Genomics Reveals Complex Population History and

Unexpected Diversity of Eurasian Otters (Lutra lutra)

in Britain Relative to Genetic Methods
Sarah J. du Plessis ,* ,1 Mark Blaxter,2 Klaus-Peter Koepfli,3,4 Elizabeth A. Chadwick,1
and Frank Hailer * ,1
1
School of Biosciences, Cardiff University, Cardiff, UK
2
Tree of Life, Wellcome Sanger Institute, Cambridge, UK
3
Smithsonian-Mason School of Conservation, George Mason University, Front Royal, VA, USA
4
Centre for Species Survival, Smithsonian’s National Zoo and Conservation Biology Institute, Washington, DC, USA

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

*Corresponding authors: E-mails: [email protected]; [email protected].
Associate editor: Emma Teeling

Abstract
Conservation genetic analyses of many endangered species have been based on genotyping of microsatellite loci and
sequencing of short fragments of mtDNA. The increase in power and resolution afforded by whole genome ap­
proaches may challenge conclusions made on limited numbers of loci and maternally inherited haploid markers.
Here, we provide a matched comparison of whole genome sequencing versus microsatellite and control region
(CR) genotyping for Eurasian otters (Lutra lutra). Previous work identified four genetically differentiated “strong­
hold” populations of otter in Britain, derived from regional populations that survived the population crash of the
1950s–1980s. Using whole genome resequencing data from 45 samples from across the British stronghold popula­
tions, we confirmed some aspects of population structure derived from previous marker-driven studies.
Importantly, we showed that genomic signals of the population crash bottlenecks matched evidence from otter
population surveys. Unexpectedly, two strongly divergent mitochondrial lineages were identified that were un­
detectable using CR fragments, and otters in the east of England were genetically distinct and surprisingly variable.
We hypothesize that this previously unsuspected variability may derive from past releases of Eurasian otters from
other, non-British source populations in England around the time of the population bottleneck. Our work highlights
that even reasonably well-studied species may harbor genetic surprises, if studied using modern high-throughput
sequencing methods.
Key words: population genomics, bottleneck, demographic history, reintroductions, genetic tools.

Article
Introduction Shapiro 2018; Hohenlohe et al. 2021). There is a trade-off
in data acquisition between the number of individuals
Genetic and Genomic Methods that can be assayed and the number of loci at which vari­
As molecular ecology expands to include whole genome se­ ation in each individual can be measured. WGS is becoming
quencing (WGS), the congruence between genetic and increasingly affordable but is still effectively limited to rela­
genomic methods has been called into question tively small sample sizes. To maximize the potential benefit
(McCartney-Melstad et al. 2018; Zimmerman et al. 2020). from the array of methods on the genetics–genomics con­
Rather than discrete categories, genetic and genomic tinuum, an understanding of when these results are likely
methods form a spectrum across marker type, number of to be congruent, or differ, will enable the most cost-
loci, and technologies used to generate data. Generally, effective marker selection for the specific research and
methods using fewer than a thousand loci are considered conservation project (Gallego-García et al. 2021).
genetic and therefore include microsatellites and mito­ Ascertainment bias, the nonrandom analysis of loci re­
chondrial fragments that are usually generated using sulting in parameter estimate biases (Nielsen 2000), has
Sanger sequencing–based methods (Hohenlohe et al. the potential to affect both genetic and genomic methods.
2021). Methods using thousands or more loci are consid­ For instance, microsatellite selection based on a single
ered genomic, therefore including WGS and reduced re­ population will be biased toward detection of variation
presentation methods such as restriction site-associated present in that population and away from detection of
DNA sequencing (RADseq), which are produced using variation in a distinct, highly differentiated population
high-throughput sequencing technologies (Supple and (Malomane et al. 2018). This results in systematic
© The Author(s) 2023. Published by Oxford University Press on behalf of Society for Molecular Biology and Evolution.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/
licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly
cited. Open Access
Mol. Biol. Evol. 40(11):msad207 https://doi.org/10.1093/molbev/msad207 Advance Access publication September 15, 2023 1
du Plessis et al. · https://doi.org/10.1093/molbev/msad207 MBE
deviations, such as the underrepresentation of rare alleles, direct estimates of otter population sizes throughout this
which can underestimate signals of a population expan­ time period, otter hunting records provide a proxy for
sion (Nielsen 2000). Genomic methods such as RADseq population size. These indicate drastic population declines
and WGS are less susceptible to ascertainment bias, due across southwest England and less drastic declines in nor­
to the largely random choice of loci. Furthermore, if few thern England and Scotland from 1957 (Chanin and
loci show strong bias, these will have a larger effect in smal­ Jefferies 1978), highlighting a potential variation in the im­
ler, genetic data sets than in larger, genomic data sets. pact of chemical contaminants across the landscape.
Due to the varying evolutionary rates of different mar­ Restrictions and bans on the use of these chemicals in
kers, more loci or greater length of sequence are required the 1970s lead to population recovery (Walker and
to obtain comparable levels of resolution and hence stat­ Newton 1998; Mason and Macdonald 2004), as tracked
istical power. For example, microsatellites are chosen to be by national surveys beginning in 1977 across Britain. The
multiallelic with an unusually high and variable mutation percentage of sites visited showing signs of otters increased
rate (from 10−3 to 10−4 mutations per locus per gener­

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

from 6% to 59% in England, from 20% to 89% in Wales, and
ation), whereas single nucleotide polymorphisms (SNPs) from 61% to 92% in Scotland, when comparing surveys
are primarily biallelic and have a lower mutation rate of from 1977 to 2009 for England and Wales and 1977 to
10−8–10−9 per nucleotide per generation (Ellegren 2000; 2003 for Scotland. The species has broadly been acclaimed
Väli et al. 2008). Microsatellites provide an effective meth­ as a conservation success story (Macdonald 1983; Strachan
od of identifying significant differences in variability even et al. 1990; Crawford and Scholey 2010; Kean and
with low numbers of loci, and SNPs are often assessed in Chadwick 2021). However, the most recent National
the thousands to millions in genomic data sets. Survey of Wales identified signals of a recent, substantial
Microsatellite loci are assumed to be largely neutral and population decline, highlighting the importance of contin­
thus useful for assessing genetic diversity and structure, ual monitoring and the value of otters as an indicator spe­
whereas approaches such as WGS provide the opportunity cies for the ecological health of aquatic systems of Britain
to investigate the evolutionary mechanisms behind the (Kean and Chadwick 2021).
measures of genetic variation, such as inbreeding depres­ Since the 1950–1980 population crash, extensive genet­
sion, genetic basis of adaptation, and functional variation ic assessments have been conducted on L. lutra in Britain,
(Supple and Shapiro 2018; Hohenlohe et al. 2021). using both microsatellites (Dallas and Piertney 1998; Dallas
It is important to evaluate characteristics of the study et al. 1999, 2002; Hobbs et al. 2006, 2011; Stanton et al.
population when considering variation in results between 2014; Thomas et al. 2022) and mitochondrial fragments
methods. The variation found within species with large ef­ (Stanton et al. 2009). These data have been incorporated
fective population sizes (e.g., Ne > 1,000) is unlikely to be into broader European assessments using both marker
accurately quantified through analysis of a few loci. This is­ types (Cassens et al. 2000; Randi et al. 2003; Ferrando
sue is less critical in analyses of divergent, small, and inbred et al. 2004; Mucci et al. 2010). These studies illustrate the
populations (Gallego-García et al. 2021). Furthermore, the consequences of the population crash by revealing distinct
number of samples taken to represent a population will genetic population strongholds. They were also critical in
constrain the power of analyses. For example, the potential developing and sharing optimized microsatellite methods
to detect fine-scale population structuring is limited with (Hobbs et al. 2011). A culmination of this work was a
few samples and the generally small sample sizes typical of microsatellite data set spanning 21 years of the population
genomic approaches may not yield the same insights as recovery in Britain, which identified a time lag in the gen­
those made with much larger genetic sample size etic connectivity between populations (Thomas et al.
(Gallego-García et al. 2021). A clear benefit of genomic ap­ 2022).
proaches is the ability to standardize results across re­ Over 4,000 otters from across Britain, primarily victims
search groups and therefore directly compare results of road traffic accidents, have been collected and pre­
between broad populations and species. For example, served by the Cardiff University Otter Project since 1994,
using a standard reference genome and short-read WGS most of which are suitable for genomic analysis. A male
data analyzed with the same bioinformatic pipeline facili­ L. lutra from Somerset, Southwest England, was sequenced
tates direct comparison of data, which is rarely possible and assembled to chromosomal completeness by the 25
using microsatellite markers. genomes for 25 years project at the Wellcome Sanger
Institute (Mead et al. 2020), and this high-quality reference
Case Study: Eurasian Otter (Lutra lutra) genome is a strong backbone on which data can be
Eurasian otter (L. lutra) populations in Britain suffered a mapped and contextualized. The Eurasian otter in
substantial population crash between 1950 and 1980 due Britain therefore presents an ideal opportunity to directly
to environmental chemical contaminants, such as persist­ compare inferences derived from parallel data sets gener­
ent organic pollutants (Chanin and Jefferies 1978). This ated from genetic and genomic methods. As the first
pattern was broadly observed across Europe, resulting in population genomic study of the species, we addressed
the species being classified as “near threatened” on the two objectives. First, we used both WGS and an established
International Union for Conservation of Nature (IUCN) microsatellite panel to directly compare population me­
Red List (Roos et al. 2015). Although there are only limited trics derived from genomic and genetic methods for the
2
Genomics Reveals Complex Population History and Unexpected Diversity · https://doi.org/10.1093/molbev/msad207 MBE
same sample set of Eurasian otters in Britain (fig. 1). assignments of the same individuals at increasing K values,
Secondly, we used the genomic data to assess inbreeding consistent with the absence of pronounced substructuring
and historic effective population size estimates for the first within the East in the PCA (fig. 2a). No pairwise relatedness
time. We predicted that: of third degree or closer was identified among the samples
(supplementary material S2.7, Supplementary Material
1) Detected genetic variation and population structur­ online).
ing will differ between genetic and genomic methods. To probe the complex population assignments of indivi­
2) The recent, anthropogenic population bottleneck duals within the East population at increasing levels of K,
(which was not previously identified using genetic fineSTRUCTURE analyses were run on the SNP data. The
methods) will be identifiable from genomic signa­ fineSTRUCTURE coancestry matrix (supplementary
tures, such as runs of homozygosity (ROH), linkage material S2.6, Supplementary Material online) showed clear
disequilibrium (LD), and changes in effective popula­ population structuring between the East and the remaining
tion size over time.

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

populations, with less distinct population structuring be­
tween the Wales and North and Southwest England popula­
Results tions. The matrix also suggested lower levels of shared
coancestry between the East and other populations, com­
Population Structure Using Genetic and Genomic pared to within and between the non-Eastern populations.
Data In contrast to the ADMIXTURE results, fineSTRUCTURE
Principal component analysis (PCA) of SNPs from the clustered mLutLut92 (East) with samples from the North.
whole genome resequencing data (fig. 2a) was dominated
on both PC1 (12.8%) and PC2 (7.7%) by variation among Genomic Diversity within Britain
samples within the East population. Removing samples Based on whole genome SNP data, heterozygosity was
from the East population and repeating the PCA of SNP higher in the East population than in other populations
data revealed geographic clustering of samples from the (mean difference of 0.00013–0.00017 higher in the East;
remaining three populations (fig. 2a inset). In the PCA of F3,41 = 33.43, P < 0.0001; fig. 3a). The realized inbreeding
the microsatellite data from the same specimens (fig. coefficient (FROH) based on the proportion of genomes
2b), PC1 accounted for 8.6% of variation in the data and in ROH ranged from 0.43 to 0.77 and was significantly low­
PC2 8.4%, without clear separation of samples by popula­ er in the East, with no significant differences among the re­
tion, except for the Southwest England population. maining populations (F3,41 = 16.21, P = 0.0004; fig. 3a). The
Pairwise FST values among the four populations were sig­ majority of ROH fell into the shortest, and therefore oldest
nificantly higher in the microsatellite than the whole gen­ class, indicating that extensive inbreeding occurred at
ome data set, with microsatellite-derived values ranging 1,024 or more generations ago (an estimated 4,096 years).
from 0.08 to 0.28 and SNP-derived values from 0.05 to Signals of longer ROH (from 4 and 8 generations, or 16 and
0.07 and a mean difference in FST between methods of 32 years ago) were visible in a few individuals in the East
0.11 (paired t-test, t5 = 3.97, P = 0.01; fig. 2c). and scarcer in the remaining populations.
We explored likely population substructuring in whole Nucleotide diversity (π) was significantly higher across
genome SNP and microsatellite data. The cross-validation genomic windows in the East and significantly different
(CV) error of ADMIXTURE analyses on the SNP data indi­ among all population pairs except Southwest England
cated the most likely value of K to be 1; however, the delta and Wales (F3,504 = 76.61, P < 0.0001; fig. 3b). Private SNPs
K of STRUCTURE analyses of the microsatellite data indi­ showed the same trend, with more SNPs identified as pri­
cated the most likely value of K to be 3 (supplementary vate to the East (2,076,325) than were private to any other
material S2.6, Supplementary Material online). Due to an population or common across all populations (1,498,476;
increase in discordance among replicate STRUCTURE ana­ fig. 3b; full private SNPs results in supplementary material
lyses and varying individual cluster assignments of the S2.8, Supplementary Material online).
ADMIXTURE analyses at higher K values, results up to LD decay over distance (up to 100 kb) showed varying
K = 6 are presented for both methods in fig. 2d and e, trends among populations (supplementary material S2.8,
whereas the results from the full range of K values are Supplementary Material online). Southwest England con­
shown in supplementary material S2.6, Supplementary sistently showed the highest overall levels of LD and the
Material online. The STRUCTURE plot of microsatellite slowest decline with increasing chromosomal distance;
data indicated more admixture among populations than the North showed consistently lowest overall levels of
identified by ADMIXTURE on whole genome SNP data. LD, with the quickest decline over increasing chromosomal
Samples clustered broadly by geographic population in distance. Wales and East showed similar patterns of LD de­
the STRUCTURE plot at K = 4, whereas in the cay, which varied among chromosomes (supplementary
ADMIXTURE analysis at K = 4, there was substructuring material S2.8, Supplementary Material online).
in the East, and Southwest England clustered with Wales.
Southwest England and Wales were separated at higher Genetic Diversity within Britain
K values. However, substructuring in the East was inconsist­ Genetic diversity statistics based on microsatellites were
ent from K = 4 to K = 6, with varying population within the range of values reported in previous
3
du Plessis et al. · https://doi.org/10.1093/molbev/msad207 MBE

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

FIG. 1. Geographic origins of
Eurasian otter samples ana­
lyzed in this study and the ref­
erence genome.

microsatellite studies (Thomas et al. 2022) despite the the East, 11 alleles private to the North, and only 1 allele pri­
lower sample size in the present study. Observed heterozy­ vate to Southwest England. No alleles were found only in
gosity based on microsatellites was highest in the North Wales (see supplementary material S2.5, Supplementary
and East (0.64 and 0.63), followed by Wales (0.53), and Material online). Rankings for FIS/FROH showed a more not­
the lowest observed heterozygosity was identified in able difference, with whole genome SNP results ranking
Southwest England (0.48). Since these statistics are not Southwest England > Wales > North > East and microsa­
comparable across methods, we report the population tellites ranking East > North > Southwest England >
rankings and full results in supplementary material S2.5, Wales.
Supplementary Material online. When populations are
ranked from highest to lowest heterozygosity, rankings Historic Ne
varied between data sets: the whole genome SNP results Historic effective population size (Ne) was estimated using
rank East > North > Wales > Southwest England, whereas GONE for the recent past (up to 800 years ago) and
microsatellites rank North > East > Wales > Southwest pairwise sequentially Markovian coalescent (PSMC) for
England. Private alleles were identified in 12 of the 15 the ancient past (10,000 to 1,000,000 years ago) (fig. 4a;
microsatellite loci, with 10 alleles across all loci private to supplementary material S2.9, Supplementary Material
4
Genomics Reveals Complex Population History and Unexpected Diversity · https://doi.org/10.1093/molbev/msad207 MBE

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

FIG. 2. Comparison of estimates of population structuring of Eurasian otters in Britain, based on WGS versus microsatellite data. PCAs of (a)
whole genome SNPs and (b) microsatellites for the whole data set and (in the inset) for the 33 non-Eastern samples and (c) pairwise FST values
for whole genome SNPs and microsatellites. (d ) ADMIXTURE results for whole genome SNPs and (e) STRUCTURE results for microsatellites.

online). The demographic population bottleneck was ex­ Both the East and Southwest England populations showed
pected to have occurred between the 1950s and 1970s, substantial bottlenecks and recoveries between the 1950s
and this expectation corresponded very accurately to and 1970s, but the Ne of the Southwest was consistently
the declines and recoveries of Ne estimated by GONE. higher than that of the East, which declined to 3.7 from
5
du Plessis et al. · https://doi.org/10.1093/molbev/msad207 MBE

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

FIG. 3. Genomic diversity in British Eurasian otters based on SNP markers. Individual-level (a) genome-wide heterozygosity and proportion of the
genome in runs of homozygosity (FROH) as calculated by RzooROH, color-coded by number of generations ago inbreeding is estimated to have
occurred (from older runs/shorter ROH to recent/longer ROH). Population-level measures (b): nucleotide diversity (π) calculated in 20 Mb slid­
ing windows and counts of SNPs private to each population (bars), relative to the number of SNPs common to all populations (black line at
1,498,476 SNPs).

1972 to 1984 (averaged across bootstrap Ne estimates for presence in the East coinciding with the low Ne estimates.
9–12 generations ago). The decline in Ne started earlier At deeper timescales, PSMC analyses of 10,000 to 1,000,000
in Wales, during the 1800s, and showed an increase in years ago showed a decline in Ne across all populations,
the past 50 years. The North showed a gradual, continual albeit with some local variation.
decline through the past 800 years. For the most recent 50
generations, we compared Ne estimates using GONE (fig. Mitochondrial Genomes versus Control Region
4b) to the survey data recording the number of sites show­ Of the 45 British samples, whole mitochondrial genomes
ing positive signs of otters by region (fig. 4c). Visually, the for 44 were assembled, of which 41 were assembled using
trends matched reasonably well, with the differences in NOVOPlasty and 3 were assembled using MITObim
trends between regions and the extremely low proportion (table 1). From these 44 sequences, 18 unique haplotypes
6
Genomics Reveals Complex Population History and Unexpected Diversity · https://doi.org/10.1093/molbev/msad207 MBE

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

FIG. 4. Estimated effective population size through time by population of Eurasian otter in Britain based on SNP data and comparable survey data.
(a) GONE analyses were conducted on a population-level scale, with ten bootstrap repeats for each population, assuming a generation time of 4
years, conducted on samples sizes of n = 8 for Southwest England, n = 12 for Wales and East, and n = 13 for North, and the expected timing of the
bottleneck highlighted (1950–1970). (b) GONE analyses for the same period of time for which surveys were conducted (0–50 years before sam­
ples were collected in 2020). (c) Survey data as percentage of sites surveyed which showed positive signs of otters, grouped by regions reflecting
stronghold populations, with the exception of the North, which was split into regions from England and Scotland based on divergent census size
trends. For surveys which spanned more than 1 year, the results are plotted at the earliest year. All plots are presented on a log scale of Ne.

were identified across 153 segregating sites (summary sta­ segregating sites and nucleotide diversity were higher in
tistics and haplotypes given in supplementary materials S1 the East than other British populations; however, haplo­
and S2.11, Supplementary Material online, respectively). type diversity was highest in the North (supplementary
The TCS network identified 2 distinct lineages, equivalent material S2.11, Supplementary Material online).
to lineages 1 and 3 in the lineage classification of Waku All five control region (CR) haplotypes which have pre­
et al. (2016), separated by a branch representing 105 mu­ viously been identified in Britain were found in our data
tations between groups (fig. 5a). Due to the presence of set. Haplotypes Lut1, Lut3, Lut6 (Stanton et al. 2009),
both divergent lineages in the East, the number of Lut4 (Cassens et al. 2000), and Lut7 (Pountney 2008)
7
du Plessis et al. · https://doi.org/10.1093/molbev/msad207 MBE
Table 1. Eurasian Otter Sample Accessions, Locations, and Publication Source of Publicly Available SRA and Whole Mitochondrial Genome Sequences
Incorporated into Analyses.

Sample Accession Location Publication Lineage

SRA reads downloaded and assembled:

a
SRR19383068 Narvik, Norway de Ferran et al. (2022) Lineage 3
a
SRR19383067 Tyumen Oblast, Russia de Ferran et al. (2022) Lineage 3
a
SRR11679564 Denmark Margaryan et al. (2021) Lineage 3
Whole mitochondrial genome sequence downloaded:
LC049377 China Waku et al. (2016) Lineage 1
LC049378 China Waku et al. (2016) Lineage 1
LC049952 Sichuan, China Waku et al. (2016) Lineage 1
LC049953 Unknown Waku et al. (2016) Lineage 1
LC049954 Sakhalin, Russia Waku et al. (2016) Lineage 2

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

LC049955 Kanagawa, Japan Waku et al. (2016) Lineage 1
LC050126 Kochi, Japan Waku et al. (2016) L. l. nippon/alineage 5
a
LC094961 Laos Waku et al. unpublished Lineage 1
a
LR822067/NC_062277 Southwest England Mead et al. (2020) Lineage 1
a
MW316682 Kinmen, Taiwan Jang-Liaw et al. unpublished Lineage 1
a
EF672696 Korea Ki et al. (2010) Lineage 2
FJ236015/NC_011358 Korea Jang et al. (2009) Lineage 2
a
MW573979 Daejeon, South Korea Kim and Jo (2021) Lineage 2
KY117556 — Mohd Salleh et al. (2017) L. sumatrana

NOTE.—Lineage names and sample allocation as defined by Waku et al. (2016).

a
Assigned in the present study.

were found in 29, 4, 9, 1, and 1 samples, respectively. The include 40 samples from Britain (across all populations)
difference between Lut1 and Lut4 is a single base indel and the SRA-derived sequences from Denmark, Russia,
at nucleotide position 96, and therefore, these haplotypes and Norway. Lineage 1 included four samples from the
were collapsed in the haplotype network (fig. 5b). East of Britain, alongside GenBank samples from China,
Geographic distribution of these haplotypes was broadly Laos, and Japan, and the reference genome sequence,
consistent with previous findings. For example, Lut1 was from a British otter (from Somerset, closest geographically
found across all populations, whereas Lut3 was only iden­ to our Southwest England region). Although the rooted
tified in the East. Interestingly, however, Lut6, previously phylogenetic tree indicated that lineage 5/L. l. nippon split
only found in Western Britain, was identified in eight sam­ first, followed by lineage 3 and then lineages 1 and 2, vari­
ples from Wales and one sample from the East in our data ation in sequences within lineage 3 appeared to be a recent
set. Only the four samples belonging to lineage 1 corre­ diversification relative to the older branching within
sponded to CR haplotype Lut3, separated from the re­ lineages 1 and 2. These results were reflected in the higher
maining haplotypes by a single mutation. number of segregating sites and nucleotide diversity ob­
We also assembled mitochondrial genomes from three served within lineages 1 and 2, relative to lineage 3.
previously generated short-read data for European L. lutra
available in the short-read archive. These read sets were Discussion
successfully assembled using NOVOPlasty and aligned
with European and Asian L. lutra mitochondrial genome se­ Genetic versus Genomic Methods
quences from GenBank (n = 13) and the British samples Our study provides a direct comparison of genetic and gen­
generated in this work, yielding a total of 60 L. lutra omic methods in assessment of genetic diversity and popu­
sequences. We added a single hairy-nosed otter (Lutra su­ lation structure in a threatened wild carnivore, the Eurasian
matrana) as outgroup. From the 60 L. lutra mitochondrial otter. We applied genomic methods to investigate the oc­
genomes, 34 unique haplotypes were identified across 772 currence and effects of a recent, anthropogenically driven
segregating sites. Phylogenetic analysis showed separation population decline and recovery of Eurasian otters in
of samples into four main lineages, all with posterior prob­ Britain. Our concurrent analysis of microsatellite and
abilities of 1.00 (fig. 5d; sample lineages in supplementary SNPs from the same sample set highlights the complexities
material S2.11, Supplementary Material online). Three inherent in interpreting results of these approaches.
lineages were named by Waku et al. (2016) as L. l. nippon Broadly, our microsatellite and SNP data sets are not
(here referred to as lineage 5), lineage 1, and lineage 2, concordant, with substantial differences in the order of
and one new lineage was identified in this study (lineage population differentiation and population groupings. For
3). Geographic locations of these lineages are provided in example, at K = 3, the SNP data set grouped Southwest
fig. 5c, with the exception of Britain and Russia for which England and Wales, whereas the microsatellite data
multiple lineages were identified. The British samples grouped East and North England. Across all samples, the
sequenced in this study identified two divergent lineages: proportion of admixture identified is higher in the micro­
lineage 1 and lineage 3. Sequences assigned to lineage 3 satellite data set than the SNP data. However, for the
8
Genomics Reveals Complex Population History and Unexpected Diversity · https://doi.org/10.1093/molbev/msad207 MBE

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

FIG. 5. TCS networks, phylogeny, and locations of lineages of mitochondrial genome variation in Eurasian otters. (a) TCS network of 44 British
samples and the reference mitochondrial genome (LR822067.1), based on 16,365 bp mitochondrial sequence (with the repeat region removed).
(b) TCS network of the same 44 British samples and known CR haplotypes from GenBank (in black). Lut4 and mLutLut29 are collapsed into Lut1,
as they only differ by a single base indel. (c) Countries shaded to indicate where each whole mitochondrial genome lineage was identified, and
hatching indicates known, global Eurasian otter range. Japan included both lineages 1 and 5 and therefore is shaded orange for the unique lineage
5. No colored shading is shown for Russia and Britain in this diagram, due to multiple lineages identified (see supplementary material S2.11,
Supplementary Material online). (d ) Phylogenetic tree of 16,392 bp whole mitochondrial sequence (with the repeat region removed) generated
in this study (n = 44), assembled from SRA (n = 3), and downloaded from GenBank (n = 13), rooted with a hairy-nosed otter (L. sumatrana),
totaling 61 sequences. Posterior probabilities of branches between lineages are all 1. Tree shading indicates lineage 1 (red), 2 (green), 3 (blue), and
5/L. l. nippon (orange). Full phylogeny and posterior probabilities are given in supplementary material S2.11, Supplementary Material online.

purpose of population assignment based on a suitable K pairwise FST estimates. These results are not unexpected,
value, there is also some concordance across data sets, as both similarities (Zimmerman et al. 2020) and differ­
with the exception of some samples between the East ences (Lah et al. 2016) such as these have been identified
and North in the microsatellite data. To summarize, simi­ in past studies of comparable genetic and genomic data
larities in the ADMIXTURE/STRUCTURE comparison sets.
broke down at higher values of K and in more complex Both the number of loci assessed and ascertainment
scenarios such as the substructuring within the East. bias are likely to be contributing to the differences be­
Similarly, SNPs showed more population structuring in tween the genetic and genomic approaches. Specifically,
PCA, and microsatellites showed significantly higher the higher resolution of data captured by almost 9 million
9
du Plessis et al. · https://doi.org/10.1093/molbev/msad207 MBE
biallelic SNPs, compared with 15 multiallelic microsatel­ absolute values of metrics calculated based on microsatel­
lites, is likely to be critical, as in several prior studies (Lah lites should be interpreted with caution, with a focus on
et al. 2016; Natesh et al. 2017; Lavretsky et al. 2019; relative values (such as a decrease in FST over time) likely
Gallego-García et al. 2021). Although the microsatellites being much more informative. Importantly, the scale of
identified the broad patterns of population structuring ac­ the effect of ascertainment bias varies depending on the
curately, they do not hold the same power to detect more specific loci and samples, and therefore, comparisons be­
fine-grained population distinction (higher values of K) tween studies based on different loci, populations, sam­
and more complex scenarios, when compared with the ples, or species would be inappropriate.
genomic results. Since the complexity of any given study Using whole mitochondrial genomes, we were able to
system is typically unknown, it is important to recognize identify two divergent lineages within Britain, which
that conclusions based on genetic methods may change were not previously identified based on CR studies
following the application of genomic data. In particular, (Stanton et al. 2009), and were not possible to identify

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

the measures which showed the most significant differ­ using CR fragments for our samples. Based on these results,
ences between methods (e.g., pairwise FST) are likely to we strongly recommend that analyses using short mito­
be inflated due to ascertainment bias of the microsatellite chondrial fragments should be upgraded to whole mito­
loci. Specifically, these microsatellite loci were identified chondrial genome analyses where possible, to avoid
and selected based on the variation they identified in misleading inferences.
Scotland and Wales (both sampled in this study), and
therefore, they are likely to represent only a subset of Demographic Analyses and Consequences
the variation present across populations in Britain Using whole genome sequence data, it was possible to de­
(Dallas and Piertney 1998; Dallas et al. 1999; Hobbs et al. tect and analyze historic trends in effective population size
2006). Although ascertainment bias is less commonly iden­ and access methods not available for genetic data sets.
tified as a source of variation between genetic and genom­ Most notably, the recent anthropogenic population
ic comparisons, Fischer et al. (2017) showed significantly bottleneck between the 1950s and 1970s was clearly evi­
larger estimates of pairwise FST using microsatellites than denced (using GONE), with a sharp decrease and subse­
SNPs, which they attributed to the ability of high-coverage quent increase in Ne occurring concurrently across
WGS in identifying rare variants, which reduce overall FST. southern England (based on samples distributed across
Similarly, Cairns et al. (2023) identified higher proportions southwest, central, and eastern England). In contrast,
of admixture among dogs and dingoes (Canis spp.) using Wales showed a decline beginning around 1830s, coincid­
microsatellites relative to SNPs, where the microsatellites ing with the “improvement of the sporting gun with per­
had been selected based on allele frequencies identified cussion detonation,” and reduced otter hunt results in
in dogs. Taken together, these findings highlight potential North Wales (Jefferies 1989). Ne estimates from Wales in­
impacts of microsatellite ascertainment bias, leading to in­ creased from the 1950s (as other British populations
flated admixture proportions (Cairns et al. 2023). were beginning to decline), which was unexpected but po­
Our utilized microsatellite panel does not contain any tentially due to the impact and changes of otter hunting
loci pairs with significant LD signals, and all loci are at least around this time, despite legal protection only being ob­
1.34 Mbp apart in the genome. Consequently, linkage ap­ tained in 1978 in Wales and England (Jefferies 1989). The
pears unlikely to have had a large effect on our microsat­ North showed an overall, gradual decline from around
ellite results. Linkage signals are higher for our analyzed 1200 to the present day, with no change in this trend
SNPs, due to their larger numbers (9 million) and greater through the 1950s–1970s bottleneck detected in southern
proximity in the genome. Therefore, although linkage oc­ England, confirming suspicions that this larger, more rural
curs in both data sets, it is likely to be affecting the genom­ population was less affected by chemical pollution
ic results more significantly than the genetic results in this (Chanin and Jefferies 1978). These findings are broadly in
study. At this point, it becomes difficult to disentangle the keeping with otter hunting records which, for example, in­
effects of number of loci, ascertainment bias, and linkage, dicate a decline in success rate of otter hunts from 1957
as using too few loci or using nonrandomly selected loci across southwest England, a lesser decline in northern
could both underrepresent rare alleles and therefore lead England and Scotland, and no evidence of a decline in
to inflated FST estimates, among other effects, further sup­ north Wales (Chanin and Jefferies 1978). The accompany­
porting the crucial role of genomic data. ing PSMC analyses also show a more long-term decline
We are not suggesting microsatellite markers are redun­ from 1,000,000 to 10,000 years ago across all populations.
dant, and we emphasize the important role they continue We hypothesized that genomic signatures of a popula­
to play as an affordable marker system for cluster assign­ tion bottleneck would be evident. Breeding between rela­
ment and other scenarios. Equally, long-term genetic tives leads to homozygous regions of the genome
data sets, such as Thomas et al. (2022), highlight the im­ separately inherited from a recent shared ancestor, known
portance of the continuity of markers to assess temporal as ROH. Longer ROH indicate recent inbreeding, and
trends in genetic variation, for example, during population shorter runs indicate older inbreeding, which have been
recovery and expansion. However, our analyses suggest broken up by recombination (Druet and Gautier 2017).
that due to numbers of loci and ascertainment bias, the We predicted that the British bottleneck occurred
10
Genomics Reveals Complex Population History and Unexpected Diversity · https://doi.org/10.1093/molbev/msad207 MBE
between the 1950s and 1970s and would therefore be consistent with a relatively complex demographic history
identifiable as ROH of a length corresponding to inbreed­ in the East population, including ADMIXTURE results,
ing occurring 9–17 generations before these samples were which indicate inconsistent population clustering, and
collected (based on a generation time of 4 years); however, fineSTRUCTURE results, which indicate that the East
ROH of this length were infrequent. ROH have been iden­ have lower shared coancestry with remaining populations
tified in populations suffering severe reductions in effect­ than they show when compared with one another but
ive population size for prolonged periods of time, such also show inconsistent coancestry between individuals
as the Florida panther, Puma concolor (Saremi et al. from the East. For example, although some pairs within
2019). We do not have census data spanning the bottle­ the East show very high shared coancestry, others show
neck, but due to the banning of some chemical contami­ very low shared coancestry, whereas all individuals in
nants in the 1970s (Walker and Newton 1998), we Wales show a very similar shared coancestry between pairs.
predicted that the bottleneck was unlikely to have per­ Lastly, the divergent mitochondrial lineage identified using

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

sisted for more than 20 years. However, the GONE analyses genomic methods was found only in samples from the East
indicated that depending on the population, Ne dropped and the reference genome (from Somerset, equidistant to
for between 30 and 45 years before increasing (with the ex­ southwest and eastern samples used in this study). In line
ception of Wales and the North). Despite the length and with previous studies (Thomas et al. 2022), SNP-based mea­
severity of this bottleneck, especially in the East (from sures of pairwise FST were not larger for pairs including the
the lowest Ne estimate of around 100 to around 3), it East, suggesting that although the East contains lots of un­
does not appear to have led to a consistent burden of ique variation, it is not more distantly related to remaining
ROH across multiple individuals. Conversely, we observe populations. Since pairwise FST is a relative measure of gen­
many, short ROH across all samples. Theory would suggest etic variation within a population relative to total variation,
this indicates very old inbreeding followed by generations it is possible that these results reflect the high variation ob­
of recombination, potentially from a bottleneck that is served within the East, when compared with total variation.
older than we are able to estimate using RzooROH or The results observed in the East do not align with the
due to background relatedness (Ceballos et al. 2018). We simple scenario of a population bottleneck and recovery,
found no evidence for bottlenecks from 10,000 to and instead, we propose alternative hypotheses to explain
1,000,000 years ago (PSMC analyses) nor from 200 to 800 our findings. In the first National Otter Survey of England,
years ago (GONE analyses); however, we have not assessed in 1977–1979, of 623 sites surveyed in East Anglia, only 20
evidence for the period 800 to 10,000 years ago due to were positive, illustrating how close the population was to
method limitations. It is unexpected, but reassuring, that local extinction (Lenton et al. 1980). Captive-bred otters
we do not see long ROH in these individuals despite the were released, 13 in the south and 81 in the east of
severe recent bottleneck in the East. However, the evi­ England, as a part of a broader reintroduction program
dence for extensive historic inbreeding and its legacy, the of over 180 otters across Britain (Green 1997). Following
generally high realized inbreeding coefficient (FROH) ob­ these releases, the second National Survey in 1984–1986
served in the modern populations, provides cause for con­ found 8 positive out of 725 surveyed sites in East Anglia,
cern with respect to the species’ long-term viability, 5 of which were from released otters (Strachan et al.
particularly given the current context of small populations 1990). One explanation for genetic distinctiveness of otters
and anthropogenic threats for a near threatened species in the East might be that their small effective population
(Roos et al. 2015; Reid et al. 2019). size led to strong genetic drift. However, the high genetic
diversity (both overall variability and the high private
Distinct Signatures Detected in the Eastern SNP count) observed in the genomic data in the East ren­
Population ders this explanation unlikely. Another explanation is that
Unexpected diversity identified in the East using genomic the East gene pool contains ancestry from Eurasian otters
methods was not previously identified in studies using gen­ of non-British origin, explaining both the unusually high
etic methods (Stanton et al. 2014). Estimates of variation proportion of unique variation and the divergent mito­
(e.g., from PCA) were dominated by variation among sam­ chondrial lineage. A study by the European Association
ples within the East, and these results were matched by of Zoos and Aquaria (EAZA) using microsatellites identi­
higher heterozygosity, nucleotide diversity, and private fied two main genetic “lines” of Eurasian otters in captivity,
SNP counts, alongside lower inbreeding coefficients relative known as A- and B-lines (E. Rey pers. comm.), where anec­
to other British populations. The GONE analyses indicate dotal evidence suggests the B-line otters were bred with
that the population from the East went through the the Asian subspecies L. l. barang found in Sumatra,
most severe bottleneck of all British populations, with Ne Thailand, and Vietnam (J. Palmer pers. comm. (Hung and
as low as 3.7, making the high variation in this region par­ Law 2016)). An organization involved in the reintroduction
ticularly surprising. Although the ROH analyses identified program was separately breeding Eurasian otters in captiv­
some recent inbreeding, it was not consistent across indivi­ ity using two founders from Thailand, thought at the time
duals from the East, indicating a more complex demo­ to be L. l. barang (J. Palmer pers. comm.). Although stud­
graphic history than hypothesized from a simple book records show no evidence of crossing between otters
anthropogenic bottleneck. Other metrics are also of Thai origin and British L. lutra (J. Palmer pers. comm.),
11
du Plessis et al. · https://doi.org/10.1093/molbev/msad207 MBE
escapes or unknown mixing between pens cannot be ex­ genomic analyses of British otters. However, based on
cluded, however unlikely (P. Chanin pers. comm.). our findings of ascertainment bias, we would recommend
Although there are no whole mitochondrial genome se­ a SNP array based on SNPs identified from a range-wide
quences available from Eurasian otters from Thailand, we sampling of the species.
have included a sequence from the neighboring Laos in Here, we provide the first clear evidence of a population
this study, which groups with samples from China and bottleneck in Eurasian otters in southern England, high­
the east of England. Therefore, when combined with our lighting an important contrast to the history of popula­
genomic results, these findings indicate the possibility of tions in northern England, Scotland, and Wales. We
individuals of Asian origin (likely Thailand) being either ac­ provide clear evidence of previously unexpected and un­
cidentally bred, released, or escaping into the south and usual signatures in the east of England and assign these
east of England. Otter releases in this region occurred be­ to a mitochondrial lineage only found, as yet, in Asia.
tween 1983 and 1996, coinciding with the increase in our Both of these findings have only been possible through

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

estimate of effective population size for the East. We there­ the application of genomic methods, enabled by the prior
fore suggest that the introduction of a few Asian-origin or publication of an extremely high-quality reference genome
admixed otters to a very small existing population had a (Mead et al. 2020), illustrating the importance of WGS for
large impact on the genetic identity of the population, conservation of this and other species.
leading to the high proportion of unique genomic vari­
ation and divergent mitochondrial lineages. The captive
B-line otters are no longer being bred in captivity Methods
(J. Palmer pers. comm.); however, there are indications Sample Collection and DNA Extraction
that some of their descendants have been released in
Samples from across the four known stronghold popula­
Europe (Hájková et al. 2007). Our results show that the
tions (the North of England and Scotland, Southwest
consequences of such introductions have left genomic sig­
England, Wales, and central and eastern England) were se­
natures across the east of England and possibly beyond.
lected based on location from the Cardiff University Otter
To provide further evidence describing the genetic his­
Project (CUOP) archive. The 45 individuals analyzed in this
tory of otters in Britain, we suggest two main paths of future
study were collected between 2016 and 2020 and com­
work. Firstly, to further investigate the history of otters in
prised 35 adult males, 5 females (1–2 per stronghold),
the East of Britain, it would be beneficial to sequence histor­
and 5 male juveniles (see supplementary material S1,
ical samples from the region (taken before the population
Supplementary Material online). Genomic DNA was ex­
bottleneck), as well as B-line captive-bred otters, to
tracted from muscle tissue using a salt extraction protocol
compare these to the population we observe presently
(Rivero et al. 2006) and stored in TLE buffer (1 M Tris, 0.5 M
and assess the likelihood of a replacement event occurring
ethylenediaminetetraacetic acid [EDTA], pH 8). DNA
through the reintroductions (Strachan et al. 1990).
quantity and quality were assessed using gel electrophor­
Secondly, we recommend sampling the range of the
esis, and final concentrations ranged from 7.3 to
Eurasian otter broadly, including historic samples, to try
117.4 ng/µL.
to identify any similarities between a potential source
population for British reintroductions, such as L. l. barang,
L. l. chinensis (found in Thailand), and other Asian lineages, Whole Genome Sequencing
alongside any other populations where B-line otters may Sequencing was conducted as part of the Darwin Tree of
have been reintroduced. This would also enable a thorough Life (DToL) program (Blaxter et al. 2022) by the
genomic assessment of subspecies classifications, clarifying Wellcome Sanger Institute (Hinxton, UK). Whole genomes
the existing confusion around Asian subspecies, such as of 45 Eurasian otters were sequenced on the Illumina
L. l. barang and L. l. chinensis. We note that population de­ NovaSeq platform, with 150 bp paired-end reads. The
clines and local extinctions across the species’ range mean standard DToL pipelines were used to quality filter, trim,
that it is possible that the source population could be and map reads to the Eurasian otter reference genome as­
missed in this sampling, highlighting the importance of his­ sembly (Mead et al. 2020). Variants were called using
toric samples (Yoxon and Yoxon 2019). DeepVariant (Poplin et al. 2018), and joint variant calling
was performed using GLNexus (Lin et al. 2018).
Conservation Implications Unplaced scaffolds (n = 23) make up 0.34% of the total
Our genetic to genomic comparison highlights the value reference genome assembly and were discarded from all
of genomic methods in conservation to avoid misinterpre­ further analyses. Only biallelic SNPs from chromosome
tations of potentially biased, low-resolution markers. 1 to 18 (i.e., excluding X, Y, and mitochondrial scaffolds)
However, this does not make the Eurasian otter microsat­ were used in population genomic analyses, unless other­
ellite panel or previous studies based on this redundant. wise stated. Vcf subsetting, viewing, variant counting, in­
Rather, this study identifies the limits of the interpreta­ dexing, and converting between file formats was
tions of these results when compared with genomic conducted using BCFTools v1.14 (Danecek et al. 2021),
data. A SNP array for capture-based SNP genotyping could and data handling and visualization was conducted in R
be designed based on our findings, to provide cheaper version 4.2.0 (R Core Team 2019), using RStudio version
12
Genomics Reveals Complex Population History and Unexpected Diversity · https://doi.org/10.1093/molbev/msad207 MBE
2022.2.3.492 (RStudio Team 2022) and the tidyverse 2.9.4 (Kamvar et al. 2014). Inbreeding coefficient, FIS, and
packages (Wickham et al. 2019) unless otherwise stated. pairwise FST values were calculated using hierfstat version
0.5-10 (Goudet 2005). PopGenReport version 3.0.7 (Gruber
Microsatellite Genotyping and Adamack 2015) was used to calculate mean and
standard errors of allelic richness by population.
Genomic locations of microsatellite loci commonly used in
STRUCTURE v2.3.4 (Pritchard et al. 2000) was used to as­
studies of Eurasian otters were identified in the reference
sess population structuring and admixture proportions.
genome using Blast (see supplementary material S2.4,
Each run consisted of a burn-in of 100,000, followed by
Supplementary Material online, for methods and results).
900,000 recorded iterations, run from K of 1 to 10, each re­
Microsatellite genotyping was carried out for all 45 sam­
peated 10 times, using the admixture model (without sam­
ples using 15 loci amplified in 3 multiplex PCR reactions
pling locations as priors) and assuming allele frequencies
(supplementary material S1, Supplementary Material on­
to be correlated. The ten repeated runs for each K value
line). PCR conditions were as in Hobbs et al. (2006)

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

were then assessed for alternative solutions and combined
(supplementary material S2.3, Supplementary Material on­
using the CLUMPAK web server (Raymond and Rousset
line). Fragment analysis on an ABI3730 capillary sequencer
1995; Dieringer and Schlötterer 2003; Rousset 2008;
(Applied Biosystems) was conducted at MRC PPU DNA
Kopelman et al. 2015).
Sequencing and Services, Dundee, Scotland, using
Rox400HD as a size marker. Microsatellite alleles were
scored using the Microsatellite plugin in Geneious Prime Genomic Diversity
v2022.0.2 (https://www.geneious.com). Four samples Nucleotide diversity was calculated using VCFtools, in
were replicated for all 3 multiplexes to ensure consistency, 20 Mb windows across all samples within each geographic
and 38 sample/multiplex combinations were rerun. One population, alongside individual heterozygosity. SNPs pri­
locus (Lut733) for one sample (mLutLut40) could not be vate to a population, or found across any combination
genotyped reliably, providing inconsistent results across of populations, were counted using the vcf-compare tool
four replicates, resulting in an overall success rate of within VCFtools. Due to the uneven sample sizes of the
99.85%. geographic populations, eight samples were randomly se­
lected from each population to use for this analysis.
Significant differences in genetic diversity measures among
Population Structure populations were assessed using ANOVA conducted in
For the whole genome SNP data, PCAs were conducted in R. LD was calculated in VCFtools, for each geographic
PLINK v1.9 (Purcell et al. 2007). Weir and Cockerham’s population, by chromosome, between all pairs of SNPs
(1984) FST was calculated using VCFtools v0.1.16 up to 100 kbp apart and averaged for each distance be­
(Danecek et al. 2011). A paired t-test was used to compare tween SNPs before plotting.
the differences in FST values between WGS and microsatel­
lite data sets. STRUCTURE could not be run on this gen­
omic data set due to computational limitations of the Runs of Homozygosity
STRUCTURE algorithm when handling genomic-scale Due to the ability of a single incorrectly called variant to
data sets, and therefore, ADMIXTURE was selected as break a ROH, variants were filtered separately for these
the most algorithmically similar software. ADMIXTURE analyses (supplementary material S2.12, Supplementary
v1.3.0 (Alexander and Lange 2011) was run on the full Material online). For ROH identification, we used
SNP data set for K values of 1 to 8, and CV error was RZooRoH v0.3.1 (Druet and Gautier 2017; Bertrand et al.
used to compare all K values. Details and results of the 2019) to model homozygous-by-descent (HBD) segments
fineSTRUCTURE v4.1.1 (Lawson et al. 2012) and related­ of the genome, where the length corresponds to the num­
ness analyses are in supplementary materials S2.6 and S2. ber of generations since the common ancestor of the
7, Supplementary Material online. haplotype. Following model selection using the Bayesian
The microsatellite data was converted to Genepop for­ Information Criterion, we used HBD segments from 4, 8,
mat using MSA v4.05 (Dieringer and Schlötterer 2003), and 16, 32, 64, 128, 256, 512, and 1,024 generations ago, along­
when assessed using Genepop v4.6 (Raymond and Rousset side identifying non-HBD segments.
1995; Rousset 2008), no significant LD was detected in any
pairwise comparisons among loci, even when compared Historic Ne
across populations and after applying sequential We used GONE (Santiago et al. 2020) to estimate recent
Bonferroni correction (correcting the nominal P = 0.05 effective population size (Ne) for each population (see
for multiple testing). All 15 loci were therefore kept for supplementary material S2.9, Supplementary Material on­
downstream analyses. Calculation of means and standard line, for full method). GONE calculates LD between pairs of
errors by population, of observed and expected microsat­ SNPs over a range of recombination rates and finds the ser­
ellite heterozygosity (HO, HE) and allele number (A), along­ ies of Ne that best explains the observed LD spectrum
side PCA was conducted using the R package adegenet (Santiago et al. 2020). We ran all simulations for 2,000 gen­
version 2.1.5 (Jombart 2008). Private microsatellite alleles erations calculated in 400 bins but only present results for
were determined using the R package poppr version the most recent 200 generations. We repeated each run 10
13
du Plessis et al. · https://doi.org/10.1093/molbev/msad207 MBE
times, with each run taking a new subsample of 100,000 analyses, alongside a hairy-nosed otter (L. sumatrana,
SNPs (comparable to bootstrapping), to incorporate vari­ KY117556) sequence as an outgroup (table 1).
ance in Ne estimates across runs. Recent estimates of Ne Geneious Prime was used to align sequences using the
using GONE were compared with survey data of MUSCLE algorithm (Edgar 2004), and due to uncertain re­
England, Wales, and Scotland from 1977 to 2010; peat numbers surrounding the tandem repeat, this region
details of these methods are in supplementary material was removed (positions 16,050–16,202 on the reference
S2.10, Supplementary Material online, and data in scaffold), leaving a total alignment length of 16,365 bp.
supplementary material S1, Supplementary Material on­ To contextualize our samples with prior studies, 255 bp
line. PSMC inference (Li and Durbin 2011) was also used of the CR were extracted and compared with previously
to estimate older effective population size (Ne) changes published CR haplotypes. PopArt v1.7 (Leigh and Bryant
(supplementary material S2.9, Supplementary Material on­ 2015) was used to produce statistical parsimony networks
line). A generation time of 4 years was used for GONE, based on the TCS algorithm (Clement et al. 2002) and to

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

PSMC, and ROH analyses (see supplementary material identify the number of mutation steps between haplotypes.
S2.9, Supplementary Material online). After aligning to previously published L. lutra and out­
group whole mitochondrial genome sequences, the repeat
Mitochondrial Genome Analyses region was again removed (positions 16,035–16,289 bp
We assembled the mitochondrial genome of each sample relative to the reference scaffold), leaving a total alignment
independently and combined these with previously pub­ length of 16,392 bp. The R packages pegas (Paradis 2010)
lished data and sequences (see supplementary material and ape (Paradis and Schliep 2019) were used to calculate
S2.11, Supplementary Material online, for full methods). summary statistics (haplotype richness, haplotype diver­
NOVOPlasty v4.3.1 (Dierckxsens et al. 2017) was used to sity, and nucleotide diversity, π). IQ-TREE (Nguyen et al.
assemble the mitochondrial genomes from the adapter- 2015) ModelFinder (Kalyaanamoorthy et al. 2017) was
trimmed reads using the 16,536 bp reference genome used to identify the best fitting model based on the
mitochondrial scaffold (LR822067.1) as the seed. Bayesian Information Criterion, which was the three-
NOVOPlasty failed to assemble four samples, of which substitution type model with unequal, empirical base
MITObim v1.9.1 (Hahn et al. 2013) was used to assemble frequencies (Kimura 1981), allowing for a proportion of
three samples. invariable sites (‘K3Pu + F + I’). A consensus tree was con­
We aimed to exclude the possibility that our results are structed based on 1,000 bootstrap replicates (Hoang et al.
assembled from nuclear sequences of mitochondrial origin 2018) and visualized using FigTree v1.4.4 (http://tree.bio.
(NUMTs) rather than true mitochondrial sequences, ed.ac.uk/software/figtree/).
therefore biasing phylogenetic interpretation (Lucas et al.
2022). Despite no prior evidence of NUMTs in any muste­
lid species, NumtFinder (Edwards 2021) identified reason­ Supplementary Material
ably long putative NUMTs (from 5 to 9 kbp) across Supplementary data are available at Molecular Biology and
chromosomal and unplaced scaffolds in the Eurasian otter Evolution online.
reference genome (mLutLut1.2). However, it was not pos­
sible to create the combination of variants identified in ei­
ther whole mitochondrial genome lineage from any Acknowledgments
combination of NUMTs we identified. When all reads S.J.d.P. was supported by the UK Natural Environment
from one sample were mapped only to a mitochondrial Research Council through the GW4+ Doctoral Training
genome sequence and sites differentiating between the Partnership (award reference: 2194854) and the collabor­
two divergent lineages in this study were interrogated, ation with K.-P.K. enabled by the Global Wales
we found that across ten randomly chosen sites, on aver­ International Mobility Fund (award: UNIW/RMF-CU/16).
age <1% of reads did not support the base assembled in We would like to thank all present and past Cardiff
the mitochondrial genome. Therefore, although we have University Otter Project staff and volunteers, and we thank
identified the presence of NUMTs in L. lutra, we do not be­ Natural Resources Wales, the Environment Agency, and
lieve they are contributing to the divergent mitochondrial the International Otter Survival Fund, for the collection,
lineages identified in this study and instead conclude that storage, and transport of otters from across Britain.
we have assembled the true mitochondrial lineages (see Otter 3329 was collected fresh via the rapid response of
supplementary material S2.11.1, Supplementary Material Somerset Otter Group and used to generate the reference
online). genome. Sequencing was conducted by the DToL program
Alongside the British samples, paired-end short-read at the Wellcome Sanger Institute, for which we are hugely
data of samples from Russia (n = 1), Norway (n = 1), and grateful. The bioinformatic analyses were performed on
Denmark (n = 1) were downloaded from the Sequence the Cardiff School of Biosciences’ Biocomputing Hub
Read Archive (SRA) and assembled using NOVOPlasty as HPC/Cloud infrastructure, and we are immensely thankful
described above (table 1). Previously published mitochon­ to Dr Ian Merrick and Andy Ells for their support. We also
drial sequences available from GenBank (n = 13, including thank Dr Matthieu Muffato (Wellcome Sanger) for bio­
the reference genome) were also incorporated into the informatic support and Dr Paul Chanin and Jason Palmer
14
Genomics Reveals Complex Population History and Unexpected Diversity · https://doi.org/10.1093/molbev/msad207 MBE
for feedback on the manuscript. In addition, we thank Dr diversity in the Eurasian otter, Lutra lutra, in Scotland.
Nia Thomas for providing the collated survey data and the Evidence from microsatellite polymorphism. Biol J Linn Soc.
68(1–2):73–86.
suggestion of, and extensive discussions around, the GONE Dallas JF, Marshall F, Piertney SB, Bacon PJ, Racey PA. 2002. Spatially re­
analyses. We also acknowledge the insightful comments by stricted gene flow and reduced microsatellite polymorphism in the
the anonymous reviewers. Eurasian otter Lutra lutra in Britain. Conserv Genet. 3(1):15–28.
Dallas JF, Piertney SB. 1998. Microsatellite primers for the Eurasian
otter. Mol Ecol. 7(9):1248–1251.
Author Contributions Danecek P, Auton A, Abecasis G, Albers CA, Banks E, DePristo MA,
Handsaker RE, Lunter G, Marth GT, Sherry ST, et al. 2011. The vari­
F.H., E.A.C., and S.J.d.P. designed the research, M.B. coordi­ ant call format and VCFtools. Bioinformatics. 27(15):2156–2158.
nated sequencing, and S.J.d.P. conducted analyses and Danecek P, Bonfield JK, Liddle J, Marshall J, Ohan V, Pollard MO,
wrote the paper with guidance from F.H. and K.-P.K.; all co­ Whitwham A, Keane T, McCarthy SA, Davies RM, et al. 2021.
authors contributed to editing the manuscript. Twelve years of SAMtools and BCFtools. GigaScience 10:2.
de Ferran V, Figueiró HV, de Jesus Trindade F, Smith O, Sinding

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

M-HS, Trinca CS, Lazzari GZ, Veron G, Vianna JA, Barbanera F,
et al. 2022. Phylogenomics of the world’s otters. Curr Biol.
Data availability 32(16):3650–3658.e4.
All raw data have been submitted to the International Dierckxsens N, Mardulyn P, Smits G. 2017. NOVOPlasty: de novo as­
Nucleotide Sequence Database Collaboration (INSDC) sembly of organelle genomes from whole genome data. Nucleic
Acids Res. 45(4):e18.
archives and are available through the DToL Data Portal Dieringer D, Schlötterer C. 2003. Microsatellite analyser (MSA): a
(https://portal.darwintreeoflife.org/data/root/details/Lutra% platform independent analysis tool for large microsatellite data
20lutra). Individual microsatellite genotype data are available sets. Mol Ecol Notes. 3(1):167–169.
in supplementary material S1, Supplementary Material online. Druet T, Gautier M. 2017. A model-based approach to characterize
Newly generated unique haplotype data are deposited to individual inbreeding at both global and local genomic scales.
Mol Ecol. 26(20):5820–5841.
NCBI Nucleotide Database (OR633269-86 and BK064833- Edgar RC. 2004. MUSCLE: a multiple sequence alignment method
BK064835), and all whole mitochondrial genome sequences with reduced time and space complexity. BMC Bioinformatics.
analysed in this study can be found on GitHub (https:// 5(1):113.
github.com/sduplessis1/EurasianOtter_PopGen) alongside Edwards RJ. 2021. NUMTFinder: nuclear mitochondrial fragment
the bioinformatic code used for all analyses. (NUMT) search tool. [https://github.com/slimsuite/numtfinder].
v0.1.0.
Ellegren H. 2000. Microsatellite mutations in the germline:: implica­
tions for evolutionary inference. Trends Genet. 16(12):551–558.
References Ferrando A, Ponsa M, Marmi J, Domingo-Roura X. 2004. Eurasian ot­
Alexander DH, Lange K. 2011. Enhancements to the ADMIXTURE al­ ters, Lutra lutra, have a dominant mtDNA haplotype from the
gorithm for individual ancestry estimation. BMC Bioinformatics. Iberian Peninsula to Scandinavia. J Hered. 95(5):430–435.
12:246. Fischer MC, Rellstab C, Leuzinger M, Roumet M, Gugerli F, Shimizu
Bertrand AR, Kadri NK, Flori L, Gautier M, Druet T. 2019. RZooROH: KK, Holderegger R, Widmer A. 2017. Estimating genomic diver­
an R package to characterize individual genomic autozygosity sity and population differentiation—an empirical comparison
and identify homozygous-by-descent segments. Methods Ecol of microsatellite and SNP variation in Arabidopsis halleri. BMC
Evol. 10(6):860–866. Genomics. 18(1):1–15.
Blaxter ML, Mieszkowska N, Di Palma F, Holland P, Durbin R, Gallego-García N, Caballero S, Shaffer HB. 2021. Are genomic up­
Richards T, Berriman M, Kersey P, Hollingsworth P, Wilson W, dates of well-studied species worth the investment for conserva­
et al. 2022. Sequence locally, think globally: the Darwin Tree of tion? A case study of the critically endangered Magdalena river
Life Project. Proc Natl Acad Sci U S A. 119(4):7. turtle. J Hered. 112(7):575–589.
Cairns KM, Crowther MS, Parker HG, Ostrander EA, Letnic M. 2023. Goudet J. 2005. Hierfstat, a package for R to compute and test hier­
Genome-wide variant analyses reveal new patterns of admixture archical F-statistics. Mol Ecol Notes. 5(1):184–186.
and population structure in Australian dingoes. Mol Ecol. 32(15): Green R. 1997. Reintroduction of otters: a successful conservation
4133–4150. strategy. IUCN Otter Specialist Group Bulletin. 14(2):62–74.
Cassens I, Tiedemann R, Suchentrunk F, Hartl GB. 2000. Gruber B, Adamack AT. 2015. Landgenreport: a new R function to
Mitochondrial DNA variation in the European otter (Lutra lutra) simplify landscape genetic analysis using resistance surface layers.
and the use of spatial autocorrelation analysis in conservation. J Mol Ecol Resour. 15(5):1172–1178.
Hered. 91(1):31–35. Hahn C, Bachmann L, Chevreux B. 2013. Reconstructing mitochon­
Ceballos FC, Joshi PK, Clark DW, Ramsay M, Wilson JF. 2018. Runs of drial genomes directly from genomic next-generation sequen­
homozygosity: windows into population history and trait archi­ cing reads—a baiting and iterative mapping approach. Nucleic
tecture. Nat Rev Genet. 19(4):220–234. Acids Res. 41(13):e129–e129.
Chanin PRF, Jefferies DJ. 1978. The decline of the otter Lutra lutra Hájková P, Pertoldi C, Zemanová B, Roche K, Hájek B, Bryja J, Zima J.
L. in Britain: an analysis of hunting records and discussion of 2007. Genetic structure and evidence for recent population de­
causes. Biol J Linn Soc. 10(3):305–328. cline in Eurasian otter populations in the Czech and Slovak
Clement M, Snell Q, Walker P, Posada D, Crandall K2002. TCS: esti­ Republics: implications for conservation. J Zool. 272(1):1–9.
mating gene genealogies. Parallel and distributed processing sym­ Hoang DT, Chernomor O, Von Haeseler A, Minh BQ, Vinh LS. 2018.
posium, international. Washington (DC): IEEE Computer Society. UFBoot2: improving the ultrafast bootstrap approximation. Mol
p. 1–7. Biol Evol. 35(2):518–522.
Crawford A, Scholey G. 2010. Fifth otter survey of England 2009– Hobbs GI, Chadwick EA, Bruford MW, Slater FM. 2011. Bayesian clus­
2010. Environment agency technical report. tering techniques and progressive partitioning to identify popu­
Dallas JF, Bacon PJ, Carss DN, Conroy JWH, Green R, Jefferies DJ, lation structuring within a recovering otter population in the
Kruuk H, Marshall F, Piertney SB, Racey PA. 1999. Genetic UK. J Appl Ecol. 48(5):1206–1217.

15
du Plessis et al. · https://doi.org/10.1093/molbev/msad207 MBE
Hobbs GI, Chadwick EA, Slater FM, Bruford MW. 2006. Landscape MTP, et al. 2021. Mitochondrial genomes of Danish vertebrate
genetics applied to a recovering otter (Lutra lutra) population species generated for the national DNA reference database,
in the UK: preliminary results and potential methodologies. DNAmark. eDNA 3(2):472–480.
Hystrix It J Mamm. 17(1):47–63. Mason CF, Macdonald SM. 2004. Growth in otter (Lutra lutra) popu­
Hohenlohe PA, Funk WC, Rajora OP. 2021. Population genomics for lations in the UK as shown by long-term monitoring. AMBIO
wildlife conservation and management. Mol Ecol. 30(1):62–82. 33(3):148–152.
Hung N, Law CJ. 2016. Lutra lutra (Carnivora: Mustelidae). Mamm McCartney-Melstad E, Vu JK, Shaffer HB. 2018. Genomic data re­
Species. 48(940):109–122. cover previously undetectable fragmentation effects in an en­
Jang KH, Ryu SH, Hwang UW. 2009. Mitochondrial genome of the dangered amphibian. Mol Ecol. 27(22):4430–4443.
Eurasian otter Lutra lutra (Mammalia, Carnivora, Mustelidae). Mead D, Hailer F, Chadwick E, Portela Miguez R, Smith M, Corton C,
Genes Genomics. 31(1):19–27. Oliver K, Skelton J, Betteridge E, Doulcan JD, et al. 2020. The gen­
Jefferies D. 1989. The changing otter population of Britain 1700– ome sequence of the Eurasian river otter, Lutra lutra Linnaeus
1989. Biol J Linn Soc. 38(1):61–69. 1758. Wellcome Open Res. 5:33.
Jombart T. 2008. adegenet: a R package for the multivariate analysis Mohd Salleh F, Ramos-Madrigal J, Peñaloza F, Liu S, Mikkel-Holger SS,
of genetic markers. Bioinformatics. 24(11):1403–1405. Riddhi PP, Martins R, Lenz D, Fickel J, Roos C, et al. 2017. An ex­

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

Kalyaanamoorthy S, Minh BQ, Wong TK, Von Haeseler A, Jermiin LS. panded mammal mitogenome dataset from Southeast Asia.
2017. ModelFinder: fast model selection for accurate phylogenet­ GigaScience 6(8):gix053.
ic estimates. Nat Methods. 14(6):587–589. Mucci N, Arrendal J, Ansorge H, Bailey M, Bodner M, Delibes M,
Kamvar ZN, Tabima JF, Grünwald NJ. 2014. Poppr: an R package for Ferrando A, Fournier P, Fournier C, Godoy JA, et al. 2010. Genetic
genetic analysis of populations with clonal, partially clonal, an­ diversity and landscape genetic structure of otter (Lutra lutra) po­
d/or sexual reproduction. PeerJ. 2:e281. pulations in Europe. Conserv Genet. 11(2):583–599.
Kean EF, Chadwick EA. 2021. Sixth otter survey of Wales 2015–18. Natesh M, Atla G, Nigam P, Jhala YV, Zachariah A, Borthakur U,
NRW report no: 519, NRW. Ramakrishnan U. 2017. Conservation priorities for endangered
Ki JS, Hwang DS, Park TJ, Han SH, Lee JS. 2010. A comparative analysis of Indian tigers through a genomic lens. Sci Rep. 7(1):1–11.
the complete mitochondrial genome of the Eurasian otter Lutra lu­ Nguyen L-T, Schmidt HA, Von Haeseler A, Minh BQ. 2015. IQ-TREE: a
tra (Carnivora; Mustelidae). Mol Biol Rep. 37(4):1943–1955. fast and effective stochastic algorithm for estimating maximum-
Kim H-N, Jo Y-S. 2021. Complete mitochondrial genome sequencing likelihood phylogenies. Mol Biol Evol. 32(1):268–274.
of Lutra lutra (Linnaeus, 1758) (Carnivora: Mustelidae) and its Nielsen R. 2000. Estimation of population parameters and recombin­
phylogenetic status in Mustelidae. Mitochondrial DNA Part B. ation rates from single nucleotide polymorphisms. Genetics
6(7):2066–2068. 154(2):931–942.
Kimura M. 1981. Estimation of evolutionary distances between hom­ Paradis E. 2010. pegas: an R package for population genetics with an
ologous nucleotide sequences. Proc Natl Acad Sci U S A. 78(1): integrated–modular approach. Bioinformatics. 26(3):419–420.
454–458. Paradis E, Schliep K. 2019. ape 5.0: an environment for modern phy­
Kopelman NM, Mayzel J, Jakobsson M, Rosenberg NA, Mayrose I. logenetics and evolutionary analyses in R. Bioinformatics. 35(3):
2015. Clumpak: a program for identifying clustering modes 526–528.
and packaging population structure inferences across K. Mol Poplin R, Chang P-C, Alexander D, Schwartz S, Colthurst T, Ku A,
Ecol Resour. 15(5):1179–1191. Newburger D, Dijamco J, Nguyen N, Afshar PT, et al. 2018. A uni­
Lah L, Trense D, Benke H, Berggren P, Gunnlaugsson P, Lockyer C, versal SNP and small-indel variant caller using deep neural net­
Öztürk A, Öztürk B, Pawliczka I, Roos A, et al. 2016. Spatially ex­ works. Nat Biotechnol. 36(10):983–987.
plicit analysis of genome-wide SNPs detects subtle population Pountney A. 2008. Analysis of the population genetics and polybro­
structure in a mobile marine mammal, the harbor porpoise. minated diphenyl ether (PBDE) burdens of otters in England and
PLoS One 11(10):e0162792. Wales: with case studies of populations in South West England.
Lavretsky P, DaCosta JM, Sorenson MD, McCracken KG, Peters JL. 2019. Exeter (UK): University of Exeter.
ddRAD-seq data reveal significant genome-wide population struc­ Pritchard JK, Stephens M, Donnelly P. 2000. Inference of population
ture and divergent genomic regions that distinguish the mallard structure using multilocus genotype data. Genetics 155(2):
and close relatives in North America. Mol Ecol. 28(10):2594–2609. 945–959.
Lawson DJ, Hellenthal G, Myers S, Falush D. 2012. Inference of popu­ Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender D,
lation structure using dense haplotype data. PLoS Genet. 8(1): Maller J, Sklar P, de Bakker PW, Daly MJ, et al. 2007. PLINK: a tool
e1002453. set for whole-genome association and population-based linkage
Leigh JW, Bryant D. 2015. POPART: full-feature software for haplo­ analyses. Am J Hum Genet. 81(3):559–575.
type network construction. Methods Ecol Evol. 6(9):1110–1116. Randi E, Davoli F, Pierpaoli M, Pertoldi C, Madsen AB, Loeschcke V.
Lenton EJ, Chanin PRF, Jefferies DJ. 1980. Otter survey of England 2003. Genetic structure in otter (Lutra lutra) populations in
1977–79. Nature conservancy council, 1–75. Europe: implications for conservation. Anim Conserv. 6:93–100.
Li H, Durbin R. 2011. Inference of human population history from in­ Raymond M, Rousset F. 1995. GENEPOP (version 1.2): population
dividual whole-genome sequences. Nature 475(7357):493–496. genetics software for exact tests and ecumenicism. J Hered.
Lin MF, Rodeh O, Penn J, Bai X, Reid JG, Krasheninina O, Salerno WJ. 86(3):248–249.
2018. GLnexus: joint variant calling for large cohort sequencing. R Core Team. 2019. R: A language and environment for statistical
bioRxiv:343970. computing. Vienna, Austria: R Foundation for Statistical
Lucas T, Vincent B, Eric P. 2022. Translocation of mitochondrial DNA Computing.
into the nuclear genome blurs phylogeographic and conserva­ Reid AJ, Carlson AK, Creed IF, Eliason EJ, Gell PA, Johnson PTJ, Kidd
tion genetic studies in seabirds. R Soc Open Sci. 9(6):211888. KA, MacCormack TJ, Olden JD, O SJ, et al. 2019. Emerging threats
Macdonald S. 1983. The status of the otter (Lutra lutra) in the British and persistent conservation challenges for freshwater biodiver­
Isles. Mammal Rev. 13(1):11–23. sity. Biol Rev Camb Philos Soc. 94(3):849–873.
Malomane DK, Reimer C, Weigend S, Weigend A, Sharifi AR, Rivero ERC, Neves AC, Silva-Valenzuela MG, Sousa SOM, Nunes FD.
Simianer H. 2018. Efficiency of different strategies to mitigate as­ 2006. Simple salting-out method for DNA extraction from
certainment bias when using SNP panels in diversity studies. formalin-fixed, paraffin-embedded tissues. Pathol Res Pract.
BMC Genomics. 19:16. 202(7):523–529.
Margaryan A, Noer CL, Richter SR, Restrup ME, Bülow-Hansen JL, Roos A, Loy A, de Silva P, Hajkova P, Zemanová B. 2015. Lutra lutra. The
Leerhøi F, Langkjær EMR, Gopalakrishnan S, Carøe C, Gilbert IUCN red list of threatened species 2015: e. T12419A21935287.

16
Genomics Reveals Complex Population History and Unexpected Diversity · https://doi.org/10.1093/molbev/msad207 MBE
Rousset F. 2008. genepop’007: a complete re-implementation of the Thomas NE, Hailer F, Bruford MW, Chadwick EA. 2022.
genepop software for Windows and Linux. Mol Ecol Resour. 8(1): Country-wide genetic monitoring over 21 years reveals lag in
103–106. genetic recovery despite spatial connectivity in an expanding
RStudio Team. 2022. RStudio: integrated development environment carnivore (Eurasian otter, Lutra lutra) population. Evol Appl.
for R. Boston (MA): R Studio. 15(12):2125–2141.
Santiago E, Novo I, Pardiñas AF, Saura M, Wang JL, Caballero A. Väli Ü, Einarsson A, Waits L, Ellegren H. 2008. To what extent do
2020. Recent demographic history inferred by high-resolution microsatellite markers reflect genome-wide genetic diversity in
analysis of linkage disequilibrium. Mol Biol Evol. 37(12): natural populations? Mol Ecol. 17(17):3808–3817.
3642–3653. Waku D, Segawa T, Yonezawa T, Akiyoshi A, Ishige T, Ueda M, Ogawa
Saremi NF, Supple MA, Byrne A, Cahill JA, Coutinho LL, Dalén L, H, Sasaki H, Ando M, Kohno N, et al. 2016. Evaluating the phylo­
Figueiró HV, Johnson WE, Milne HJ, O’Brien SJ, et al. 2019. genetic status of the extinct Japanese otter on the basis of mito­
Puma genomes from North and South America provide insights chondrial genome analysis. PLoS One 11(3):e0149341.
into the genomic consequences of inbreeding. Nat Commun. 10: Walker CH, Newton I. 1998. Effects of cyclodiene insecticides on the
4769. doi: 10.1038/s41467-019-12741-1 sparrowhawk (Accipiter nisus) in Britain—a reappraisal of the
Stanton DWG, Hobbs GI, Chadwick EA, Slater FM, Bruford MW. evidence. Ecotoxicology. 7(4):185–189.

Downloaded from https://academic.oup.com/mbe/article/40/11/msad207/7275014 by guest on 27 February 2024

2009. Mitochondrial genetic diversity and structure of the Weir BS, Cockerham CC. 1984. Estimating F-statistics for the analysis
European otter (Lutra lutra) in Britain. Conservation Genetics of population-structure. Evolution 38(6):1358–1370.
10(3):733–737. Wickham H, Averick M, Bryan J, Chang W, McGowan L, François R,
Stanton DWG, Hobbs GI, McCafferty DJ, Chadwick EA, Philbey AW, Grolemund G, Hayes A, Henry L, Hester J, et al. 2019. Welcome to
Saccheri IJ, Slater FM, Bruford MW. 2014. Contrasting genetic the Tidyverse. J Open Source Softw. 4(43):1686.
structure of the Eurasian otter (Lutra lutra) across a latitudinal Yoxon P, Yoxon B. 2019. Eurasian otter (Lutra lutra): a review of the
divide. J Mammal. 95(4):814–823. current world status. IOSF 5:53–73.
Strachan R, Birks JDS, Chanin PRF, Jefferies DJ. 1990. Otter survey of Zimmerman SJ, Aldridge CL, Oyler-McCance SJ. 2020. An empirical
England 1984–1986. Nature conservancy council, Peterborough. comparison of population genetic analyses using microsatellite
Supple MA, Shapiro B. 2018. Conservation of biodiversity in the gen­ and SNP data for a species of conservation concern. BMC
omics era. Genome Biol. 19(1):1–12. Genomics. 21(1):1–16.

17