Introduction
In the past decades, environmental DNA (eDNA) analysis has been
developed and successfully applied in multiple fields in ecology,
fisheries, and environmental science (Ficetola et al., 2008; Bálint et
al., 2018; Ruppert et al., 2019; Spear et al., 2021). Environmental DNA
is defined as a total pool of DNA isolated from environmental samples
such as water and sediment (Pawlowski et al., 2020); in a narrower
sense, it is generally defined as extra-organismal DNA released from
macro-organisms in the form of feces, skin, mucus, and gamete (Barnes &
Turner, 2016; Rodriguez-Ezpeleta et al., 2021). Contrary to traditional
methods, PCR-based detection of target eDNA does not require capturing
nor observing individuals, and thus eDNA analysis is a feasible approach
for non-disruptive, highly-sensitive, and cost-effective biomonitoring
(Takahara et al., 2013; Yamanaka & Minamoto, 2016; Deiner et al., 2017;
Djurhuus et al., 2020). Therefore, eDNA analysis has potential to
improve the monitoring of biodiversity and ecosystems, allowing for more
effective conservation and management of biodiversity and resources.
In addition to species presence/absence, eDNA analysis can be used to
estimate species abundance from target eDNA quantity. Several studies
have reported positive relationships between eDNA concentrations and
species abundance for various taxa and environments (e.g., Takahara et
al., 2012; Pilliod et al., 2013; Klymus et al., 2015; Salter et al.,
2019). However, a recent meta-analysis demonstrated that the
relationships between eDNA concentration and species abundance was
weaker in natural environments than in controlled laboratory conditions
(i.e., aquaria, tanks, or mesocosms) (Yates et al., 2019). According to
the study, the mean R2 values were 0.81 and 0.57 in
laboratory conditions and natural environments, respectively. This
finding is intuitively unsurprising given that abundance can be
precisely set in laboratory experiments, but we cannot know ‘true’
species abundance in natural environments where some individuals are not
analyzable depending on their developmental stage and the survey method
(Yates et al., 2019). In addition, the effects of diffusion and
degradation on eDNA detection/quantification would be more substantial
in natural environments due to compounding and complicated environmental
conditions, including temperature, water chemistry, flow rate, and
substrate (Strickler et al., 2015; Jane et al., 2015; Shogren et al.,
2018; Jo et al., 2019a). Such factors could hamper the practical
application of eDNA-based abundance estimation in natural environments
(Hansen et al., 2018). Therefore, toward effective conservation
management of biodiversity and precise stock assessment via eDNA
analysis, it is critical to assess the factors affecting such
variabilities concerning the estimation accuracy and improve the
accuracy of eDNA-based abundance estimation.
Given that the amount of eDNA is determined by a function of its
production, transport, and degradation (Strickler et al., 2015; Barnes
& Turner, 2016; Jo & Minamoto, 2021), the relationships between eDNA
quantity and species abundance may also be affected by target eDNA
characteristics, including its production source and cellular/molecular
state. For example, eDNA production sources and processes may differ
among taxa, which could accordingly influence the estimation accuracy of
species abundance via eDNA analysis, as well as detection
sensitivity of target eDNA. Andruszkiewicz et al. (2021) estimated eDNA
shedding rates (pg/hour) of multiple taxa under similar experimental
conditions and found that crustaceans (Palaeomenes spp.) had
lower shedding rates than fish (Fundulus heteroclitus ) and
scyphomedusae (Aurelia aurita and Chrysaora spp.). These
findings suggest that external morphology and/or physiology could
explain the difference in eDNA production sources and processes among
taxa.
Cellular and molecular states of eDNA can also be closely associated
with its transport and degradation processes, consequently influencing
the spatiotemporal range of target eDNA signals (Barnes & Turner, 2016)
and even eDNA-based estimation accuracy of species abundance. Although
studies linking eDNA state to its spatiotemporal dynamics are scarce, it
has been reported that intra-cellular eDNA collected from larger size
fraction (i.e., larger eDNA particles) contained longer DNA fragments
more frequently (Jo et al., 2020a), and eDNA decay rates could be
determined by eDNA states, such as target gene (mitochondrial/nuclear)
and particle size, as well as abiotic factors, including temperature and
water chemistry (Strickler et al., 2015; Jo & Minamoto, 2021). In the
context of abundance estimation, given the rapid degradation of longer
eDNA fragments (Jo et al., 2017) and persistence of smaller-sized eDNA
particles (i.e., eDNA from smaller size fractions) in water due to the
inflow of degraded eDNA from larger to smaller fractions (Jo et al.,
2019b), biological signals from longer eDNA fragments and larger eDNA
particles (i.e., eDNA from larger size fractions) could be fresher and
spatiotemporally finer in the field, which may consequently improve the
accuracy of eDNA-based abundance estimation. Nevertheless, aside from
Stewart (2019), who reviewed how biotic factors, such as developmental
stage, life history, and species interaction might influence eDNA
production and eDNA-based abundance estimation performance, exploration
of the effects of eDNA production sources and states on estimation
accuracy has been limited.
Furthermore, the estimation accuracy of species abundance can rely on
some technical aspects in eDNA analysis. First, although the earlier
studies collected eDNA in water samples by centrifugation and
precipitation (e.g., Ficetola et al., 2008; Takahara et al., 2012),
filtration of water samples is now the most common method for collecting
aqueous eDNA (Kumar et al., 2020). Between these collection methods, the
volume of water samples (typically 15 mL in the former while hundreds to
thousands of milliliters in the latter) and the state of eDNA collected
(both membranous and dissolved DNA in the former while only membranous
DNA in the latter) could be different. Second, compared to the eDNA
concentration estimated by real-time and digital PCR, the eDNA read
number estimated by metabarcoding is expected to reflect species
abundance less precisely, given the biases introduced during PCR,
sequencing, and bioinformatics steps required high-throughput sequencing
(Lamb et al., 2019). Nevertheless, some studies have reported a positive
relationship between the relative abundance of eDNA read detectedvia high-throughput sequencing and species abundance (e.g., Evans
et al., 2016; Li et al., 2021). Third, it is unclear which metrics of
species abundance, biomass or number of individuals, exhibit a stronger
relationship with eDNA quantity. Although a precedent meta-analytic
study found no evidence that there was no significant difference in
these metrics for the relationship with eDNA quantity (Yates et al.,
2019), the study also acknowledged the need of accumulating future
research to specify the difference. As well as the production source and
state of eDNA, these technical aspects in eDNA analysis have potentials
affecting the estimation accuracy of species abundance via eDNA,
whereas these points had not been assessed so far.
As far as we know, there is no study to directly assess the importance
of eDNA production source and state, as well as technical aspects in
eDNA analysis, for the accuracy of eDNA-based abundance estimation.
However, meta-analyses, synthesizing previous findings and statistically
re-analyzing them, may shed light on the relationship between eDNA-based
estimation of species abundance and eDNA characteristics and
methodology. In this study, we investigated how different eDNA
production sources, states, and methodology influenced eDNA-based
species abundance estimation accuracy by performing meta-analyses of
eDNA studies targeting macro-organisms. We conducted a literature search
and extracted data on factors influencing eDNA characteristics and
methodology. Moreover, since it is unclear how the relationship between
species abundance and eDNA concentration differs among various natural
environments (e.g., freshwater/marine, lentic/lotic), we also assessed
the effect of target environments on eDNA-based abundance estimation
accuracy. Integrating and collating previous findings viameta-analysis will enable us to find generalizable patterns in these
relationships and to elucidate the hitherto unknown findings in the
estimation accuracy of species abundance via eDNA.