Introduction
Landscape genetics is an integrated field aiming to understand the
relationship between landscape features and microevolutionary processes
that generate local genetic differences (Manel et al. 2003). This field
has developed not only as a basic discipline but also as an applied
science, because knowledge of local genetic structure is useful for
strategic conservation planning through the identification of dispersal
barriers or corridors (Sommer et al. 2013; Bowman et al. 2016).
Landscape genetics is becoming increasingly important under ongoing
climate change and habitat loss (Manel and Holderegger 2013; Nakajima et
al. 2023).
In freshwater ecosystems, environmental DNA (eDNA) metabarcoding is
rapidly developing and becoming widespread as a cost-efficient and
non-invasive tool for acquiring species information (Ruppert et al.
2019; Doi and Nakamura 2023). In recent years, eDNA metabarcoding has
gradually gained attention not only for species detection but also for
population-based analysis using intraspecific variation (Reviewed in
Adams et al. 2019; Sigsgaard et al. 2020; Andres et al. 2023b; Couton et
al. 2023). Previous studies drawing population-based inferences, such as
by examining genetic diversity and differentiation, reported that
despite some limitations such as the lack of individual information or
the difficulty of distinguishing false positives and negatives from
correct data, eDNA-based analysis has high applicability, given the
sampling effort of traditional population studies (Tsuji et al. 2020a;
Adams et al. 2023). However, most population-based studies using eDNA to
date have been experimental or, when conducted in the field, have
focused on intrapopulation genetic diversity, phylogenetic
relationships, or fragmentation among populations at large scales (e.g.,
between watersheds or dams) (Turon et al. 2020; Snyder and Stepien 2020;
Weitemier et al. 2021; Tsuji et al. 2023). Notably, there is a lack of
research specifically focused on landscape genetics and its metrics
dealing with local genetic variation. As landscape genetics typically
requires a great effort to sample a large number of individuals, the
application of eDNA has the potential to significantly simplify
landscape genetics research.
The analysis of landscape genetics is essentially different from that of
phylogeography (Wang 2010). In inferring recent gene flow and its
limiting factors, the key data is the proportion of a given genotype in
the gene pool within each population (gene frequency), not the ancestry
or phylogenetic relationships between genotypes/individuals (Hudson et
al. 1992; Bohonak and Roderick 2001; Bohonak and Vandergast 2011).
Although some analytical methods require information on individuals
(e.g., sibship analysis or assignment tests), the frequency-based
statistics of genetic differentiation that are most often the focus when
evaluating gene flow do not require information on individuals.
Furthermore, since false positives and negatives usually display low
abundance, they should only have a minor effect on the calculation of
these frequency-based estimates (Couton et al. 2023). Consequently, the
question of whether landscape genetic statistics can be calculated would
come down to whether gene frequencies within each population reflect
actual frequencies. In eDNA metabarcoding, gene frequencies can be
obtained as relative read counts instead of numbers of individuals (or
numbers of genomes for non-haploids). Despite differences in the nature
of the obtained data, previous studies performed in tanks or even in the
field have shown a good congruence in gene frequencies (typically as
haplotype frequencies) between eDNA- and tissue-based approaches
(Sigsgaard et al. 2016; Marshall and Stepien 2019; Andres et al. 2021,
2023a; Couton et al. 2023; Wakimura et al. 2023). Therefore, landscape
genetics analyses using eDNA are qualitatively considered feasible.
However, it remains unclear whether the statistics of genetic
differentiation calculated from eDNA samples in the field are sufficient
to withstand landscape genetic analysis. At the same time, analytical
treatments that have been claimed to be effective in previous
population-based eDNA studies, such as the removal of low-frequency
reads (Tsuji et al. 2023) or the conversion of data to semi-quantitative
rankings (Turon et al. 2020), would also need to be investigated; it is
unclear whether they are also effective in landscape genetics.
As a model case, we targeted the fluvial sculpin Cottus nozawaein the upper watershed of the Sorachi River in Japan, for which detailed
landscape genetics studies had been previously conducted. In this study,
we performed an eDNA metabarcoding analysis of the mitochondrial DNA
(mtDNA) D-loop region of this species and compared the results with
inferences based on traditional tissue-based approaches. The aims of
this study are (i) to clarify whether eDNA-based local genetic structure
is consistent with tissue-based inferences, (ii) to reveal the extent to
which the statistics of genetic differentiation calculated by eDNA are
consistent with those obtained from tissue samples, as well as to what
extent the analytical treatments of eDNA datasets change the results,
and (iii) to discuss the applicability of eDNA to landscape genetics.