TITLE PAGE
TITLE: Red fish, blue fish, native fish, new fish: eDNA as a tool to
monitor fish in estuarine systems
RUNNING TITLE eDNA as a tool to monitor fish in estuarine systems
AUTHORS
Corresponding author: Alison W. Watts, University of New Hampshire,
Alison.watts@unh.edu
Contributing authors:
Laura C. Crane, Wells National Estuarine Research Reserve,
lcrane@wellsnerr.org
Jason Garwood, Apalachicola National Estuarine Research Reserve,
Jason.garwood@FloridaDEP.gov
Jason S. Goldstein, Wells National Estuarine Research Reserve,
jgoldstein@wellsnerr.org
Megan S. Lamb, Apalachicola National Estuarine Research Reserve,
Megan.Lamb@FloridaDEP.gov
Christopher R. Peter, Great Bay National Estuarine Research Reserve,
christopher.r.peter@wildlife.nh.gov
Yoshimi M. Rii, Hawaiʻi Institute of Marine Biology, University of
Hawaiʻi at Mānoa,
shimi@hawaii.edu
Shon Schooler, South Slough National Estuarine Research Reserve,
shon.schooler@dsl.oregon.gov
Devin W. Thomas, University of New Hampshire,
Devin.thomas@unh.edu
W. Kelley Thomas, University of New Hampshire,
Kelly.thomas@unh.edu
Bree K. Yednock, South Slough National Estuarine Research Reserve,
bree.yednock@dsl.oregon.gov
ABSTRACT
Environmental DNA (eDNA) is emerging as a potentially powerful tool to
assess aquatic ecosystems, and to monitor fish assemblages. We conducted
a pilot eDNA water sampling program at 5 Reserves within the National
Estuarine Research Reserve System from temperate and tropical estuaries
to better understand how eDNA data can support detection of fish species
within an standardized monitoring network. Samples were collected in
coordination with an existing long term monitoring program, and the
sampling plan was designed to minimize additional work for field teams.
We found that results varied; in some estuaries the number and type of
fish species was consistent with expected occurrence, and eDNA analysis
detected fish that often eluded traditional sampling methods. In warm
turbid waters, however, we detected fewer species than expected,
suggesting that additional processing is required at these sites.
Managers interested in applying standardized eDNA monitoring across a
range of sites may want to consider a decision tree protocol, where a
baseline methodology is developed for all sites, enhanced by additional
laboratory or analysis steps when specific conditions are present.
1. INTRODUCTION
Estuaries are among the most prolific habitats in the world, comprising
complex and diverse ecosystems that are home to a wide range of species.
They serve as both primary and nursery habitat for numerous fish
including commercially important, rare and endangered, and invasive
species (Kelley, 1988; Lenanton et al., 1987). Estuaries are also areas
of high human activity that can influence fish communities and habitats,
through shoreline development, global transportation, commercial and
recreational fishing, and other industries (Kennish, 2002; Schulz et
al., 2020). Management, conservation, and restoration in these regions
rely on accurate assessments of the species present and how they vary
over space and time (Fleishman et al., 2011). Although comprehensive
species and habitat assessments with conventional monitoring methods
provide a strong basis for understanding estuarine structure and
function (Neckles et al., 2015), they are often time-consuming,
expensive, and subject to technical and resource limitations. For
example, traditional capture-and-release methods are labor-intensive and
require extensive capacity, limiting the frequency of assessments and
the ability to track temporal dynamics of fish communities. Many fish
and invertebrates may also evade capture; species that are larger and
faster may escape, while very small organisms may be able to move
through nets, resulting in biased data. In addition, capture-and-release
techniques are not universally applicable within all estuarine
environments, due to bank and bottom topography, water depth,
limitations on capture of sensitive species, and other environmental
factors (Bulleri & Chapman, 2010; Kennish, 2002).
Because of these constraints, traditional monitoring methods may miss
early detection of newly-arrived invasive species or losses of rare
native species that occur in low densities or at locations that are
difficult to survey. These factors highlight the importance of
developing alternative methods and tools that may individually, or in
combination with traditional methods, support more comprehensive
monitoring in diverse environments. Underwater video, passive and active
acoustic monitoring and environmental DNA (eDNA) are potentially
powerful methods of augmenting traditional monitoring. Here we will
explore the use of eDNA-based monitoring for fish species at five
estuaries within the National Estuarine Research Reserve System (NERRS).
Advances in DNA methods and rapid reductions in analytical costs present
an opportunity to harness this new technology and fundamentally improve
our capacity to monitor biological communities and individual species
(Baillie et al., 2019; Gilbey et al., 2021; Ip et al., 2021; Pawlowski
et al., 2018; Thomsen & Willerslev, 2015). DNA present in a water
sample is comprised of whole microorganisms (microalgae, bacteria,
etc.), fragments of tissue, reproductive and waste products, and other
cellular material. eDNA analysis allow resource managers to identify
species in an aquatic ecosystem without having to actually capture and
physically identify individual organisms. eDNA monitoring protocols are
being developed for monitoring invasive species targeted for control,
and for native species that may be in decline for special management
attention (e.g. Crane et al., 2021; Darling & Mahon, 2011; Goldberg et
al., 2016). eDNA methods have limitations, including the potential for
false negative or positive results, vulnerability to contamination,
misinterpretation of transported DNA and database limitations that may
preclude identification of place-specific species (Jerde, 2019). In
addition, eDNA results yield relative abundance of species within the
sample, making data analysis and comparison beyond presence/absence
difficult. Estuaries present particular challenges to eDNA studies;
turbid water, transport from rivers, tidal forces, and shipping can
complicate interpretation of results, and the use of eDNA to support
coastal management has been limited to date. In this study, we evaluate
the use of eDNA for fish surveys in estuarine settings as an additional
tool for resource managers to support monitoring and assessment of
coastal ecosystems.
1.1. National Estuarine Research Reserve System
(NERRS)
The NERRS is a network of estuarine research sites founded in 1972 under
the United States Coastal Zone Management Act
(https://coast.noaa.gov/czm/). At present, a total of 29 NERRs are
designated and funded through a partnership between the National Oceanic
and Atmospheric Administration (NOAA) and a state agency, university, or
non-profit organization. Each NERR is subdivided into four distinct, but
integrated sections: Stewardship, Research, Coastal Training, and
Education sectors are each focused on the overarching goal of coastal
management. The research and monitoring sector of the NERRS includes the
System-Wide Monitoring Program (SWMP), which encompasses monitoring for
water quality, biological species, and watershed and land-use/land-cover
characterizations to examine short- and long-term change (Kennish, 2004;
Sanger, 2002). The SWMP program is a required element of all NERRs and
serves as a system-wide network of ecosystem-based monitoring with a
large quantitative database produced through standardized operating
procedures and data quality and data assurance metrics. The SWMP program
thus results in initiatives to prioritize research in estuarine systems
for understanding the structure and function of coastal wetlands for
sustainability and future resilience. The SWMP initiative enables
inter-and intra-NERR comparability of environmental characteristics of
these systems; however, NERRs span multiple biogeographic regions
representing multiple unique ecological habitats. Therefore, monitoring
programs focused on biotic factors are more difficult to standardize.
These constraints amplify the need to formulate methods to link
biological research initiatives in a systematic manner among the NERRs.
The addition of standardized methods and recommendations for eDNA
sampling to the existing, long-term monitoring program at these sites
would be crucial to examining changes in biological diversity and
understanding the underlying mechanisms driving such changes across the
Reserve system.
The use of eDNA has a tremendous range of research applicability from
tracking rare or invasive species to assessing biodiversity at all
trophic levels (Andruszkiewicz et al., 2017; Baird & Hajibabaei, 2012;
Giakoumi et al., 2016; Gilbey et al., 2021). For management areas
focusing on fish communities, eDNA is a promising monitoring tool to
help identify species composition by augmenting or potentially replacing
traditional surveys to track communities over time affected by local or
global driving factors (e.g., eutrophication, dam removal, climate
change), and further improve habitat management for fish production
(Aglieri et al., 2020; Carim et al., 2017; Gilbey et al., 2021; Nguyen
et al., 2020). However, as with any new method, eDNA must be field
tested and evaluated to better understand where application is
appropriate and useful, and where other methods might better serve
management needs.
NERRS staff have identified management needs that generally center
around identifying and mitigating impacts from climate change,
eutrophication and other forms of contamination, advancing our
understanding of ecosystem function, and investigating traditional and
novel options for habitat restoration, all of which have the potential
for advancement through eDNA monitoring. However, eDNA is a relatively
new tool, with limitations associated with species identification,
analytic and bioinformatic capacity, and interpretation of results. As
part of the process of understanding the strengths and weakness of using
eDNA to support estuarine management, we designed and implemented an
eDNA sampling program at 5 NERR sites in Apalachicola FL, Great Bay NH,
Heʻeia HI, South Slough OR, and Wells ME (Figure 1). The Hudson Bay NERR
in NY also participated, but samples were collected in a tributary
stream rather than an estuary, so the results are not included in this
analysis. The overall goal was to deploy eDNA-based biological
assessments as a complement to existing, long-term monitoring programs
in a range of coastal settings to support estuarine management. There
are many viable approaches to eDNA sampling, and project specific
methods were selected based on project goals, end user needs, and the
capacity of the science team, including field, lab, and analytic
resources. We attempted to identify a practical, time efficient approach
with a level of detail and specificity that complement existing NERRS
SWMP and resources. The selected sampling methods are readily applied by
a field team with access to a clean workspace and filtration apparatus.
2. MATERIALS and METHODS
2.1. Field sampling
Samples were collected from January-November 2019 in 5 estuaries.
Between 4 and 15 locations were sampled in each estuary with most of the
samples collected at existing SWMP sites (Table 1). Replicate samples
were collected from just below the water surface in new, sterile
Whirl-pak bags or acid-washed or bleached sample bottles, then
transported to a clean laboratory for filtering. Larger volume (5-L)
samples were collected at South Slough in June and August 2019 and
field-filtered using a portable peristaltic pump. Smaller volumes
(500ml) were collected at Wells. A negative field control, consisting of
lab or bottled water was opened at the site, then closed and processed
with field samples. Collected samples were kept in a cooler during
transport, and either frozen or filtered within 24 hours. Disposable
nitrile gloves were worn throughout this process and all sampling
equipment was either new/single use or sterilized with bleach or 10%
w/v hydrochloric acid solution. Samples were filtered through 1.5 µm
glass fiber filters (VWR 691) using a vacuum or peristaltic pump and
stored at -20°C or -80°C. Frozen filters were packed in ice and
delivered or shipped overnight to the lab. Prior to extraction, sample
filters were cut in half, with one half retained for storage and one
half processed for DNA extraction.
2.2. Extraction and PCR
All samples were extracted at the Uviversity of New Hampshire Hubbard
Center for Genome Studies in a dedicated lab space separate from PCR and
raw sample handling. Extraction was performed following the protocols
given in the QIAamp DNA Mini Hanbook (Qiagen, 2016); filters were placed
in a lyse and spin basket with 400 mL of buffer ATL and 20 µL of
proteinase K and incubated at 56 °C for 1 hr, then centrifuged. The
remainder of the filter extraction was performed on a QIAcube Connect
system (QIAGEN®, Hilden, Germany) following the QIAamp DNA Mini Kit
program. Tissue collected for voucher specimens was extracted with
DNeasy Blood and Tissue kits. One negative control sample was extracted
with every 11 field samples. DNA concentration was measured using an
Invitrogen Qubit 2.0 (Thermo Fisher®, Waltham, MA). The manufacturer’s
protocol for the Qubit dsDNA HS Assay kit was followed using 1 uL of
sample. Samples were kept at -20°C until processed, then transferred to
-80°C storage.
A 12S rRNA metabarcoding assay designed for fish (Miya et al., 2015) was
applied in a two-step process. After the first round of thermocycling,
the presence of target DNA was confirmed with gel electrophoresis, then
TrueSeq adapters were added, and a second round of PCR was conducted. A
negative and positive control was included with each PCR set. Most of
the samples were processed using the following thermocycling profile:
Step 1) 95 °C for 3 min, 2) 98 °C for 30 sec, 3) 55 °C for 30 sec, 4) 72
°C for 30 sec, 5) repeat steps 2 to 4 43 times, 6) 72 °C for 5 min, and
finally 7) hold at 4°C. Samples collected during the summer in South
Slough contained bacteria sequences that masked the 12S response. These
samples were processed using a Touchdown procedure (Pitz et al., 2020)
for step 1, then processed in the same manner as other samples for step
2. Samples were then sequenced on an Illumina HiSeq 2500 (Illumina Inc.,
San Diego, CA), demultiplexed, and returned for bioinformatic analysis.
All samples were diluted 1:10 to reduce inhibition, and amplification in
each sample was confirmed by reviewing gels after the first PCR
application. The extraction method used (Qiagen QIAmp DNA) includes
inhibition removal, and was generally very effective. In rare cases
where inhibition was apparent, samples were re-extracted (from the
reserved ½ filter) using PowerSoil Pro (Qiagen) which has enhanced
inhibition removal. Bands of competing sequences could also be observed
on gels, particularly 16S bacterial sequences, which in some samples
completely overwhelmed the fainter 12S bands. In these samples, gel
bands were excised to isolate target sequences. These efforts improved
the number of reads associated with each detection, but did not increase
the number of fish species detected. We also applied a cytochrome c
oxidase subunit I (COI) primer (Leray et al., 2013) to a subset of
samples, without substantial increase in number of species detected.
2.3. Bioinformatics and data analysis The demultiplexed sequence data was processed through QIIME2 dada2
(Caporaso et al., 2018) with a truncation length of 150 bp for both the
forward and reverse reads, and with a trim length of either 21 bp
forward, 27 bp reverse or 27 bp forward and 33 bp reverse depending on
whether the MiFish primers used included a 6 bp spacer. Each lane was
denoised separately, and the results were merged together. An initial
approximate taxonomy was done comparing the sequences to a reference set
of 12S sequences downloaded from NCBI
https://github.com/dwthomas/MiFish-Reference-Database.
Unassigned and off target sequences are filtered from the samples, and
the finalized taxonomy was curated by hand using the web interface to
BLAST to manually assign taxonomy, based on similar reference sequences,
and knowledge of the geographic range of place-specific species. This
draft species ass In a few cases, improbable species identified in a
sample were traced to lab contamination. In these cases the suspect
species were flagged and removed from the sample list. Non-fish
chordates, such as mammals and birds, were segregated from the fish
results.
Species accumulation curves were generated for each site; incidence data
was uploaded to iNEXT online, which computes seamless interpolation and
extrapolation using Hill numbers. An order hyperparameter of q = 0 was
selected, corresponding to the species richness. The hill number with
incidence data counts the effective number of equally frequent species
in the assemblage (Chao et al., 2014). Species accumulation curves were
calculated using iNEXT online and plotted with matplotlib (Hsieh et al.,
2016).
2.4. Traditional fish surveys The approximate number of fish species expected to be present was
provided by resource managers at each site based on long term surveys
and local knowledge. Long term survey methods vary, but included seine
surveys (Great Bay, Apalachicola, South Slough), larval fish tows
(Wells), cast net and visual surveys (He’eia). Managers augmented these
data with information on species likely present, but not always captured
by common survey methods. For example American Eel (Anguilla
rostrata ) is present in the Great Bay region, but rarely caught in
estuarine seine surveys. They used this information to estimate the
number of fish species likely present in the estuaries at each site to
allow comparison with total number of species detected with eDNA. In
Great Bay a more detailed species comparison was conducted by pairing
eDNA sampling with seine surveys performed by New Hampshire Fish and
Game during the same time frame; seining was conducted using a bag seine
(30.5 m long and 1.8 m high, 6.4 mm mesh). Seine hauls were performed
during daylight within 2 hours of low tide and were set by boat 15–25 m
from shoreline. In September 2019, water samples were collected for DNA
analysis in coordination with seine sampling at four locations. Water
samples were collected a day after the seining to avoid cross
contamination and species disturbance associated from the net and boat.
3. RESULTS
Overall, a total of 124 unique fish species were detected in over 370
environmental samples (Figure 2). Most of the species detected represent
fish known to be in the region, with some exceptions discussed below.
Taxa from the same family which are clearly different, but cannot be
resolved to the species level are indicated by ‘sp.’ Many fish were only
detected in one sample, while the more abundant species are generally
detected more frequently.
In cases where sequence data identifiedd a species that was not likely
to be present in the system, we re-reviewed all lab and bioinformatics
QA steps to ensure that neither contamination nor incorrect sequence
assignment were likely. We then considered the local estuarine
environment to identify potential DNA transport pathways. In several
cases, known commercial or recreational fisheries were noted as likely
sources: In South Slough both hagfish (Eptatretus sp. ) and
albacore (Thunnus alalunga ) were identified in eDNA samples
during this study. Although not likely to be living in the region, both
species are handled at local commercial facilities. In Apalachicola, an
estuarine grouper (Epinephelus sp .) was detected in a freshwater
tributary, probably transported by recreational fishing staging from a
nearby dock. At all sites, DNA from freshwater species known to be
present in the watershed was detected occasionally in estuarine samples.
Detections of freshwater fish were most common near the entrance to
tributary rivers, indicating transport from upstream reaches. In
contrast, at Heʻeia, DNA from 3 different native goby species was
detected at the specific watershed sites where they have been detected
through other methods, indicating that DNA transport may be limited in
that system. This is consistent with studies finding that degradation
and transport of DNA fragments in aquatic environments will vary
depending on characteristics such as tidal flow, water chemistry and
temperature (Caza-Allard et al., 2021). Additional species which do not
have a clear source were occasionally detected; some of these are fish
that are known to be present in the region, but further offshore such as
Atlantic flyingfish (Cheilopogon melanurus ).
The number of fish species detected in each sample is shown in Figure 3.
The number of species ranged from 0 to 7 in individual samples. In the
cases where few or no fish species were detected the samples were
reviewed to ensure that inhibition or other issues were not interfering
with amplification.
At most locations triplicate samples were collected. Replicate samples
provide additional information, without adding additional field travel
time. Figure 4 shows the number of fish species detected in triplicate
vs single 1-liter samples in Great Bay in June 2019. Triplicate samples
approximately doubled the number of detected species, although costs
associated with analyses also increase.
4. DISCUSSION
The estuaries sampled in this study represent very different site
conditions (e.g., latitude, geomorphology, temperature, salinity, etc.)
and fish communities. By applying the same sampling and analysis method
at each location we are able to compare the results from a standardized
method to better understand the practical use of eDNA monitoring in
estuarine systems. We considered two questions commonly asked by
resource managers: How many species are detected by eDNA sampling
in a given estuary? How does eDNA-based monitoring compare to
traditional fish surveys? And what is the relative cost of these
methods?