Harmoinen et al. BMC Genomics
(2021) 22:473
https://doi.org/10.1186/s12864-021-07761-5
RESEARCH ARTICLE
Open Access
Reliable wolf-dog hybrid detection in
Europe using a reduced SNP panel
developed for non-invasively collected
samples
Jenni Harmoinen1* , Alina von Thaden2,3, Jouni Aspi1, Laura Kvist1, Berardino Cocchiararo2,4, Anne Jarausch2,3,
Andrea Gazzola5, Teodora Sin5,6, Hannes Lohi7,8,9, Marjo K. Hytönen7,8,9, Ilpo Kojola10, Astrid Vik Stronen11,12,
Romolo Caniglia13, Federica Mattucci13, Marco Galaverni14, Raquel Godinho15,16, Aritz Ruiz-González13,17,
Ettore Randi18,19, Violeta Muñoz-Fuentes2,20† and Carsten Nowak2,4†
Abstract
Background: Understanding the processes that lead to hybridization of wolves and dogs is of scientific and
management importance, particularly over large geographical scales, as wolves can disperse great distances.
However, a method to efficiently detect hybrids in routine wolf monitoring is lacking. Microsatellites offer only
limited resolution due to the low number of markers showing distinctive allele frequencies between wolves and
dogs. Moreover, calibration across laboratories is time-consuming and costly. In this study, we selected a panel of
96 ancestry informative markers for wolves and dogs, derived from the Illumina CanineHD Whole-Genome
BeadChip (174 K). We designed very short amplicons for genotyping on a microfluidic array, thus making the
method suitable also for non-invasively collected samples.
Results: Genotypes based on 93 SNPs from wolves sampled throughout Europe, purebred and non-pedigree dogs,
and suspected hybrids showed that the new panel accurately identifies parental individuals, first-generation hybrids
and first-generation backcrosses to wolves, while second- and third-generation backcrosses to wolves were
identified as advanced hybrids in almost all cases. Our results support the hybrid identity of suspect individuals and
the non-hybrid status of individuals regarded as wolves. We also show the adequacy of these markers to assess
hybridization at a European-wide scale and the importance of including samples from reference populations.
Conclusions: We showed that the proposed SNP panel is an efficient tool for detecting hybrids up to the thirdgeneration backcrosses to wolves across Europe. Notably, the proposed genotyping method is suitable for a variety
of samples, including non-invasive and museum samples, making this panel useful for wolf-dog hybrid assessments
and wolf monitoring at both continental and different temporal scales.
Keywords: Canis lupus, Canis lupus familiaris, Hybridization, SNP genotyping, Non-invasive sampling, Museum samples
* Correspondence: jenni.harmoinen@oulu.fi
†
Violeta Muñoz-Fuentes and Carsten Nowak shared senior authorship.
1
Ecology and Genetics Research Unit, University of Oulu, Oulu, Finland
Full list of author information is available at the end of the article
© The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this article are included in the article's Creative Commons
licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons
licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain
permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the
data made available in this article, unless otherwise stated in a credit line to the data.
Harmoinen et al. BMC Genomics
(2021) 22:473
Background
Gray wolves (Canis lupus) are currently expanding to
areas in Europe from which they had been temporarily
absent [1]. This increase in population size and range is
due to effective legal protection measures, reforestation,
expansion of wild ungulate populations, and increased
public awareness. During the last three decades, wolves
have increased in numbers in several regions in Europe,
including Fennoscandia (e.g. Finland, Sweden), the Alps
(e.g. France, Italy, Switzerland), Central Europe (e.g.
Czech Republic, Germany, western Poland) and the
northern part of the Iberian Peninsula [2, 3]. In many of
these regions, a wealth of genetic data on wolf dispersal
has been collected over the years to track the
recolonization process (e.g. [4–9]).
Analyses based on genetic markers, such as microsatellites and mitochondrial sequences, have greatly improved our knowledge of wolves, including estimates of
pack structure, population censuses and effective population sizes, and inference of the population of origin for
migrating individuals, among others (see [2]). Further,
microsatellites markers, either solely or in combination
with other markers, have been used to assess the admixture of wolves and domestic dogs (C. l. familiaris);
reported rates of admixed animals in local wolf populations range between 0 and 10% (e.g., [10–14], but see for
instance [15] for locally higher admixture rates). However, identification of wolf-dog hybrids based on microsatellite data is far from trivial, due to the low number
of alleles with distinctive frequencies between wolves
and dogs, the rather limited number of loci used in
many studies, and the fact that results strongly depend
on reference samples and the extent of population substructure in the dataset [14, 16–18]. Moreover, the fact
that most laboratories have relied on different panels of
microsatellite markers has hampered the comparability
of data on wolf-dog admixture across populations, limiting our knowledge on the extent of hybridization [2].
Genome-wide approaches have allowed previously unattainable resolution in wolf-dog hybrid identification,
such as later-generation hybrids and the differentiation
of ancient and recent hybridization events [19–21].
Analyses have confirmed the genetic separation of
wolves and dogs, but also found strong support for widespread existence of historic introgression of dog DNA in
virtually all wolf populations across Europe, Asia and
North America [20, 22, 23]. These results have unveiled
a complex evolutionary history of wolves, in which
hybridization occurring at multiple time scales nevertheless resulted in wolves maintaining their genetic distinctiveness from dogs. While such genome-wide approaches
importantly contributed to our knowledge on wolves,
their application in routine wolf monitoring for wildlife
management purposes is unpractical due to high costs,
Page 2 of 15
extensive analysis procedures and the requirement for
high-quality DNA samples [24]. Genetic wolf monitoring, however, often relies on the analysis of numerous
non-invasively collected samples, in which DNA is often
in low-quantity and has low-quality [25].
Here we describe the development of a Single Nucleotide Polymorphism (SNP) panel selected for maximum
discrimination power between European wolves and
dogs that allows for the reliable identification of pure
and admixed individuals. The method relies on the
utilization of a microfluidic array designed to simultaneously genotype 96 SNPs from 96 samples, which we
have optimized for samples with low DNA quality and
quantity. The method works reliably with all sample
types commonly collected in wolf monitoring, including
scats or saliva traces from wolf kills and, notably, also
museum samples. The results are readily comparable
across different laboratories, making this method suitable to comprehensively assess hybridization of wolves
and dogs at both local and continental scales.
Results
Assay performance
Genotyping success with the selected panel was high
across samples and markers. Only 2.7% of the samples
failed in all reactions (n = 14) and hence were discarded
from further analyses. The average genotyping success
rate was 0.97. As expected, genotyping success was the
highest for concentrated buccal swabs (1.00) and tissue
samples (0.99), while it was only slightly lower (0.93–
0.97) for the other sample types, including museum
samples (Table 1).
Genotyping consistency was also generally high. When
genotypes of high-quality samples (tissue) were compared with non-invasively collected samples from the
same individual, we detected only one individual with
one allele in the non-invasive sample that was not found
in the tissue sample (0.04% rate), as well as one missing
allele in three different non-invasive samples (2.88%
rate), while the missing data was obtained for 0–12 loci
per sample (2.42% rate) (Table S1). When comparing
results from 22 tissue samples with the Illumina
CanineHD chip results and assuming that the Illumina genotype was the correct one, only one allelic
discrepancy was found, namely a missing allele in the
Fluidigm genotype (0.49% rate) (Table S2).
Cross-species amplification testing resulted in valid genotypes only for other Canidae species (Table S3).
Samples from golden jackals produced genotypes with
0.97–0.99 genotyping success rates and red foxes (three
out of four) genotypes with 0.77, 0.78 and 0.85 call rates.
No successful amplifications were observed for the case
of the tested prey species for wolves (roe and red deer,
Harmoinen et al. BMC Genomics
(2021) 22:473
Page 3 of 15
Table 1 Genotyping success rates (proportion of successfully scored loci over the 93 genotyped SNP loci) for different samples
types. “Removed samples” were not included in the calculations due to genotyping failure for all markers
Sample type
Samples (n)
Removed samples (n)
Genotyping success rate (%)
Tissue
149
1
99
Concentrated buccal swab
28
0
100
Saliva swab
13
1
97
Hair
10
0
95
Scat
63
2
93
Urine
4
0
97
Blood
3
0
96
Museum samples
40
6
97
wild boar, goat, and sheep), nor for humans or carnivores that are not members of Canidae.
Allele frequencies in wolves and dogs using the selected
SNP panel
FST calculated for each of the 93 SNPs in our panel indicated high discriminatory power between wolves and
dogs (FST = 0.40–0.88; average 0.70). All markers were
polymorphic in dogs, with allele frequencies > 0.10, except for one (BICF2P263751, allele frequency = 0.04),
and most markers had one allele with frequency 0.7–0.8.
Wolves, on the contrary, had 18 markers with a fixed allele (all populations considered) and 77 markers had one
allele that was rare (frequency < 0.1). For all markers, the
most frequent allele in one species was the least frequent
in the other (or absent, in the case of wolves; Fig. 1).
Population differentiation and admixture analysis
Using this final SNP panel, wolves (n = 288) and dogs
(n = 300; excluding wolf-dog breeds, n = 14) were significantly differentiated (FST = 0.72, p < 0.05), and so were
wolf-dog breeds (n = 14) and dogs (n = 300; FST = 0.20,
p < 0.05). Different wolf populations also showed significant differentiation. The divergence was highest between
wolves from Italy and other European populations, with
FST = 0.17–0.28 (p < 0.05; Table S4). Wolves from the
Iberian Peninsula showed lower divergence from wolves
from Central and Eastern Europe (FST = 0.07–0.17, p <
0.05) and there was very low divergence between
remaining wolf populations (FST = 0.03–0.11).
A PCA analysis (Fig. 2) of multilocus genotypes based
on the selected 93 markers reflected substantial differentiation between wolves and dogs and showed more
genetic diversity in dogs than in wolves. Individuals
identified as suspected hybrids were either placed in an
approximately equidistant position between the wolf and
dog clusters or closer to the wolf cluster. Wolves formed
one tight cluster, including both contemporary and museum samples, as well as those from the animal parks.
Wolves sampled in Italy and the seven immigrants from
the Alpine population that were sampled across Germany
clustered together and only partially overlapped with the
remaining wolves (shown more clearly in Figure S1, PCA
for only wolves). Golden jackals clustered closely with
Fig. 1 Allele frequencies for the 93 selected SNPs in wolves and dogs. High discriminating power is due diverging allele frequencies in the wolf
and dog groups, accompanied by the presence of private alleles for dogs
Harmoinen et al. BMC Genomics
(2021) 22:473
Page 4 of 15
Fig. 2 Principal component analysis (PCA) based on 93 SNPs selected to maximize discriminatory power between wolves and dogs. Wolves are
color-coded based on sampling locations, except seven immigrant wolves from the Alpine population sampled in Germany that were colorcoded as wolves from Italy (in agreement with previous microsatellite and haplotype data, see text). Purebred dogs were sampled in Finland and
non-pedigree dogs in Germany and Romania. Saarloos Wolfdogs and Czechoslovakian Wolfdogs were sampled in Finland and Germany.
Suspected wolf-dog hybrids were identified based on previous microsatellite analysis and ancillary evidence (see text). Foxes and golden jackals
were included to assess cross-species amplification
wolves, while foxes were slightly separated. The PC1-axis
discriminated wild canids from dogs, while the PC2-axis
explained some of the variation found in wolves and dogs.
Wolf-dog breeds were located close to the dog cluster, but
were closer to the wolf cluster than other dog breeds. A
similar pattern was observed for Siberian Huskies and
Alaskan Malamutes, with the PC2-axis separating the artic
breeds from the wolf-dog breeds.
Clustering analysis implemented in STRUCTURE [26]
assigned wolves and dogs to two distinct clusters. When
all the wolves and dogs were analyzed together, wolves
had individual assignment values qw > 0.93 and except
for two individuals from Germany (one an immigrant
from the Alpine population), one from Italy and three
from the Iberian Peninsula, all assignment values were
qw > 0.97 (Table S5). Suspected hybrids had assignment
values (qw = 0.52–0.92; Table 2), in agreement with previous knowledge (Table S6). Dogs were preferentially
assigned to the other cluster, showing higher variation in
assignment values (qd = 0.59–1.00). Out of the dog
breeds, Siberian Huskies and Alaskan Malamutes had
the lowest assignment values (qd = 0.59–0.73). Similarly,
other dog breeds with roots in Siberia (East Siberian
Laika, West Siberian Laika, Russo-European Laika and
Samoyed) had somewhat lower assignment values (on
average qd = 0.86). Individuals from the wolf-dog breeds
(Saarloos Wolfdog and Czechoslovakian Wolfdog) had a
wider range of assignment values (qd = 0.64–0.99), with
an average assignment value qd = 0.73. The remaining
purebred dogs and non-pedigree dogs had assignment
values qd = 0.84–1.00.
When performing the analysis for wolf samples only,
the most likely number of populations was K = 2 (Figure
S2), which separated wolves from Italy from the
remaining wolves (Figure S3). A less supported K = 3
assigned Iberian wolf samples to another cluster (Figure
S4). Because differentiation between populations was significant, we also performed separate runs with K = 2 for
wolves from Central and Eastern Europe (including
Finland, Russia, Romania and Germany), Italy and the
Iberian Peninsula, with dogs. These analyses assigned all
the wolves to one cluster qw > 0.97 (Table S5). Despite
somewhat higher assignment values of some of the
wolves to that cluster, the assignments were similar than
in the run including all the wolf samples.
NEWHYBRIDS [27] analyses were run four times
(with four different prior combinations based on the two
available priors, Jeffreys and Uniform) for all individuals
Harmoinen et al. BMC Genomics
(2021) 22:473
Page 5 of 15
Table 2 Results from NEWHYBRIDS and STRUCTURE analyses for suspected hybrids of wolves and dogs. Analyses were run with the
four possible prior combinations (see main text). The range of results from different runs is indicated. Assignment values based on
STRUCTURE qw values were obtained for K = 2
NEWHYBRIDS
STRUCTURE
Origin
ID
Assigned Category
qi
Wolf qw
Germany
GW01xf
F1
1.00
0.56
GW02xm
F1
1.00
0.54
GW03xm
F1
1.00
0.54
Romania
RO022m
BC2w
0.98–0.99
0.85
Czech Republic
GW05xf
F1
1.00
0.52
Finland
CL134
F1
1.00
0.55
CL370
F1
1.00
0.52
CL309
F2
1.00
0.63
CL307
BC1w
0.81–0.98
0.76
CL308
BC2w
0.98–0.99
0.83
CL419
BC2w
0.74–0.92
0.89
CL420
BC2w/BC3w
0.59–0.64/0.81–0.84
0.92
together, without prior assumption of parental populations. All assumed wolf individuals, including museum
samples from Finland and wolves from animal parks
(Table S7; Table S8), were categorized as wolves (qi >
0.87 using Uniform priors and qi > 0.93 when using Jeffreys priors for theta, qi = 1 for 283–284 individuals depending on the priors). The 12 suspected wolf-dog
hybrids were assigned to different hybrid categories in
NEWHYBRIDS (Table 2), in agreement with our field
observations (Table S6). As for the dogs, most of the
purebred individuals, except for wolfdogs, were classified
as dogs (224–228 out of 264 individuals, depending on
the priors used). The individuals that were not classified
as dogs but rather as hybrids (F2, BC1d or BC2d) were
mostly from breeds with Siberian roots (n = 27–28) or
wolf-dog breeds (n = 13), which also had the lowest
STRUCTURE assignment values among dogs. Among
non-pedigree dogs, 29–30 out of 36 were assigned as
dogs, while the rest were assigned as BC2d or were not
clearly assigned to any category (posterior probability <
0.5 to several categories). All the samples from golden
jackals and red foxes were categorized as wolves in
NEWHYBRIDS and assigned to the wolf cluster in
STRUCTURE, with assignments to wolves qw = 0.95–
0.98 (Table S9).
For testing purposes, and because there was significant
pairwise genetic differentiation between different populations, we also performed three separate runs for different sample sets (wolves from Central and Eastern
Europe and dogs, wolves from the Alpine population
and dogs, and wolves from the Iberian Peninsula and
dogs). The categorizations of individuals were very similar, but the assignment values of wolves to wolf cluster
were higher (Table S5), as expected when dataset is
more homogenous. Hybrids were assigned to the same
hybrid category, with almost identical assignment values.
Assignment accuracy of simulated hybrids
When we analyzed simulated hybrids between wolves
from Central and Eastern Europe and dogs, the STRUCT
URE assignment distributions for wolves, simulated firstgeneration hybrids and first-generation backcrosses to
wolves showed no overlap (F1 qw = 0.46–0.60, BC1w qw =
0.68–0.84, wolves qw = 0.97–1.00) (Fig. 3), while there was
some degree of overlap for later-generation hybrid classes
(BC2w qw = 0.79–0.95 and BC3w qw = 0.87–0.99). When
analyzed with NEWHYBRIDS, the number of correct assignments to the corresponding hybrid class was very
high, even for third-generation backcrosses to wolf (89–
92%) (Table 3). The highest accuracy in the correct assignment of wolf backcrosses was achieved using Jeffreys
priors. The accuracy to distinguish a simulated hybrid
from a pure individual by adding up the individual assignments of all hybrid categories was 100% for all wolf hybrid
categories except BC3w (96–99%). Due to the larger variation found in dogs with these markers, the accuracy of
categorizing dog backcrosses to the correct hybrid class
dropped from 86 to 87% for BC1d to 76–77% for BC2d,
and was zero for BC3d. The assignment accuracies were
similar for wolves from Italy or the Iberian Peninsula
(Table S10; Table S11).
Harmoinen et al. BMC Genomics
(2021) 22:473
Page 6 of 15
Fig. 3 Individual assignment values to belong to the wolf cluster (qw) for wolves from Central and Eastern Europe (n = 162), dogs (n = 300) and
simulated hybrids from each of the eight simulated genealogical classes (n = 100 per class) using STRUCTURE with K = 2. Means and quartiles are
highlighted, while whiskers illustrate the range of values with outliers (circles)
Discussion
Discriminating power of the selected SNP panel
We developed a 96 SNP panel from which 93 SNPs were
finally selected based on performance (three SNPs were
dropped as they had low genotyping success rate, < 0.7
across samples). The 93 selected SNPs allowed for reliable discrimination of wolves, dogs and their hybrids.
Table 3 Assignment accuracy of simulated hybrid individuals
between dogs and wolves from Central and Eastern Europe
(Finland, Russia, Germany and Romania) from eight different
hybrid classes to the correct category (> 0.5) or to any hybrid
category (sum of assignments to hybrid categories > 0.7) based
on results from NEWHYBRIDS runs with all the four possible
prior combinations (see main text). Range of results from
different runs is indicated
Hybrid Category n
Correct
Assignments
(%) (qi > 0.5)
Assigned to
Hybrid Categories (%)
(qi > 0.7)
F1
100 100
100
F2
100 100
100
BC1w
100 99
100
BC2w
100 81–82
100
BC3w
100 89–92
96–99
BC1d
100 86–87
100
BC2d
100 76–77
79–80
BC3d
100 0
19
This high discriminating power is due diverging allele
frequencies in the wolf and dog groups, accompanied
by the presence of private alleles for dogs. For all
loci, alle frequencies were > 0.69 for one of the
groups. While this panel was chosen to maximize the
differentiation between wolves and dogs, significant
differentiation between the wolf populations was detected. However, panels designed specifically to study
population differentiation are available and better
suited for this purpose (e.g. Illumina CanineHD chip,
Affymetrix Canine SNP array or specifically designed
SNP chips).
The fact that golden jackals and foxes had high amplification success and were not distinguishable from
wolves requires caution. However, there are several genetic methods for differentiating these species from
wolves that could be applied in routine laboratory analyses. Stronen et al. [28] have shown that only 11 microsatellite markers are sufficient to differentiate golden
jackals from wolves. Even more convenient is to sequence a targeted region of mtDNA that allows to differentiate between the two species, e.g. cytochrome oxidase
I (the barcoding gene), cytochrome b [29] or control region [30]. Amplifying the targeted mtDNA sequence a
priori would not require much resources and could be
implemented routinely for all non-invasive samples before SNP genotyping. As golden jackals were about as
distinguishable from dogs as wolves were, this SNP
Harmoinen et al. BMC Genomics
(2021) 22:473
panel could potentially also be used for detecting hybrids
between these two species, albeit that would require further testing. Golden jackals have been shown to rarely
hybridize with domestic dogs in the wild [30], which
might be more common in the future, as golden jackals
are expanding extensively throughout Europe [31], particularly if suitable mates are scarce, as seen for wolves
[32]. Golden jackals and dogs have also been bred
intentionally to develop a new breed (Sulimov dog) with
good olfactory capabilities [33]: however, although used
for narcotic detection at the Sheremetyevo Airport in
Moscow, their superior olfactory skills have been questioned [34].
Although the discriminating power between wolves
and dogs with this SNP panel was high (100% for F1
and F2, 99% for BC1w), and we were able to assign
even third-generation backcrosses to wolves to the
right category with high accuracy (89–92%), the assignment accuracy for second-generation backcrosses
to wolves was slightly lower (81–82%). This hybrid
category’s lower assignment accuracy is due to the
fact that reliably distinguishing between second- and
third-generation backcrosses is difficult; most of the
incorrectly assigned hybrids from this category were
assigned as third-generation hybrids (the remaining
two or three individuals were assigned as firstgeneration hybrids). However, unless the criteria for
defining a hybrid requires the distinction between
these two hybrid categories, the lower assignment
accuracy in this category is not relevant for management as the individuals would be anyway categorized
as advanced hybrids. The software could not assign
any third-generation backcrosses to dogs into the
right category, possibly because the analysis was
hampered by the large variation in allele frequencies
in dogs. The amount of genetic variation is higher in
dogs than in wolves when all dogs are combined,
but variation within each single breed is less than
that found in wolves [35]. In this SNP set, variation
in dogs is emphasized by the fact that the SNPs are
selected from the Illumina CanineHD Chip, the
SNPs of which are in turn selected from the dog reference genome. Somewhat higher variation in dogs,
wolves and different hybrid categories can be observed in a study testing 100 SNPs chosen from the
Affymetrix Canine Mapping SNP Array 2.0, with
SNPs also originally chosen from the dog reference
genome [21]. Here we attempted to develop an efficient and reliable genotyping method that would
allow to detect wild wolf-dog hybrids during routine
wolf monitoring based on samples with low DNA
quality and quantity. Therefore, reliable discrimination of dogs from backcrosses to dogs falls beyond
the scope of this study.
Page 7 of 15
Extent of hybridization detected in the investigated wolf
populations
During the 18th and 19th centuries, wolf populations in
Central and Western Europe experienced large-scale
contractions in their distribution and reductions in their
population sizes. In the last decades, wolves have increased their distribution range and numbers in many
parts of Europe [3]. During a recolonization phase
when the population size is small, there is an increased risk for hybridization due to the lack of available mates [32]. The same holds for intensively
hunted populations [13, 36]. Severe anthropogenic
disturbance, such as intense hunting or poaching, has
been shown to disrupt the normal social structure of
wolf packs, turning them more tolerant towards individuals outside of the pack [37]. Because of these reasons, removing advanced generation backcrosses from
nature needs to be carefully evaluated on case-by-case
basis. Below, we discuss the evidence for hybridization
detected in this study for each country. It should be
noted that these samples are not representative of actual hybridization rates, as suspected hybrids were
overrepresented for assay testing purposes.
Finland
Despite the fact that the Finnish wolf population experienced severe bottlenecks in the 1920s and 1970s [38,
39], we did not find any sign of admixture in museum
wolves from the 1850s to 1980s. Wolves started to
recolonize Finland in mid-1990 [40]. At present, the
population size is estimated at 185–205 individuals [41].
Up to now, the only hybridization event reported in
Fennoscandia was that of a lone female wolf breeding
with a dog in southern Norway [42, 43]. Here, seven individuals were identified as hybrids; of those, three corresponded to a single hybridization event involving a
male hybrid mating with a female wolf and their two
backcross pups. Six of these individuals had already been
identified as hybrids in Harmoinen et al. (in prep). That
study included one additional hybridization event (comprising five pups), indicating that the total number of
genetically confirmed hybridization events in Finland
amounts to six, involving 12 individuals.
Romania
With a population size of around 2500 individuals
[44], Romania has one of the largest wolf populations
in Europe (note the northwest of the Iberian Peninsula encompasses comparable numbers of about
2200–2500 individuals [3];). In this study, we confirmed the hybrid status of a suspected individual
(second-generation backcross to wolf), which was previously identified using microsatellites (A. Jarausch,
unpublished No signs of introgression were obtained
Harmoinen et al. BMC Genomics
(2021) 22:473
for the 22 remaining samples, suggesting that
hybridization occurs, but may not be a widespread
phenomenon. More samples collected across the entire region are needed to provide a reliable estimate
of wolf-dog hybridization rate in this area.
Germany
After extirpation and absence for almost a century,
wolves of Polish origin have been reported to reproduce
in Germany since 2000 [45]. As of 2019, 105 packs were
documented [46]. Only three cases of hybridization with
male domestic dogs (Saxony in 2003 and Thuringia
2017/ 2019) have been documented in the frame of intense, microsatellite-based national genetic wolf monitoring [46]. No signs of recent dog introgression were
found in the 100 German wolf samples analyzed in this
study, confirming that hybridization rate in Central Europe is very low despite of the ongoing recolonization
process.
Immigrants from the Alpine population
This study supports that the seven individuals sampled
in Germany are immigrants from Italy or the Alpine region. These individuals have a mtDNA haplotype previously only seen in wolves from Italy, and had been
previously assigned to the Italian wolf population based
on microsatellite markers (unpublished). FST between
Italian wolves and these seven individuals was very low
and clustered together in the PCA analysis (Figure S1).
These individuals probably originate from the Alpine region, which was recolonized by wolves in the 1990’s,
after 70 years of absence [47]. There has been only one
study showing low level of hybridization in the Alps
([12]; however, see hybrid detection in regions close by,
[14, 19, 20]). In this study, the seven suspected immigrants were assigned to wolves when Italian wolves were
used as a reference population. In the absence of wolves
from Italy, these individuals were incorrectly assigned to
later-generation backcrosses to wolves. Thus, even for
marker sets with a low population signal, such as this
SNP panel, including individuals belonging to the appropriate reference populations is of critical importance.
Iberian Peninsula
There is some evidence indicating that one individual
may be an advance backcross to wolves (Fig. 2), while
NEWHYBRIDS indicated this individual would be a
wolf. A larger sample size would be needed for results to
be conclusive.
Captive wolves
We found no sign of recent hybridization for the wolves
from both Tierpark Berlin and Wildpark Poing.
Page 8 of 15
Suitability of the SNP panel for non-invasively collected
samples
Our study confirms high SNP amplification success rates
with good genotyping consistency for non-invasively collected samples with the Fluidigm microfluidic array
technology, confirming similar findings in previous studies (e.g., [24, 25]), particularly when protocols are
adapted for samples with low DNA quality and quantity
[25, 48]. In the light of those studies, high levels of genotyping success indicate that more intensive replication
effort is not necessary. This was further supported by
the fact that identical or almost identical genotypes were
obtained from invasive and non-invasive samples from
the same individuals in this study (only one false allele
and three missed alleles in three out of the 30 noninvasive samples examined). We note that, when genotyping success rate of the samples is high (Table 1), disagreements across replicates and thus errors are low (see
[25] for an extensive discussion on this particular point).
Notably, the museum samples (from 1850s and later)
were genotyped with high call rate. Therefore, this
method would be well suited in a variety of scientific
studies, including those based on samples of lower DNA
quality and/or quantity.
Implications for wolf monitoring and research across
Europe
Whereas the obtaining wolf-dog hybridization rates remains a central issue in wolf monitoring and management, relying on non-standardized microsatellite-based
analysis of non-invasively collected samples has so far
hampered the comparability of regional data, resulting in
a lack of over-regional, European-wide hybridization rate
estimates [2]. The application of this novel panel would
solve the technical issues that prevent us from obtaining
data that are comparable across regions.
We found an overall low population signal in this
study. Nevertheless, our results show the importance of
including samples from the relevant populations. Indeed,
including reference samples from wolves from Central
or Northern Europe, the Iberian Peninsula and Italy
and/or the Alpine region when testing for admixture in
these regions is of critical importance. In contrast to
microsatellites, obtaining reference data can be easily
achieved through extracting genotypes from already
available genome-wide SNP or sequence data.
Laboratories that have already established the Fluidigm
genotyping workflow could offer genotyping services
to other institutions, or provide assistance in establishing those protocols. We assume that for most national wolf monitoring programs, only one or two 96
sample array runs per year would be sufficient to
screen for potential hybridization events on a routine
basis, which produces consumable costs of around
Harmoinen et al. BMC Genomics
(2021) 22:473
800 € (without tax) per array plus a couple of working days for one staff member [24, 25, 48].
Wolves and dogs have co-existed for millennia. Even if
dog genomic introgression into wolves is more common
than initially appreciated in studies using a small number of markers, our results show that wolves have kept
their genetic distinctiveness, in agreement with genomewide studies [19, 20, 23]. In addition to a correct management of dogs, maintaining viable population sizes of
wolves and limiting human disturbance on wolf pack
structure is probably the best way to minimize the risk of
hybridization. Wolves play an important ecological role
and perturbations to wolf social structure by removing individuals, particularly advanced backcrosses to wolves,
could in some cases be detrimental and promote further
hybridization. Plans to routinely monitor hybridization in
Europe should be initiated to help identify areas where actions may be directed to better control feral dogs and to
promote measures that would support ecological separation of dogs and wolves. Standardized, concerted assessment of hybridization rates across Europe may serve as a
basis for further research aiming at understanding regional
differences in hybridization rates and degrees of dog introgression in wolf populations.
Conclusions
The designed 96 SNP panel is a highly discriminative
new tool that could be used in routine wolf monitoring
to detect wolf-dog hybrids up to third-generation backcrosses to wolves. We demonstrated a high genotyping
success rate for all sample types, including different
types of non-invasive samples commonly collected in
monitoring practices and even museum samples, making
the panel suitable for various types of studies. Moreover,
the developed SNP panel is applicable at a Europeanwide scale, making it possible to produce comparable results of hybridization rates across the continent, as long
as all the potential reference populations are included in
the analyses. Extensive collection of wolf and dog reference samples is not required, as already published genotypes of wolves and dogs can be added to the analyses.
The study reduces the gap between genomic research
and real-world application by developing a fast and affordable method for wolf monitoring and management
purposes.
Methods
SNP selection
Our initial SNP panel consisted of 300 wolf-dog
ancestry-informative markers (AIMs) obtained from
Harmoinen et al. (in prep). The SNPs were initially selected from a total of 173,662 SNPs on the CanineHD
Whole-Genome BeadChip microarray (Illumina, Inc., San
Diego, California, USA) which was used to genotype
Page 9 of 15
wolves sampled in most of their Eastern European range
(Finland, Sweden, Russia, Estonia, Latvia, Poland, Belarus,
Ukraine, Slovakia, Croatia, Bulgaria and Greece; n = 180)
and dogs from 58 different breeds (collected in Finland,
n = 352). In the study, unlinked (r2 < 0.2) data with MAF >
0.1 was used to select SNPs with the highest FST between
wolves and dogs as AIMs (FST 0.67–0.86). Due to strict
pruning, SNPs were evenly distributed across the 38 autosomal chromosomes. We then excluded SNPs located
near another polymorphic site (minimum separation distance 100 base pairs; based on the dog genome, [49], with
UCSC Genome Browser, [50]) to avoid problems in the
interpretation of results and to simplify primer design
(n = 63 excluded). This resulted in 237 markers, from
which we selected 192 markers with the highest FST values
for downstream testing using microfluidic arrays (Table
S12).
Assay development and testing
SNPtype™ genotyping assays were designed for the 192
selected AIMs and tested on microfluidic 96.96 Dynamic
Arrays™ (Fluidigm Corp., South San Francisco, USA) following the recommendations and testing scheme in
vonThaden et al. [25, 48]. The Fluidigm platform uses
chips containing integrated fluidic circuits (IFCs), harbouring nanoscale PCR reaction chambers that allow the
simultaneous genotyping of 96 samples and 96 loci [51].
We chose samples with high DNA concentration (n =
92, tissue and concentrated buccal swabs; ~ 20–80 ng/μl
DNA) for the initial assessment of the 192 AIMs following in silico design. Samples included wolves (n = 51),
non-pedigree dogs (n = 30), known hybrids (n = 7) and
three species that may be a source of DNA contaminations in non-invasively collected samples (red fox, Vulpes
vulpes, n = 1; golden jackals, Canis aureus, n = 2; and red
deer, Cervus elaphus, n = 1; see next section for more information on the samples). All 192 AIMs were initially
run without a multiplexed pre-amplification step to exclude primer interference as a cause of potential performance failure. Results were then examined to exclude
markers that either: (i) produced ambiguous genotype
clusters or fluorescence for non-template controls (n =
38); or (ii) showed genotype disagreements compared to
the genotypes generated with the Illumina CanineHD
chip (n = 6). Subsequently, the best performing 96 SNPs
were selected and tested on the same reference sample
set, but now including a multiplexed pre-amplification
step (specific target amplification; STA) according to the
manufacturer’s protocol, which is recommended for
samples with moderate DNA concentration. In subsequent runs of samples with low DNA quality and quantity, we adjusted the manufacturer’s STA protocol as
indicated in vonThaden et al. ([25]; i.e., 3.2 μl instead of
Harmoinen et al. BMC Genomics
(2021) 22:473
1.25 μl DNA template and 18 instead of 14 PCR cycles
in the STA step).
Application of the selected 96 SNP panel
Using the final 96 SNP panel, we genotyped samples collected both invasively (tissue) and non-invasively (scats,
saliva from kills, urine, hairs) to generate sufficient data
for subsequent analyses of marker performance and discriminative power (Table 4; Table S5). Tissue samples
were selected from our collections of wolves, dogs and
other canids, which were obtained from road-kills and
other carcasses. For 11 individuals we had both invasively
and non-invasively collected samples, which allowed to
compare marker performance in samples with high versus
low DNA quantity and quality, respectively. Wolf samples
were collected from three areas within the European distribution range (Central European population: Germany,
n = 117; Carpathian population: Romania, n = 28; and Karelian population: Finland, n = 65 and Russia, n = 5). We
also included 9 samples collected in Germany previously
assigned to the Italian wolf lineage (clustering analyses
based on microsatellite genotypes, data not shown, and
with the most frequent mitochondrial haplotype of the
Italian lineage, haplotype HW22, see [52], corresponding
Page 10 of 15
to haplotype W14 described by [53]). These samples were
obtained as part of the German official wolf monitoring
program and we refer to them here as ‘immigrants from
the Alpine population’. For samples collected in Finland,
more than half (n = 34) were museum samples (tissue,
teeth, bone, footpad, dry blood, skin and claw) collected
between the 1850’s and 1980’s [54]. We also included wolf
samples from two different zoos in Germany, Tierpark
Berlin (n = 3) and Wildpark Poing (n = 1).
As dog reference, we sampled non-pedigree dogs from
Germany (n = 35) and Romania (n = 2), collected from
animal shelters, private owners, and from a carcass
found in the field. We also sampled four individuals belonging to wolfdog breeds (Saarloos Wolfdog, n = 2; and
Czechoslovakian Wolfdog, n = 2).
Furthermore, we had 12 suspected wolf-dog hybrids,
which were identified as such based on previouslyconducted microsatellite genotyping (Germany, n = 3;
Romania, n = 1; Czech Republic, n = 1; and Finland, n =
7). These individuals were found to have less than 0.85
posterior probability to be assigned to the wolf cluster
when analyzed with Bayesian assignment procedures implemented in STRUCTURE (unpublished data, see Table
S6 for more information on these individuals). Five of
Table 4 Number of genotyped samples with (a) the 96-SNP panel and (b) the Illumina CanineHD BeadChip, as well as the number
of individuals included in the analyses after removal of samples with low genotyping success and construction of consensus
genotypes from repeatedly genotyped individuals. See Table S5 for a complete sample list
a) 96-SNP panel dataset
Species
Sampling location
Genotyped samples (n)
Analyzed individuals (n)
Gray wolf
Germany
117
100
Dog
Wolf-dog hybrid
Germany (immigrants from Alps/Italy)
9
7
Romania
28
21
Finland
65
61
Russia
5
4
Captive (Germany)
4
4
Germany
39
38
Romania
2
2
Germany
4
3
Romania
1
1
Czech Republic
1
1
Finland
7
7
Golden jackal
Germany
3
3
Fox
Germany
4
3
b) Illumina CanineHD BeadChip datasets
Species
Sampling location (n)
Analyzed individuals (n)
Gray wolf
Italy
70
Dog
Iberian Peninsula
25
Finland
274
Harmoinen et al. BMC Genomics
(2021) 22:473
the suspected hybrids from Finland had also been genotyped with the Illumina CanineHD chip data and their
hybrid status was supported (Harmoinen et al. in prep).
To test for cross-amplification of DNA from species
that may be present in non-invasively collected wolf
samples, we included samples (n = 20) from human
(Homo sapiens), roe deer (Capreolus capreolus), red deer
(Cervus elaphus), Eurasian goat (Capra aegagrus hircus),
sheep (Ovis sp.), wild boar (Sus scrofa), red fox (Vulpes
vulpes), golden jackal (Canis aureus) and other European
carnivore species (Table S5).
Genomic DNA from tissue and blood samples was extracted using the DNeasy® Blood & Tissue Kit (Qiagen),
from scat and urine samples using the DNA Stool Mini
Kit (Qiagen), and from hairs and saliva swabs using the
QIAamp DNA Investigator Kit (Qiagen). For the museum samples, DNA extraction procedures are described
in Jansson et al. [54] and they were genotyped under the
same conditions as non-invasive samples. All genotyping
reactions were set up in a laminar flow hood that was
previously irradiated with UV light for 40 min. The
STA-PCRs were set up in a laboratory dedicated for
non-invasive samples. PCRs were performed in a physically separated laboratory to avoid contaminations. To assess potential genotyping errors, 50 of the 149 tissue
samples were genotyped 2–3 times, all scat samples were
replicated 1–3 times and all the remaining non-invasive
samples and museum samples 1–5 times. Some individuals were genotyped using several different sample types
and consensus genotypes were constructed over all samples and replicates (see number of replicates per sample
and samples per individual from Table S5). For that purpose we used a custom script following the simple rules
that the same genotype (i) has to be observed at least
twice, otherwise it is marked as missing data, and (ii)
must be the most commonly observed genotype over all
replicates.
Assessment of assay performance
We removed three SNPs with low genotyping success
rate (< 0.7; BICF2P1334457, BICF2S2305845 and
BICF2G630504215) and thus performed all subsequent
analyses using genotypes based on 93 SNPs. We also
removed a few samples that failed to amplify in all reactions e.g. due to poor sample quality (see Table 1).
Assay performance was assessed using three different
measures:
(i) Genotyping success rates for DNA from different
sources (tissue, concentrated buccal swab, saliva
swab, hair, scat, urine, blood, museum samples). For
each sample the proportion of scored loci over all
loci was calculated, and an average was obtained for
the corresponding tissue category (Table 1).
Page 11 of 15
(ii) Genotyping consistency
a) between non-invasive and tissue samples from
the same individuals (tissue samples n = 11,
non-invasive samples n = 30). For 11 individuals,
we compared the genotype from the tissue sample against each genotype from a non-invasively
collected sample. We counted a false allele when
an allele found in the genotype of a non-invasive
sample was not present in the genotype of a tissue sample. A missing allele was counted in the
cases in which two alleles were present in the
tissue sample, and only one in the non-invasive
sample. The proportion of false alleles was calculated as the number of false alleles divided by
number of homozygous genotypes (n = 2664)
and the proportion of missing alleles as the
number of divided by the number of heterozygous genotypes (n = 104) in the tissue samples,
due to the fact that the selected SNPs were biallelic. In addition, we counted the number of loci
with missing data and divided by number of all
loci to get the missing rate per sample. Proportion of loci with missing genotypes in the study
was calculated by taking average over samples.
b) between microfluidic array-based and Illumina
CanineHD chip genotypes of the same individuals (n = 22 tissue samples). Illumina CanineHD
genotypes of the wolves were taken from
Harmoinen et al. (in prep), extracting the
genotypes for the corresponding 93 SNPs. To be
able to calculate the genotyping error rates, we
assumed that the genotype based on the
Illumina chip was the true genotype.
(iii)Cross-species amplifications. We checked if any of
the samples we included in the assays that were not
wolves or dogs yielded genotypes.
Samples with < 0.8 genotyping success rate (proportion
of scored loci per sample) were removed from all analyses (wolves, n = 14; potential wolf-dog hybrids, n = 1;
potential cross-species contaminants, n = 15), except for
two foxes which were included with genotyping rates of
0.77 and 0.78 (removed samples indicated in Table S5).
Statistical analyses
For the statistical analysis of hybridization and population differentiation, we added additional genotypes to
the dataset. The genotypes of 70 Italian and 25 Iberian
wolves were extracted from the Illumina CanineHD chip
data (Table 2; Table S5) and included in the analyses to
test the performance of the SNP panel on wolves from
Southern and Western Europe, which are genetically differentiated from other European wolf populations based
on earlier genome-wide analyses [55]. Similarly, we
Harmoinen et al. BMC Genomics
(2021) 22:473
extracted the genotypes of 274 dogs belonging to 55
breeds from the CanineHD chip dataset, in order to capture a larger proportion of the genetic diversity in dogs
for the admixture and assignment analysis. Among the
274 dogs, there were ten individuals from two wolfdog
breeds (Saarloos Wolfdog, n = 5 and Czechoslovakian
Wolfdog, n = 5).
The total dataset used in the analyses consisted of 288
wild wolves, 4 wolves from zoos, 314 dogs (including 14
individuals from wolf-dog breeds), 12 suspected hybrids,
3 golden jackals and 3 foxes (Table 2).
We conducted principal component analysis (PCA)
using the SMARTPCA package of the EIGENSOFT software [56] to visualize the genetic distance between individuals. Then we analyzed the dataset using a Bayesian
clustering approach implemented in STRUCTURE ver
2.3.4 [57]. We conducted 5 independent runs for each
value of K between 1 and 6 with a burn-in length of 50,
000 and a run length of 500,000 Markov Chain Monte
Carlo (MCMC) repetitions. We used the admixture
model and correlated allele frequencies. We use the
STRUCTURE HARVESTER program [58] and estimated
the most likely number of populations (K) using the
Evanno method [59]. The most likely number of clusters
was two (Figure S5) and we used the mean over the 5 independent runs with K = 2 to estimate the assignment of
each individual as wolf or dog. We also ran STRUCT
URE analysis in the same way just for wolf genotypes in
order to identify the most likely number of subpopulations among the wolves. As STRUCTURE is known to
be affected by unequal sample sizes [60], we reduced the
sample size in each geographical area to 20 individuals
(the smallest sample size in our dataset) by excluding
samples based on pairwise relatedness.
To test for population differentiation, and for differentiation between wolves and dogs, we calculated FST
values between different groups of samples using ARLE
QUIN 3.5.2.2 [61]. We considered all wolves as one
group (n = 288) or as separate groups based on sampling
location (for groups, see Table S4). In the case of dogs,
we excluded individuals from wolf-dog breeds (n = 300).
We performed 1000 bootstraps in order to get p-values
around pairwise FST values.
We used the Bayesian model-based software NEWH
YBRIDS without prior information about parental individuals (i.e., the z-option was not used), in order to see
how the software categorized the empirical dataset into
different hybrid classes. The software estimates the posterior probability of individuals falling into one of four
default categories: two parental populations, F1, F2 and
the two first-generation backcrosses to wolves (BCw)
and dogs (BCd). We included four additional classes
(second and third-generation backcrosses; BC2w, BC2d,
BC3w and BC3d) using the corresponding derived
Page 12 of 15
frequencies. We analyzed all samples together but, because we found significant differentiation between different wolf populations in the analyses described above, we
also conducted three additional analyses for wolf samples from (i) Central and Eastern Europe including
Finland, Russia, Germany, Romania (n = 186); (ii) Italy
(n = 70) and (iii) the Iberian Peninsula (n = 25). In all the
analysis, we included all of the dog samples, including
individuals from wolf-dog breeds (n = 314). In the analysis of wolf samples from Central and Eastern Europe
(n = 186), we also included the suspected wolf-dog hybrids (n = 12), the wolves from the animal parks (n = 4)
and the golden jackals and foxes (n = 6). In the analysis
of wolf samples from Italy (n = 70), we also included the
immigrants from the Alpine population sampled in
Germany (n = 7). All the runs were conducted with four
different prior combinations to explore the sensitivity of
the results. We ran the program for an initial burn-in of
100,000 sweeps followed by 500,000 MCMC sweeps. A
posterior probability value of ≥0.5 was used to assign individuals to a specific class.
To assess the power of the 93 SNPs in detecting recent
hybrids between wolves and dogs, we used simulated genotypes generated with the software HYBRIDLAB v1.0
[62]. The simulated genotypes represented individuals of
eight different hybrid classes (100 individuals for each
class), as described above. We generated genotypes separately for wolves from Central and Eastern Europe (n =
162), Italy (n = 70) and the Iberian Peninsula (n = 25).
For the parental population comprising wolves from
Central and Eastern Europe we included tissue, hair and
saliva samples (n = 162), and excluded scat samples to
minimize the risk of potential DNA contamination in
the field that may affect the allele frequencies. Independently of sample type, all wolf samples had > 0.97 assignment to the wolf cluster with STRUCTURE using K = 2,
in analyses conducted separately for the different wolf
datasets, Central and Eastern Europe, Italy, Iberian Peninsula. The other parental population comprised all the
dog genotypes, except the ones from wolf-dog breeds
(n = 300). Simulated hybrids were subsequently analyzed
using STRUCTURE with K = 2, as well as NEWH
YBRIDS. Simulated genotypes were run with the parental populations using the z-option, which allows to define wolf and dog parental individuals. Runs and
analyses were performed in the same way as described
above for the empirical data.
Abbreviations
AIM: Ancestry-informative marker; BCd: First-generation backcross to dog;
BC2d: Second-generation backcross to dog; BC3d: Third-generation
backcross to dog; BCw: First-generation backcross to wolf; BC2w: Secondgeneration backcross to wolf; BC3w: Third-generation backcross to wolf;
F1: First generation hybrid between parental species; F2: Second generation
hybrid (F1 x F1); IFC: Integrated fluidic circuit; SNP: Single nucleotide
polymorphism; STA: Specific target amplification
Harmoinen et al. BMC Genomics
(2021) 22:473
Supplementary Information
The online version contains supplementary material available at https://doi.
org/10.1186/s12864-021-07761-5.
Additional file 1: Table S1. Comparison of non-invasive samples to the
consensus of corresponding invasive samples. Corresponding samples are
on the same row. False allele: an allele seen in non-invasive sample but
not in corresponding invasive sample; Missing allele: an allele seen in invasive sample but not detected in the corresponding non-invasive sample; Missing data: sample didn’t produce readable genotype, Hom:
homozygous, Het: heterozygous.
Additional file 2: Table S2. Comparison of genotypes from same
individuals genotyped with CanineHD chip (Illumina) and microfluidic
array (Fluidigm). We assumed Illumina genotype as the true genotype of
individual. False allele: an allele seen in Fluidigm genotype but not in
corresponding Illumina genotype; Missing allele: an allele seen in Illumina
genotype but not in Fluidigm genotype; Hom: homozygous; Het:
heterozygous.
Additional file 3: Table S3. Genotyping success (proportion of
successfully scored loci over all SNP loci) for samples from species that
are potential sources of DNA in non-invasively collected samples.
Additional file 4: Table S4. FST values for dogs and wolves grouped
based on the sampling location, except for Italian immigrants that were
sampled in Germany. Analysis was performed also without Italian
immigrants (n = 7) and Russian wolves (n = 4) due to low sample sizes in
these groups. However, the FST values between the remaining groups
were the same. When all wolves were combined as one group (n = 288),
the overall FST to dogs (without wolf-dog breeds, n = 300) was 0.72 (p <
0.05). The overall FST between wolf-dog breeds (n = 14) and dogs (n =
300) was FST = 0.20, p < 0.05.
Additional file 5: Table S5. The first tab contains names, locations and
sampling dates for the samples that were genotyped in this study.
Column named “Replicates” in the first tab shows how many times the
same sample was genotyped, and the number in the parenthesis shows
how many times the same individual was genotyped. Samples that were
not included in the analysis are indicated. The second tab contains
names, locations and sampling dates for the samples from the wolves
from Italy and the Iberian Peninsula that were genotyped with the
CanineHD Whole-Genome BeadChip (Illumina). The third tab contains
sample names and breeds for the dog samples that were genotyped
with the Illumina CanineHD chip. In each tab, there are results from the
STRUCTURE and NEWHYBRIDS runs, with all possible prior combinations,
for all the individuals included in the runs. When the result differed between the runs, several results were included. If the analysis was been
done using a consensus genotype based on several samples from the
same individual, the same result is indicated for all samples.
Additional file 6: Table S6. Description of suspected hybrid samples
and discussion of the corresponding results. Microsatellite results, used
for comparison, are unpublished.
Additional file 7: Table S7. NEWHYBRIDS and STRUCTURE results for
Finnish museum samples categorized in different time periods as in
Jansson et al. [54].
Page 13 of 15
possible prior combinations. Range of results from different runs is
indicated.
Additional file 12: Table S12. Description of SNPs and primer
sequences used in this study. Allele frequencies for the 93 SNPs are also
reported. The first tab contains 96 SNPs that were included in the final
SNP panel. Three SNPs that were removed before the analysis are
indicated. The remaining SNPs that were tested but not included in the
final panel and their corresponding primer sequences are in the second
tab.
Additional file 13: Figure S1. Principal component analysis (PCA) for
wild wolves based on 93 SNPs selected to maximize discriminatory
power between wolves and dogs. Wolves are labeled based on sampling
locations, except immigrants from the Italian wolf population, which
were sampled in Germany.
Additional file 14: Figure S2. Delta K values for 1 ≤ K ≤ 8 when
analyzed wolves with STRUCTURE.
Additional file 15: Figure S3. STRUCTURE analysis for the wolf dataset
using the best K value (K = 2).
Additional file 16: Figure S4. STRUCTURE analysis for the wolf dataset
using K = 3.
Additional file 17: Figure S5. Delta K values for 1 ≤ K ≤ 8 when
analyzed whole dataset with STRUCTURE. K = 2 had highest value.
Acknowledgements
We thank the Tierpark Berlin and the Wildpark Poing, the Tierschutz- und
Wildgehegeverein im Tierzentrum e.V., and several private pet owners for
sample provision. We thank the Portuguese Institute for Nature Conservation
and Forestry (SMLM/ICNF), Consejería de Medio Ambiente del Principado de
Asturias, Consellería de Medio Ambiente de la Xunta de Galicia, Consultora
de Recursos Naturales S.L. (E. Arberas and M.A. Campos) and Luís Llaneza for
providing Iberian wolf samples.
Authors’ contributions
Samples were provided by AG, CN, IK, JA and TS. Additional Illumina datasets
were provided by AR-G, ER, FM, HL, JH, MG, RC and RG. Laboratory work was
carried out by AJ, AvT and BC. All the analyses were conducted by JH. First
draft of the manuscript was written by JH, VM-F, AvT and CN. Final manuscript
was edited and approved by all co-authors.
Funding
JH received funding from the Maj and Tor Nessling Foundation, the Finnish
Cultural Foundation and Emil Aaltonen Foundation. Laboratory analyses
were cofounded by the (LOEWE) program (Landes-Offensive zur Entwicklung
Wissenschaftlich-ökonomischer Exzellenz) of the German Federal State of
Hessen. AvT received funding from the Karl und Marie Schack-Stiftung. RG
was supported by the Portuguese Foundation for Science and Technology
(FCT, DL57/2016). The funding bodies played no role in the design of the
study and collection, analysis, and interpretation of data and in writing the
manuscript.
Additional file 8: Table S8. NEWHYBRIDS and STRUCTURE results for
wolves living in two animal parks in Germany.
Availability of data and materials
The dataset supporting the conclusions of this article is available in the
Dryad repository: https://doi.org/10.5061/dryad.76hdr7stk.
Additional file 9: Table S9. NEWHYBRIDS and STRUCTURE results for
other canid species that successfully amplified with the SNP panel.
Declarations
Additional file 10: Table S10. Assignment accuracy for the selected 93
SNPs to categorize simulated individuals between dog and Italian wolves
from 8 different hybrid classes to the correct category (> 0.5) or assign it
to any hybrid category (sum of assignments to hybrid categories > 0.7)
by the software NEWHYBRIDS. Analysis was run with the four possible
prior combinations. Range of results from different runs is indicated.
Additional file 11: Table S11. Assignment accuracy for the selected 93
SNPs to categorize simulated individuals between dog and Iberian
wolves for 8 different hybrid classes to the correct category (> 0.5) or
assign it to any hybrid category (sum of assignments to hybrid categories
> 0.7) by the software NEWHYBRIDS. Analysis was run with the four
Ethics approval and consent to participate
No animals were trapped, harmed or killed for this study. German samples
were obtained from the responsible authorities of the Federal States in the
frame of the legal national wolf monitoring. Finnish samples were obtained
from the Natural Resources Institute Finland, one of the prescribed tasks of
which is to act as a sample bank. Romanian samples were collected in the
frame of an EU-funded LIFE project (LIFE13NAT/RO/000205). No permissions
were needed for this study specifically.
Consent for publication
Not applicable.
Harmoinen et al. BMC Genomics
(2021) 22:473
Competing interests
The authors declare no conflict of interest.
Author details
1
Ecology and Genetics Research Unit, University of Oulu, Oulu, Finland.
2
Conservation Genetics Group, Senckenberg Research Institute and Natural
History Museum Frankfurt, Gelnhausen, Germany. 3Institute for Ecology,
Evolution and Diversity, Johann Wolfgang Goethe-University, Biologicum,
Frankfurt am Main, Germany. 4LOEWE Centre for Translational Biodiversity
Genomics (LOEWE-TBG), Frankfurt am Main, Germany. 5Association for the
Conservation of Biological Diversity, Focşani, Romania. 6Department of
Systems Ecology and Sustainability, Faculty of Biology, University of
Bucharest, Bucharest, Romania. 7Department of Veterinary Biosciences,
University of Helsinki, Helsinki, Finland. 8Department of Medical and Clinical
Genetics, University of Helsinki, Helsinki, Finland. 9Folkhälsan Research Center,
Helsinki, Finland. 10Natural Resources Institute Finland (Luke), Eteläranta 55,
FI-96300 Rovaniemi, Finland. 11Department of Biology, Biotechnical Faculty,
University of Ljubljana, Ljubljana, Slovenia. 12Department of Biotechnology
and Life Sciences, Insubria University, Varese, Italy. 13Unit for Conservation
Genetics (BIO-CGE), Department for the Monitoring and Protection of the
Environment and for Biodiversity Conservation, Italian Institute for
Environmental Protection and Research, Bologna, Italy. 14Scientific Area, WWF
Italy, Rome, Italy. 15CIBIO/InBIO, Centro de Investigação em Biodiversidade e
Recursos Genéticos, Universidade do Porto, Campus de Vairão, 4485-661
Vairão, Portugal. 16Department of Biology, Faculty of Science, University of
Porto, Porto, Portugal. 17Department of Zoology and Animal Cell Biology,
Zoology Laboratory, University of the Basque Country (UPV/EHU),
Vitoria-Gasteiz, Spain. 18Department of Biological, Geological and
Environmental Sciences, University of Bologna, Bologna, Italy. 19Department
of Chemistry and Bioscience, Faculty of Engineering and Science, University
of Aalborg, Aalborg, Denmark. 20European Molecular Biology Laboratory,
European Bioinformatics Institute, Wellcome Trust Genome Campus, Hinxton,
Cambridge, UK.
Received: 12 May 2021 Accepted: 1 June 2021
References
1. Chapron G, Kaczensky P, Linnell JDC, Von Arx M, Huber D, Andrén H, et al.
Recovery of large carnivores in Europe’s modern human-dominated
landscapes. Science. 2014;346(6216):1517–9. https://doi.org/10.1126/
science.1257553.
2. de Groot GA, Nowak C, Skrbinšek T, Andersen LW, Aspi J, Fumagalli L, et al.
Decades of population genetic research reveal the need for harmonization
of molecular markers: the grey wolf Canis lupus as a case study. Mammal
Rev. 2016;46(1):44–59. https://doi.org/10.1111/mam.12052.
3. Hindrikson M, Remm J, Pilot M, Godinho R, Stronen AV, Baltrūnaité L, et al.
Wolf population genetics in Europe: a systematic review, meta-analysis and
suggestions for conservation and management. Biol Rev. 2017a;92(3):1601–
29. https://doi.org/10.1111/brv.12298. https://doi.org/10.1111/brv.12298.
4. Aspi J, Roininen E, Ruokonen M, Kojola I, Vilà C. Genetic diversity, population
structure, effective population size and demographic history of the Finnish
wolf population. Mol Ecol. 2006;15(6):1561–76. https://doi.org/10.1111/j.13
65-294X.2006.02877.x.
5. Czarnomska SD, Jędrzejewska B, Borowik T, Niedziałkowska M, Stronen AV,
Nowak S, et al. Concordant mitochondrial and microsatellite DNA structuring
between polish lowland and Carpathian Mountain wolves. Conserv Genet.
2013;14(3):573–88. https://doi.org/10.1007/s10592-013-0446-2.
6. Fabbri E, Miquel C, Lucchini V, Santini A, Caniglia R, Duchamp C, et al. From
the Apennines to the Alps: colonization genetics of the naturally expanding
Italian wolf (Canis lupus) population. Mol Ecol. 2007;16(8):1661–71. https://
doi.org/10.1111/j.1365-294X.2007.03262.x.
7. Flagstad, Walker CW, Vilà C, Sundqvist AK, Fernholm B, Hufthammer AK,
et al. Two centuries of the Scandinavian wolf population: patterns of
genetic variability and migration during an era of dramatic decline. Mol
Ecol. 2003;12:869–80 https://doi.org/10.1046/j.1365-294x.2003.01784.x.
8. Granroth-Wilding H, Primmer C, Lindqvist M, Poutanen J, Thalmann O,
Aspi J, et al. Non-invasive genetic monitoring involving citizen science
enables reconstruction of current pack dynamics in a re-establishing
wolf population. BMC Ecol. 2017;17. https://doi.org/10.1186/s12898-0170154-8(1):44.
Page 14 of 15
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
Szewczyk M, Nowak S, Niedźwiecka N, Hulva P, Špinkytė-Bačkaitienė R,
Demjanovičová K, et al. Dynamic range expansion leads to establishment of
a new, genetically distinct wolf population in Central Europe. Sci Rep. 2019;
9. https://doi.org/10.1038/s41598-019-55273-w(1):19003.
Godinho R, Llaneza L, Blanco JC, Lopes S, Álvares F, García EJ, et al. Genetic
evidence for multiple events of hybridization between wolves and domestic
dogs in the Iberian Peninsula. Mol Ecol. 2011;20(24):5154–66. https://doi.
org/10.1111/j.1365-294X.2011.05345.x.
Godinho R, López-Bao JV, Castro D, Llaneza L, Lopes S, Silva P, et al. Realtime assessment of hybridization between wolves and dogs: combining
noninvasive samples with ancestry informative markers. Mol Ecol Resour.
2015;15:317–28. https://doi.org/10.1111/1755-0998.12313.
Dufresnes C, Remollino N, Stoffel C, Manz R, Weber JM, Fumagalli L. Two
decades of non-invasive genetic monitoring of the grey wolves
recolonizing the Alps support very limited dog introgression. Sci Rep. 2019;
9(1):1–9. https://doi.org/10.1038/s41598-018-37331-x.
Moura AE, Tsingarska E, Dabrowski MJ, Czarnomska SD, Jedrzejewska B, Pilot
M. Unregulated hunting and genetic recovery from a severe population
decline: the cautionary case of Bulgarian wolves. Conserv Genet. 2014;15(2):
405–17. https://doi.org/10.1007/s10592-013-0547-y.
Randi E, Hulva P, Fabbri E, Galaverni M, Galov A, Kusak J, et al. Multilocus
detection of wolf x dog hybridization in Italy, and guidelines for marker
selection. PLoS One. 2014;9(1):e86409. https://doi.org/10.1371/journal.pone.
0086409.
Salvatori V, Godinho R, Braschi C, Boitani L, Ciucci P. High levels of recent
wolf × dog introgressive hybridization in agricultural landscapes of central
Italy. Eur J Wildl Res. 2019;65(5):1–14. https://doi.org/10.1007/s10344-01
9-1313-3.
Caniglia R, Galaverni M, Velli E, Mattucci F, Canu A, Apollonio M, et al. A
standardized approach to empirically define reliable assignment thresholds
and appropriate management categories in deeply introgressed
populations. Sci Rep. 2020;10(1):1–14. https://doi.org/10.1038/s41598-02059521-2.
Steyer K, Tiesmeyer A, Muñoz-Fuentes V, Nowak C. Low rates of
hybridization between European wildcats and domestic cats in a humandominated landscape. Ecol Evol. 2018;8(4):2290–304. https://doi.org/10.1002/
ece3.3650.
Vähä JP, Primmer CR. Efficiency of model-based Bayesian methods for
detecting hybrid individuals under different hybridization scenarios and
with different numbers of loci. Mol Ecol. 2006;15(1):63–72. https://doi.org/1
0.1111/j.1365-294X.2005.02773.x.
Galaverni M, Caniglia R, Pagani L, Fabbri E, Boattini A, Randi E. Disentangling
timing of admixture, patterns of introgression, and phenotypic indicators in
a hybridizing wolf population. Mol Biol Evol. 2017;34(9):2324–39. https://doi.
org/10.1093/molbev/msx169.
Pilot M, Greco C, vonHoldt BM, Randi E, Jędrzejewski W, Sidorovich VE, et al.
Widespread, long-term admixture between grey wolves and domestic dogs
across Eurasia and its implications for the conservation status of hybrids.
Evol Appl. 2018;11(5):662–80. https://doi.org/10.1111/eva.12595.
vonHoldt BM, Pollinger JP, Earl DA, Parker HG, Ostrander EA, Wayne RK.
Identification of recent hybridization between gray wolves and
domesticated dogs by SNP genotyping. Mamm Genome. 2013;24:80–8
https://doi.org/10.1007/s00335-012-9432-0.
Fan Z, Silva P, Gronau I, Wang S, Armero AS, Schweizer RM, et al. Worldwide
patterns of genomic variation and admixture in gray wolves. Genome Res.
2016;26(2):163–73. https://doi.org/10.1101/gr.197517.115.
Gómez-Sánchez D, Olalde I, Sastre N, Enseñat C, Carrasco R, Marques-Bonet
T, et al. On the path to extinction: inbreeding and admixture in a declining
grey wolf population. Mol Ecol. 2018;27(18):3599–612. https://doi.org/1
0.1111/mec.14824. https://doi.org/10.1111/mec.14824.
Kraus RHS, vonHoldt B, Cocchiararo B, Harms V, Bayerl H, Kühn R, et al. A
single-nucleotide polymorphism-based approach for rapid and cost-effective
genetic wolf monitoring in Europe based on noninvasively collected samples.
Mol Ecol Resour. 2015;15:295–305 https://doi.org/10.1111/1755-0998.12307.
vonThaden A, Cocchiararo B, Jarausch A, Jüngling H, Karamanlidis AA,
Tiesmeyer A, et al. Assessing SNP genotyping of noninvasively collected
wildlife samples using microfluidic arrays. Sci Rep. 2017;7:1–13 https://doi.
org/10.1038/s41598-017-10647-w.
Pritchard JK, Stephens M, Donnelly P. Inference of population structure
using multilocus genotype data. Genetics. 2000;155(2):945–59. https://doi.
org/10.1093/genetics/155.2.945.
Harmoinen et al. BMC Genomics
(2021) 22:473
27. Anderson EC, Thompson EA. A model-based method for identifying species
hybrids using multilocus genetic data. Genetics. 2002;160(3):1217–29.
https://doi.org/10.1093/genetics/160.3.1217.
28. Stronen AV, Bartol M, Boljte B, Jelenčič M, Kos I, Potočnik H, et al. “Passive
surveillance” across species with cross-amplifying molecular markers: the
potential of wolf (Canis lupus) genetic monitoring in tracking golden jackal
(C. aureus) colonization and hybridization. Hystrix. 2020;31:1–3 https://doi.
org/10.4404/hystrix-00259-2019.
29. Rueness EK, Asmyhr MG, Sillero-Zubiri C, Macdonald DW, Bekele A, Atickem
A, et al. The cryptic African wolf: Canis aureus lupaster is not a golden jackal
and is not endemic to Egypt. PLoS One. 2011;6. https://doi.org/10.1371/
journal.pone.0016385(1):e16385.
30. Galov A, Fabbri E, Caniglia R, Arbanasić H, Lapalombella S, Florijančić T, et al.
First evidence of hybridization between golden jackal (Canis aureus) and
domestic dog (Canis familiaris) as revealed by genetic markers. R Soc Open
Sci. 2015;2(12):150450. https://doi.org/10.1098/rsos.150450.
31. Trouwborst A, Krofel M, Linnell JDC. Legal implications of range expansions
in a terrestrial carnivore: the case of the golden jackal (Canis aureus) in
Europe. Biodivers Conserv. 2015;24(10):2593–610. https://doi.org/10.1007/s1
0531-015-0948-y.
32. Muñoz-Fuentes V, Darimont CT, Paquet PC, Leonard JA. The genetic legacy
of extirpation and re-colonization in Vancouver Island wolves. Conserv
Genet. 2010;11(2):547–56. https://doi.org/10.1007/s10592-009-9974-1.
33. Sulimov KT: PhD thesis. Kinologicheskaya identifikaciya individuma po
obonuatel’nym signalam. Institut Problem Ekologii I Evolucii imini A. N.
Severtsova; 1995
34. Hall NJ, Protopopova A, Wynne CDL. Olfaction in wild canids and Russian
canid hybrids. In: Jezierski T, Ensminger J, Paper E, editors. Canine olfaction
science and law: advances in forensic science, medicine, conservation, and
environmental remediation. Boca Raton: Taylor and Francis; 2016. p. 57–66.
https://doi.org/10.1201/b20027.
35. Wang GD, Zhai W, Yang HC, Wang L, Zhong L, Liu YH, et al. Out of
southern East Asia: the natural history of domestic dogs across the world.
Cell Res. 2016;26(1):21–33. https://doi.org/10.1038/cr.2015.147.
36. Rutledge LY, White BN, Row JR, Patterson BR. Intense harvesting of eastern
wolves facilitated hybridization with coyotes. Ecol Evol. 2012;2(1):19–33.
https://doi.org/10.1002/ece3.61.
37. Jȩdrzejewski W, Branicki W, Veit C, Medugorac I, Pilot M, Bunevich AN, et al.
Genetic diversity and relatedness within packs in an intensely hunted
population of wolves Canis lupus. Acta Theriol (Warsz). 2005;50:3–22 https://
doi.org/10.1007/BF03192614.
38. Ermala A. A survey of large predators in Finland during the 19 th −20 th
centuries. Acta Zool Litu. 2003;13(1):15–20. https://doi.org/10.1080/13921
657.2003.10512538.
39. Pulliainen E. Studies on the wolf (Canis lupus L.) in Finland. Ann Zool Fenn.
1965;2:215–59.
40. Kojola I, Helle P, Heikkinen S, Lindén H, Paasivaara A, Wikman M. Tracks in
snow and population size estimation: the wolf Canis lupus in Finland. Wildl
Biol. 2014;20(5):279–84. https://doi.org/10.2981/wlb.00042.
41. Heikkinen S, Kojola I, Mäntyniemi S, Holmala K, Härkälä A. Susikanta
Suomessa maaliskuussa 2019. Luonnonvara- ja Biotalouden tutkimus 35/
2019. Luonnonvarakeskus: Helsinki; 2019. http://urn.fi/URN:ISBN:978-952-326767-1. Accessed 10 May 2021
42. Vilà C, Sundqvist AK, Flagstad Ø, Seddon J, Björnerfeldt S, Kojola I, et al.
Rescue of a severely bottlenecked wolf (Canis lupus) population by a single
immigrant. Proc R Soc B Biol Sci. 2003a;270(1510):91–7. https://doi.org/10.1
098/rspb.2002.2184.
43. Vilà C, Walker C, Sundqvist AK, Flagstad, Andersone Z, Casulli A, et al.
Combined use of maternal, paternal and bi-parental genetic markers for the
identification of wolf-dog hybrids. Heredity. 2003b;90:17–24 https://doi.org/1
0.1038/sj.hdy.6800175.
44. Kaczensky P, Chapron G, von Arx M, Huber D, Andrén H, Linnell J. Status,
management and distribution of large carnivores - bear, lynx, wolf &
wolverine - in Europe. Part 1 - Europe summaries. Report: 1-72. A Large
Carnivore Initiative for Europe Report prepared for the European
Commission; 2013. https://www.europarc.org/wp-content/uploads/2017/02/
Kaczensky_et_al_2013_Status_management_and_distribution_of_large_ca
rnivores_in_Europe_1.pdf. Accessed 10 June 2021.
45. Reinhardt I, Kluth G, Nowak C, Szentiks CA, Krone O, Ansorge H, et al.
Military training areas facilitate the recolonization of wolves in Germany.
Conserv Lett. 2019;12(3):e12635. https://doi.org/10.1111/conl.12635.
Page 15 of 15
46. Dokumentations- und Beratungsstelle des Bundes zum Thema Wolf (DBBW).
Wölfe in Deutschland, Statusbericht 2018/2019. 2019. https://dbb-wolf.de/
mehr/literatur-download/statusberichte. Accessed 10 May 2021.
47. Marucco F, Avanzinelli E, Boitani L. Non-invasive integrated sampling design
to monitor the wolf population in Piemonte, Italian Alps. Hystrix. 2012;23:5–
13 https://doi.org/10.4404/hystrix-23.1-4584.
48. vonThaden A, Nowak C, Tiesmeyer A, Reiners TE, Alves PC, Lyons LA, et al.
Applying genomic data in wildlife monitoring: Development guidelines for
genotyping degraded samples with reduced single nucleotide
polymorphism panels. Mol Ecol Resour. 2020;20 https://doi.org/10.1111/1
755-0998.13136.
49. Lindblad-Toh K, Wade CM, Mikkelsen TS, Karlsson EK, Jaffe DB, Kamal M,
et al. Genome sequence, comparative analysis and haplotype structure of
the domestic dog. Nature. 2005;438(7069):803–19. https://doi.org/10.1038/na
ture04338.
50. Kent WJ, Sugnet CW, Furey TS, Roskin KM, Pringle TH, Zahler AM, et al. The
human genome browser at UCSC. Genome Res. 2002;12(6):996–1006.
51. Wang J, Lin M, Crenshaw A, Hutchinson A, Hicks B, Yeager M, et al. Highthroughput single nucleotide polymorphism genotyping using nanofluidic
dynamic arrays. BMC Genomics. 2009;10. https://doi.org/10.1186/1471-21
64-10-561(1):561.
52. Pilot M, Branicki W, Jȩdrzejewski W, Goszczyski J, Jȩdrzejewska B, Dykyy I,
et al. Phylogeographic history of grey wolves in Europe. BMC Evol Biol.
2010;10(1):104. https://doi.org/10.1186/1471-2148-10-104.
53. Randi E, Lucchini V, Christensen MF, Mucci N, Funk SM, Dolf G, et al.
Mitochondrial DNA variability in Italian and east European wolves: detecting
the consequences of small population size and hybridization. Conserv Biol.
2000;14(2):464–73. https://doi.org/10.1046/j.1523-1739.2000.98280.x.
54. Jansson E, Harmoinen J, Ruokonen M, Aspi J. Living on the edge:
reconstructing the genetic history of the Finnish wolf population. BMC Evol
Biol. 2014;14(1):64. https://doi.org/10.1186/1471-2148-14-64. https://doi.org/1
0.1186/1471-2148-14-64.
55. Pilot M, Greco C, Vonholdt BM, Jȩdrzejewska B, Randi E, Jȩdrzejewski W,
et al. Genome-wide signatures of population bottlenecks and diversifying
selection in European wolves. Heredity. 2014;112(4):428–42. https://doi.org/1
0.1038/hdy.2013.122.
56. Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich D. Principal
components analysis corrects for stratification in genome-wide association
studies. Nat Genet. 2006;38(8):904–9. https://doi.org/10.1038/ng1847.
57. Falush D, Stephens M, Pritchard JK. Inference of population structure using
multilocus genotype data: dominant markers and null alleles. Mol Ecol
Notes. 2007;7(4):574–8. https://doi.org/10.1111/j.1471-8286.2007.01758.x.
58. Earl DA, vonHoldt BM. STRUCTURE HARVESTER: a website and program for
visualizing STRUCTURE output and implementing the Evanno method. Conserv
Genet Resour. 2012;4(2):359–61. https://doi.org/10.1007/s12686-011-9548-7.
59. Evanno G, Regnaut S, Goudet J. Detecting the number of clusters of
individuals using the software STRUCTURE: a simulation study. Mol Ecol.
2005;14(8):2611–20. https://doi.org/10.1111/j.1365-294X.2005.02553.x.
60. Puechmaille SJ. The program structure does not reliably recover the correct
population structure when sampling is uneven: subsampling and new
estimators alleviate the problem. Mol Ecol Resour. 2016;16:608–27. https://
doi.org/10.1111/1755-0998.12512.
61. Excoffier L, Lischer HEL. Arlequin suite ver 3.5: a new series of programs to
perform population genetics analyses under Linux and windows. Mol Ecol
Resour. 2010;10(3):564–7. https://doi.org/10.1111/j.1755-0998.2010.02847.x.
62. Nielsen EE, Bach LA, Kotlicki P. HYBRIDLAB (version 1.0): A program for
generating simulated hybrids from population samples. Mol Ecol Notes.
2006;6:971–3 https://doi.org/10.1111/j.1471-8286.2006.01433.x.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.