39]. In addition to fish eDNAs, human, domestic animal, and terrestrial wildlife eDNAs were commonly obtained (S6 Table), as routinely observed in aquatic eDNA studies.
in silico analysis confirmed highly conserved primer binding sites among local vertebrates for the 12S eDNA target, and new sequences reported in this study boosted target coverage to 80% of local fish. As expected, the short amplicon sequence matched multiple local species in some cases, limiting species resolution. Alternate targets may enable discrimination [39].
Contamination did not appear to be an issue (see METHODS for details), suggesting fish eDNA work can be safely performed in a benchtop setting with standard molecular biology methods, at least when assaying relatively common species. More advanced precautions, such as those used in ancient DNA investigations, may be needed in some circumstances, particularly if the aim is to detect very rare DNAs [16,34,40]. The workflow was relatively slow, on average about three months between collecting samples and obtaining sequence results, primarily because the fixed cost of a MiSeq run mandated accumulating multiple samples before submission. Sample collection aside, the entire process could be accomplished without stress with present technology in one week, and in about 24 hours if urgency demanded. Total direct costs excluding salaries of personnel were low, about $10,000.
This study demonstrates amplified aquatic eDNA correlates with fish abundance, seasonal movements, and habitat preference in an urban estuary with large fresh and saltwater inflows. Taking into account inherent limitations and multiple aspects needing further study, we and others anticipate that the relatively low cost of sampling, which can be performed by diverse persons with modest equipment, will facilitate surveys at much finer temporal and geographic scales than possible with traditional techniques [2,41–44]. Early applications in regional waters could include assessments of aquaculture on proximate marine ecosystems and when closures of dredging or fishing are needed to protect threatened migratory species. In environments lacking a comprehensive reference sequence library, e.g., in the deep sea and coral reefs, eDNA could at least offer a list of “dark taxa” [45] that inspires further exploration and efforts to capture specimens. Given the simplicity of collecting water samples, a wide variety of interested persons could participate in surveys [46]. Looking ahead, how best to assemble and display aquatic eDNA analyses generated by different researchers deserves attention. Portals that compile wildlife observations (e.g., FISHBase, MARCO, eBIRD) benefit both the public and science.
Materials and methods
Hudson River estuary checklist
A compiled occurrence list of 85 fish species was constructed from 12 local surveys conducted between 1988–2015 [26–28] (S1 Table). Chondrichthyans were excluded due to primer mismatch as described below. Species were classified according to number of surveys in which they were detected: abundant (9–12), common (5–8), or uncommon (1–4). Most (80%) of the 85 species have 12S target region sequences in GenBank or generated in this study (S1 and S4 Tables).
Primer evaluation
We first tested whether a broad-range vertebrate 12S-V5 region primer pair [29] would amplify local freshwater and marine vertebrate species. in silico evaluation demonstrated highly conserved primer binding sites (S7 Table). Sharks and rays were an exception; most local species have a mismatch at the 5’ primer terminal nucleotide, likely to inhibit amplification. The amplified segment differs among most local species, allowing species-level identification. A pilot study (23 samples at 10 estuary sites) confirmed 12S eDNA amplification of one or more marine fish species in all samples.
New 12S reference sequences
The pilot study described above generated a number of sequences with absent, incomplete, or non-exact GenBank matches. Based on these findings and our checklist of local fish, we sought tissues of relevant species. 31 specimens representing 18 species were obtained from fisheries researchers, local fish stores, or bait shops (S8 Table). Reference specimens were sequenced for an approximately 725 bp fragment of 12S that encompasses the eDNA target site and for 648 bp COI barcode region, the latter to confirm species identification. Genetic identification of reference specimens was desirable because some were fin clips or filets, thus identifications could not be morphologically verified. The newly generated COI barcode region sequences were matched to existing reference sequences in GenBank using BLAST and in BOLD using BOLD ID engine. Species-level assignments with COI barcodes were possible because 1) most local marine fish species have COI reference sequences in GenBank or BOLD (whereas the representation of fish 12S sequences in the databases is much sparser) and 2) most fish species have diagnostic differences in this gene fragment. For those species which already had both COI and 12S sequences in GenBank, there were no differences in taxonomic assignments of reference specimens based on COI as compared to 12S. For amplification of 12S, M13-tailed versions of primer pair 12S229F/12S954R were utilized with amplification parameters as described [47]. For COI barcode region, COI-3 fish primer cocktail and amplification parameters were as described [48]. Bidirectional sequencing using M13 primers was done at GENEWIZ. Sequences are deposited in GenBank (12S accession nos. KX686069-KX686099; COI, KX688296-KX688324).
Collection sites
Sampling was done at two estuary locations with contrasting hydrology [18,49] (Figs 1 and 2). The East River site (40.760443, -73.956354) is a narrow (275 m), deep (8 m at shoreline to 30 m mid-channel), high flow (2.7 m/sec and 2.5 m/sec, on ebb and flood tides, respectively) tidal strait, with salinity and transparency close to seawater. The Hudson River location (40.794711, -73.978902) is more typical estuarine: broad (1325 m), sloping (1 m at shoreline to 14 m mid-channel), relatively low flow (1.4 and 1.0 m/sec on ebb and flood tides, respectively), with lower salinity (on average about one-half that of seawater) and higher turbidity than the East River.
Sampling method and schedule
Water sampling in protected areas was collected under permit from New York City Department of Parks and Recreation. Fish specimens were collected under permit from New York State Department of Environmental Conservation by NYSDEC personnel, or were obtained from scientific collections at Monmouth University. All sampling procedures were approved as part of the field permit. No animals were housed or experimented upon as part of this study. No endangered or protected species were collected. One liter surface water samples were collected from the shoreline at both sites approximately weekly from January 2016 to July 2016. Assuming that many species would be difficult to detect, on most (60%) days a second collection was made on the opposite tide (approximately 6 h later) to maximize the chance of detection. Analyzing possible differences by tide would require many more samples at standardized points in the tidal cycle. Because the shoreline at the sampling sites is armored and elevated, water was collected with a bucket on a rope and transferred to two 500 mL plastic bottles (Fig 2).
Filtration
All work described below including DNA analysis was performed on a benchtop, with a separate work area for post-PCR procedures. After collection, samples were filtered within 1 h or stored at 4°C for up to 72 h before filtration. Filtration apparatus consisted of a 1000 mL side arm flask attached to wall suction, a frittered glass filter holder (Millipore), and a 47 mm, 0.45 µM pore size nylon filter (Millipore). Before filtration, water was poured through two paper coffee filters (Melitta) to exclude large particulate matter. After filtration, nylon filters were folded to cover the retained material and stored in 15 mL conical tubes at -20° C until DNA extraction. As environmental controls, we filtered water samples from controlled or remote environments that do not share fish taxa with the Hudson estuary (New York Aquarium; Marco Island, FL). There was no evidence of cross-contamination of fish eDNAs, i.e., none of the fish eDNAs present in local samples were detected in controls and vice versa (S1 File). The results themselves support the validity of detections—consistent differences over time and between sites, and the congruence of eDNA and seining studies. This strategy was not informative as to the source(s) of human and domestic animal DNA, which was detected in samples from experimental and control environments—as frequently observed in eDNA studies [38].
DNA extraction, amplification, and library construction
Filters were processed using MoBio Powersoil Kit with modifications from the manufacturer’s protocol to accommodate extraction from a filter as described in S9 Table. Following extraction, DNA concentration was measured using Qubit (Thermo Fisher Scientific) (typical yield 1–5 µg DNA/L water), samples were further purified with AMPure XP (Beckman Coulter) following manufacturer’s protocol, and resuspended in 50 µl of Elution Buffer (10 mM Tris, pH 8.3) (Qiagen).
To facilitate library construction, we adapted the Illumina 16S metabarcoding protocol, adding tails to 12S-V5 primers [29] described above. Primers were obtained from Integrated DNA Technologies with the following sequences (Illumina tails in italics):
Forward: 5’- TCG TCG GCA GCG TCA GAT GTG TAT AAG AGA CAG ACT GGG ATT AGA TAC CCC -3’. Reverse: 5’- GTC TCG TGG GCT CGG AGA TGT GTA TAA GAG ACA GTA GAA CAG GCT CCT CTA G -3’. The amplified target not including primers is approximately 110 bp; the entire amplicon including tailed primers is approximately 200 bp.
Amplifications were done with Illustra puReTaq Ready-To-Go PCR beads (GE Healthcare), 5 μl DNA (representing eDNA from 100 mL of water from estuary or control samples) or 5 μl H2O for negative PCR control, 200 nM each primer, in final volume 25 μl. Parameters were 95°C x 7 m, then 40 cycles of (95°C x 30 s, 52°C x 30 s, 72°C x 30 s), followed by 72°C for 10 m, and hold at 4°C. 5 μl of each reaction were run on a 2% agarose gel with SYBR Safe dye (Thermo Fisher Scientific) to visualize amplifications and confirm negative control. PCR products were treated with AMPure XP to remove unincorporated primers and nucleotides and resuspended in 40 μl Elution Buffer.
To enable pooling of libraries, Nextera index primers (Illumina) were added following manufacturer’s protocol, using 10 μl of 12S PCR product, 2.5 μl each primer, GE Illustra beads with final volume of 25 μl, and amplification parameters 95°C x 3 m, then 8 cycles of (95°C x 30 s, 55°C x 30 s, 72°C x 30 s), extension at 72°C x 5 m, hold at 4°C. 5 μl of each reaction were run on a 2% agarose gel with SYBR Safe dye to confirm amplification.
Indexed PCR products were treated with AMPure XP, resuspended in 40 μl Elution Buffer, and DNA concentration was determined with Qubit. A pooled sample containing 27 ng of each library at 15 nM was sequenced at GENEWIZ on an Illumina MiSeq (2 x 150 bp). Negative library controls as described above were included in each pool. The 76 experimental and 11 control libraries, plus samples from other studies not reported here, were analyzed in four MiSeq runs with 35–60 libraries per run.
To assess reproducibility, 42 DNA samples were re-amplified, indexed, and submitted for MiSeq sequencing, and a two-way comparison of each pair of amplifications was performed. Each taxon detection was classified by number of reads and detection of that taxon in the paired sample (S5 Table).
Bioinformatics
The paired FASTQ files generated by the MiSeq instrument were analyzed using DADA2 [50]. DADA2 was chosen because it uses an error model to infer exact sample sequences that can vary by as little as a single nucleotide. This is an alternative to cluster-based methods that traditionally lump sequences at 3% identity. This technical detail is important because the 12S amplicon is short (~100bp not including primers) and some fish species differ at only one or a few nucleotide positions; clustering would potentially lump such taxa together. DADA2 was used to merge paired FASTQ files and infer sequence variants using the default error model parameters, with one modification. We changed the sentence description of DADA2’s error model. The default behavior for DADA2 is to build an error model for each basepair for every fastq file that is provided. Alternatively, you can build an error model using a subset of the total reads from a sequencing run and provide this model to DADA2. We used the default value and have modified the description of this choice in the methods to read as follows:
“DADA2 was run in “self-consist” mode so that the error-model was independently built for each sample. This error model uses FASTQ quality scores to assess the likelihood at each base-pair that small mutations are due to true changes in the underlying biological sample rather than an error introduced during sequencing.” The primary outputs were a FASTA file of unique sequences (S2 File) and an operational taxonomic unit (OTU) table (S1 File) providing the abundance of each sequence in each experimental sample.
For each MiSeq run, OTU sequence counts were filtered to reduce library assignment errors. In pilot studies, we found misassigned reads in a given library for a given taxa were present on average at about 0.02% of total reads per taxa in the pooled library sample or 0.02% total reads per library. Misassignment classification was based on detection of species from non-overlapping environments (e.g., marine species in freshwater samples). Based on these results, in the present study we excluded detections representing less than 0.1% of reads per taxa or per library, which usually worked out to exclusion of read counts less than 100. Identical thresholds are applied in a recent aquatic eDNA study [25]; more detailed approaches to false-positive errors are recently proposed [51]. The average for non-excluded read counts was much higher, about 5,000. No fish species were entirely eliminated from a MiSeq run by these thresholds.
Environmental library controls were negative for estuary fish eDNAs after filtering (S1 File). Total reads per MiSeq run were roughly similar (about 10 M reads), regardless of the number of libraries in the run (35–60). We therefore considered that reads were relatively inflated in runs with fewer libraries. To enable compiling of results from different runs, we first chose 48 libraries per run as an arbitrary standard. OTU table reads were then adjusted accordingly (#reads x 48/#libraries in the run). Maximum adjustment between library runs was 1.7x. The adjustment did not change any presence/absence results and did not change the timing of eDNA increases for the three species in Fig 4. Alternate approaches are described [52]. The sequential modifications of each DADA2 OTU table are included in S1 File.
Species identifications were made in a multi-step process. Reads that were 100% identical to an internal library (comprised of GenBank 12S sequences for local species and the new 12S sequences generated in this study) were recorded. In the second step, all DADA2 unique reads (including those with 100% match to internal library) were submitted to BLAST engine in GenBank and results were screened for similarity (>90% identity) to vertebrate 12S sequences; those lacking similarity were excluded. About 1% of fish reads were minor variants (1 or 2 nucleotide differences) of 100% matching sequences present in the same run (for details see OTU tables in S1 File); these were excluded. Most of the minor variants were unique to a particular sample and library run—these likely represent sequence errors; others may be biological variants. Final species assignments were made using BLAST results, the checklist of local species, and fish distributions as recorded in FishBase. The identification process applied to all 76 libraries yielded 51 unique fish sequences. Most (45) were named based on 100% full-length identity to one or more reference sequences in GenBank or from this study; three cases were based on 99% identity; and there were three incomplete identifications (to genus or family level) based on 92%-99% identity.
As expected given the short fragment of 12S analyzed, some amplicons gave 100% full-length matches to multiple species. In some cases, a species-level identification could be made based on geographic range, i.e., only one of the matching species is present in the northwest Atlantic. In two cases (herrings, hakes), the amplicon sequence matched multiple local species. These detections were assigned to genus or family level, based on the lowest taxonomic group that comprised all matches. Identification details for individual sequences are in S4 Table. Species identifications including DNA sequences are listed in in S4 Table. Original fastq files with metadata are deposited in NCBI Sequence Read Archive (NCBI BioProject ID PRJNA358446).
Supporting information
S1 Table. Lower Hudson River estuary checklist and eDNA detection.
https://doi.org/10.1371/journal.pone.0175186.s001
(XLSX)
S2 Table. Statisical analysis of fish eDNA detection vs checklist abundance.
https://doi.org/10.1371/journal.pone.0175186.s002
(XLSX)
S5 Table. eDNA detection reproducibility with repeat PCR.
https://doi.org/10.1371/journal.pone.0175186.s005
(XLSX)
S9 Table. Modified MOBIO PowerSoil DNA extraction protocol.
https://doi.org/10.1371/journal.pone.0175186.s009
(XLSX)
Acknowledgments
We thank Jesse Ausubel for encouragement and editorial comment, Sean Brady for advice on experimental design, Jeanne Garbarino and Anna Zeidman for laboratory space and assistance, Keith Dunton for contributing fish specimens and sharing Long River Survey report, Melissa Cohen for contributing fish specimens, Howard Rosenbaum for permission to sample at New York Aquarium, and Julie Nadel, Iman Nassef, and Alden Liang for trialing protocols.
Author Contributions
- Conceptualization: MYS.
- Data curation: MYS ZCP.
- Formal analysis: MYS ZCP.
- Funding acquisition: MYS ZCP.
- Investigation: MYS LS.
- Methodology: MYS ZCP.
- Project administration: MYS.
- Resources: MYS LS ZCP.
- Software: ZCP.
- Supervision: MYS.
- Validation: MYS ZCP.
- Visualization: MYS.
- Writing – original draft: MYS ZCP.
- Writing – review & editing: MYS ZCP LS.
References
- 1. Ficetola GF, Miaud C, Pompanon F, Taberlet P. Species detection using environment DNA from water samples. Biol Lett. 2008;4: 423–425. pmid:18400683
- 2. Bourlat SJ, Borja A, Gilbert J, Taylor MI, Davis N, Weisberg SB, et al. Genomics in marine monitoring: new opportunities for assessing marine health status. Marine Pollution Bull. 2013;74: 19–31.
- 3. Kelly RP, Port JA, Yamahara KM, Martone RG, Lowell N, Thomsen PF, et al. Harnessing DNA to improve environmental management. Science. 2014; 344: 1455–1456. pmid:24970068
- 4. Matsui K, Honjo M, Kawabata Z. Estimation of the fate of dissolved DNA in thermally stratified lake water from the stability of exogenous plasmid DNA. Aquatic Microb Ecol. 2001;26: 95–102.
- 5. Dejean T, Valentini A, Duparc A, Pellier-Cuit S, Pompanon F, Taberlet P, et al. Persistence of environmental DNA in freshwater ecosystems. PLoS ONE. 2011;6: e23398. pmid:21858099
- 6. Takahara T, Minamoto T, Yamanaka H, Doi H, Kawabata Z. Estimation of fish biomass using environmental DNA. PLoS ONE 2012;7: e335868.
- 7. Takahara T, Minamoto T, Doi H. Using environmental DNA to estimate the distribution of an invasive fish species in ponds. PLoS ONE. 2013;8: e56584. pmid:23437177
- 8. Turner CR, Barnes MA, Xu CY, Jones SE, Jerde CL, Lodge DM. Particle size distribution and optimal capture of aqueous microbial eDNA. Methods Ecol Evol. 2014;5: 676–684.
- 9. Laramie MB, Pilliod DS, Goldberg CS. Characterizing the distribution of an endangered salmonid using environmental DNA analysis. Biol Conservation. 2015; 183: 29–37.
- 10. Klymus KE, Richter CA, Chapman DC, Paukert C. Quantification of eDNA shedding rates from invasive bighead carp Hypophthalmichthys nobilis and silver carp Hypophthalmichthys molitrix. Biol Conservation. 2015;183: 77–84.
- 11. Evans NT, Olds BP, Renshaw MA, Turner CR, Li Y, Jerde CL, et al. Quantification of mesocosm fish and amphibian species diversity via environmental DNA metabarcoding. Mol Ecol Resources. 2016;16: 29–41.
- 12. Thomsen PF, Kielgast J, Iversen LL, Moller PR, Rasmussen M, Willerslev E. Detection of diverse marine fish fauna using environmental DNA from seawater samples. PLoS ONE. 2012;7: e41732. pmid:22952584
- 13. Foote AD, Thomsen PF, Sveegaard S, Wahlberg M, Kielgast J, Kyhn LA, et al. Investigating the potential use of environmental DNA (eDNA) for genetic monitoring of marine mammals. PLoS ONE. 2012;7: e41781. pmid:22952587
- 14. Kelly RP, Port JA, Yamahara KM, Crowder LB. Using environmental DNA to census marine fish in a large mesocosm. PLoS ONE 2014;9: e86175 pmid:24454960
- 15. Port JA, O’Donnell JL, Romero-Maraccini OC, Leary PR, Litvin SY, Nickols KJ, et al. Assessing vertebrate biodiversity in a kelp forest using environmental DNA. Mol Ecol. 2016;25: 527–541. pmid:26586544
- 16. Thomsen PF, Moller PR, Sigsgaard EE, Knudsen W, Jorgensen OA, Willerslev E. Environmental DNA from seawater samples correlate with trawl catches of subarctic, deepwater fishes. PLoS ONE 2016;11: e0165252. pmid:27851757
- 17. Cooper JC, Cantelmo FR, Newton CE. Overview of the Hudson River estuary. Amer Fisheries Soc Monograph. 1988;4: 11–24.
- 18.
New York City Department of Environmental Protection. New York harbor water quality report, 2011. Available 11/30/2016 from www.nyc.gov/html/dep/pdf/hwqs2011.pdf.
- 19.
Riverkeeper. How’s the water? Fecal contamination in the Hudson River and its Tributaries. 2015. Available 11/30/2016 from https://www.riverkeeper.org/wp-content/uploads/2015/06/Riverkeeper_WQReport_2015_Final.pdf.
- 20.
National Oceanographic and Atmospheric Administration. Habitat use and requirements of important fish species inhabiting the Hudson River estuary: availability of information. NOAA Technical Memorandam NMFS-NE-121. 1999. Available 11/30/2016 from www.nefsc.noaa.gov/publications/tm/tm121/tm121.pdf.
- 21. Jerde CL, Mahon AR, Chadderton WL, Lodge DM. “Sight-unseen” detection of rare aquatic species using environmental DNA. Conservation Lett. 2010;4: 150–157.
- 22. Dejean T, Valentini A, Miquel C, Taberlet P, Bellemain E, Miaud C. Improved detection of an alien invasive species through environmental DNA barcoding: the example of the American bullfrog Lithobates catesbeianus. J Applied Ecol. 2012;49: 953–959.
- 23. Thomsen PF, Kielgast J, Iversen LL, Wiuf C, Rasmussen M, Gilbert MTP, et al. Monitoring endangered freshwater biodiversity using environmental DNA. Mol Ecol. 2012;21: 2565–2573. pmid:22151771
- 24. Wilson C, Wright E, Bronnenhuber J, MacDonald F, Belore M, Locke B. Tracking ghosts: combined electrofishing and environmental DNA surveillance efforts for Asian carps in Ontario waters of Lake Erie. Management Biol Invasions. 2014;5: 225–231.
- 25. Valentini A, Taberlet P, Miaud C, Civade R, Herder J, Thomsen PF et al. Next-generation monitoring of aquatic biodiversity using environmental DNA barcoding. Mol Ecol. 2016;25: 929–942. pmid:26479867
- 26.
U.S. Army Corps of Engineers. New York and New Jersey harbor deepening project, Appendix E: Essential fish habitat assessment. 2004. Available 11/20/2016 from http://www.nan.usace.army.mil/Portals/37/docs/harbor/Harprogrep/appE.pdf.
- 27.
The River Project. Fish caught at Hudson River Park 1988–2015. Available 12/1/2015 from http://www.riverprojectnyc.org/riverdive_fish.php.
- 28.
ASA Analysis & Communication. 2013 Year Class Report for the Hudson River Estuary Monitoring Program. 2015.
- 29. Riaz T, Shehzad W, Viari A, Pompanon F, Taberlet P, Coissac E. ecoPrimers: inference of new DNA barcode markers from whole genome sequence analysis. Nucl Acids Res. 2011;39: e145. pmid:21930509
- 30. Bigelow HB, Schroeder WC. Fishes of the Gulf of Maine, First Revision. Fishery Bull Fish Wildlife Service. 1953;53: 1–577.
- 31.
Able KW, Fahay MP. The First Year in the Life of Estuarine Fishes in the Middle Atlantic Bight. New Brunswick: Rutgers University Press; 1992.
- 32.
Waldman JR. Diadromous fish fauna of the Hudson River: life histories, conservation concerns, and research avenues. In: Levinton JS, Waldman JR, editors. The Hudson River Estuary. New York: Cambridge University Press: New York; 2006, pp. 171–188.
- 33.
Kells V, Carpenter K. A Field Guide to Coastal Fishes: from Maine to Texas. Baltimore: The Johns Hopkins University Press; 2011.
- 34. Ficetola GF, Pansu J, Bonin A, Coissac E, Giguet-Covex C, DeBarba M, et al. Replication levels, false presences and the estimation of the presence/absence from eDNA metabarcoding data. Mol Ecol Resources. 2015;15: 543–556.
- 35. Simmons M, Tucker A, Chadderton WL, Jerde CL, Mahon AR. Active and passive environmental DNA surveillance of aquatic invasive species. Can J Fish Aquat Sci. 2015;73: 76–83.
- 36. Sassoubre LM, Yamahara KM, Gardner LD, Block BA, Boehm AB. Quantification of environmental DNA (eDNA) shedding and decay rates for three marine fish. Environ Sci Technol. 2016;50: 10456–10464. pmid:27580258
- 37. Deagle BE, Jarman SN, Coissac E, Pompanon F, Taberlet P. DNA metabarcoding and the cytochrome c oxidase subunit I marker: not a perfect match. Biol Lett. 2014;10: 20140562. pmid:25209199
- 38. Leonard JA, Shanks O, Hofreiter M, Kreuz E, Hodges L, Ream W, et al. Animal DNA in PCR reagents plagues ancient DNA research. J Anthrop Sci. 2007;34: 1361–1366.
- 39. Miya M, Sato Y, Fukunaga T, Sado T, Poulsen JY, Sato K, et al. MiFish, a set of universal PCR primers for metabarcoding environmental DNA from fishes: detection of more than 230 subtropical marine species. R Soc Open Sci. 2015;2: 150088. pmid:26587265
- 40. Krause J, Fu Q, Good JM, Viola B, Shunkov MV, Derevianko AP et al. The complete mitochondrial DNA genome of an unknown hominin from southern Siberia. Nature. 2010;464: 894–897. pmid:20336068
- 41. Leese F, Altermatt F. Bouchez A, Ekrem T, Hering D, Meissner K, et al. DNAqua-Net: developing new genetic tools for bioassessment and monitoring of aquatic ecosystems in Europe. Res Ideas Outcomes. 2016;2: e11321.
- 42. Yamamoto S, Masuda R, Sato Y, Sado T, Araki H, Kondoh M, et al. Environmental DNA metabarcoding reveals local fish communities in a species-rich coastal sea. Sci Rep 2017;7: 40368. pmid:28079122
- 43. Kelly RP, Closek CJ, O’Donnell JL, Kralj JE, Shelton AO, Samhouri JF. Genetic and manual survey methods yield different and complementary views of an ecosystem. Front Marine Sci 2017;3: 1–11.
- 44. Sigsgaard EE, Nielsen IB, Bach SS, Lorenzen ED, Robinson DP, Knudsen SW, et al. Population characteristics of a large whale shark aggregation inferred from seawater environmental DNA. Ecol Evol 2016;1: 1–4.
- 45.
Page R. Dark taxa: GenBank in a post-taxonomic world. iPhylo blog, 4/12/2011. Available 12/1/2016 from http://iphylo.blogspot.com/2011/04/dark-taxa-genbank-in-post-taxonomic.html.
- 46. Biggs J, Ewald N, Valentini A, Gaboriaud C, Dejean T, Griffiths RA, et al. Using eDNA to develop a national citizen science-based monitoring programme for the great crested newt (Triturus cristatus). Biol Conservation. 2015;183: 19–28.
- 47. Li C, Orti G. Molecular phylogeny of Clupeiformes (Actinopterygii) inferred from nuclear and mitochondrial DNA sequences. Mol Phylogenet Evol. 2007;44: 386–398. pmid:17161957
- 48. Ivanova NV, Zemlak TS, Hanner RH, Hebert PDN. Universal primer cocktails for fish DNA barcoding. Mol Ecol Notes. 2007;7: 544–548.
- 49. Blumberg AF, Pritchard DW. Estimates of transport through the East River, New York. J Geophysical Res. 1997;102: 5685–5703.
- 50. Callahan BJ, McMurdie PJ, Rosen NJ, Han AW, Johnson AJA, Holmes SP. DADA2: High-resolution sample inference from Illumina amplicon data. Nature Methods. 2016;13: 581–583. pmid:27214047
- 51. Lahoz-Monfort JJ, Guillera-Arroita G, Tingley R. Statistical approaches to account for false-positive errors in environmental DNA samples. Mol Ecol Res 2016;16:673–683.
- 52. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 2014;15: 550. pmid:25516281