ORIGINAL RESEARCH
published: 02 December 2016
doi: 10.3389/fevo.2016.00134
Frontiers in Ecology and Evolution | www.frontiersin.org 1 December 2016 | Volume 4 | Article 134
Edited by:
Badri Padhukasahasram,
Illumina, USA
Reviewed by:
Klaus-Peter Koepfli,
Smithsonian Conservation Biology
Institute, USA
Jouni Aspi,
University of Oulu, Finland
*Correspondence:
Erik Ersmark
Erik.ersmark@nrm.se
†
These authors have contributed
equally to this work.
Specialty section:
This article was submitted to
Evolutionary and Population Genetics,
a section of the journal
Frontiers in Ecology and Evolution
Received: 22 July 2016
Accepted: 09 November 2016
Published: 02 December 2016
Citation:
Ersmark E, Klütsch CFC, Chan YL,
Sinding M-HS, Fain SR, Illarionova
NA, Oskarsson M, Uhlén M, Zhang
Y-P, Dalén L and Savolainen P (2016)
From the Past to the Present: Wolf
Phylogeography and Demographic
History Based on the Mitochondrial
Control Region.
Front. Ecol. Evol. 4:134.
doi: 10.3389/fevo.2016.00134
From the Past to the Present: Wolf
Phylogeography and Demographic
History Based on the Mitochondrial
Control Region
Erik Ersmark
1, 2
*
, Cornelya F. C. Klütsch
3
, Yvonne L. Chan
1
, Mikkel-Holger S. Sinding
4
,
Steven R. Fain
5
, Natalia A. Illarionova
6
, Mattias Oskarsson
7
, Mathias Uhlén
7
,
Ya-ping Zhang
8
, Love Dalén
1 †
and Peter Savolainen
7 †
1
Department of Bioinformatics and Genetics, Swedish Museum of Natural History, Stockholm, Sweden,
2
Department of
Zoology, Stockholm University, Stockholm, Sweden,
3
Biology Department, Trent University, Peterborough, ON, Canada,
4
Centre for GeoGenetics, Natural History Museum of Denmark, Copenhagen University, Copenhagen, Denmark,
5
National
Forensics Laboratory, U. S. Fish and Wildlife Service, Ashland, OR, USA,
6
Laboratory of Microevolution and Domestication of
Mammals, A. N. Severtsov Institute of Ecology and Evolution, Moscow, Russia,
7
Science for Life Laboratory, Department of
Gene Technology, KTH Royal Institute of Technology, Solna, Sweden,
8
Laboratory of Cellular and Molecular Evolution, and
Molecular Biology of Domestic Animals, Kunming Institute of Zoology, Chinese Academy of Sciences, Kunming, China
The global distribution of the gray wolf (Canis lupus) is a complex assembly consisting of a
large number of populations and described subspecies. How these lineages are related
to one another is still not fully resolved, largely due to the fact that large geographical
regions remain poorly sampled both at the core and periphery of the species’ range.
Analyses of ancient wolves have also suffered from uneven sampling, but have shown
indications of a major turnover at some point during the Pleistocene-Holocene boundary
in northern North America. Here we analyze variation in the mitochondrial control region
in 122 contemporary wolves from some of the less studied populations, as well as six
samples from the previously unstudied Greenland subspecies (Canis l. orion) and two
Late Pleistocene samples from Siberia. Together with the publicly available control region
sequences of both modern and ancient wolves, this study examines genetic diversity on
a wide geographical and temporal scale that includes both Eurasia and North America.
We identify 13 new haplotypes, of which the majority is found in northern and eastern
Asia. The results show that the Greenland samples are all represented by one haplotype,
previously identified in North American wolves, among which this population seems to
trace its maternal lineage. The phylogeny and network analyses show a wide spatial
distribution of several lineages, but also some clusters with more distinct geographical
affiliation. In North America, we find support for an end-Pleistocene population bottleneck
through coalescent simulations under an approximate Bayesian framework in contrast
to previous studies that suggested an extinction-replacement event. However, we find
no support for a similar bottleneck in Eurasia. Overall, this global analysis helps to clarify
our understanding of the complex history for wolves in Eurasia and North America.
Keywords: Canis lupus, phylogeography, mtDNA, turnover, control region
Ersmark et al. Wolf Phylogeography and Demographic History
INTRODUCTION
The gray wolf (Canis lupus) exhibits a tremendous ecological
flexibility with respect to different environments, ranging from
the Arctic tundra to the deserts and dry shrub lands of the Middle
East. This iconic canid was one of the most widely distributed
large terrestrial mammals in the Late Quaternary with a historical
range that covered most of the northern hemisphere (
Nowak,
2003), and its range has expanded even further a longside
humans as the domestic dog ( C. lupus familiaris). In addition
to domestication, humans have impacted the wolf considerably
by restricting its habitat through active persecution, which has
resulted in a dramatic decrease in population sizes, especially
over the last two centuries (Boitani, 2003; Leonard et al., 2005 ).
Shrinking habitats overlapping with closely related dogs and
coyotes (Canis latrans) have also led to numerous oc currences
of hybridization (
Andersone et al., 2002; Lucchini et al., 2004;
Fain et al., 2010; vonHoldt et al., 2011, 2013
). Along with
recent turnovers, t hese characteristics and events make the
phylogeographic history of the wolf in many ways difficult to
disentangle.
The divergence between wolves and coyotes most likely took
place in America at some point between 4.5 and 1.8 million
years ago (Mya;
Nowak, 2003). However, the more recent time
point seems more plausible, considering that a sudden expansion
of the genus Canis, sometimes referred to as t he “wolf event,”
took place at t he beginning of the Pleistocene ( ∼2.5–1. 8 Mya;
Azzaroli, 1983; Wang et al., 2010). This expansion was most lik ely
facilitated by intense continental glaciations, which created open
landscapes including the “mammoth steppe biome” (Azzaroli,
1983; Azzaroli et al., 1988
).
According to the fossil record, the wolf C. lupus ssp. appeared
in Europe around 800 thousand years ago (kya) during t he
Middle Pleistocene and in the mid-latitudes of North A merica
around 100 kya, where its ancestor previously had gone extinct
(Wang et al., 2010). The wolf appears to have been well-
established in Europe from around 400 kya and onwards
(Meloro et al., 2007). Older records are known only from Siberia
and Alaska (Beringia), leading to the assumption that wolves
originated somewhere in this region, whence they spread all
across the Holarctic (Wang et al., 2010).
Despite the wide geographical distribution and complex
evolutionary h istory of the gray wolf, most genetic analyses to
date have concentrated on specific geographical regions (
Randi
et al., 2000; Aggarwal et al., 2007; Pilot et al., 2010; Weckworth
et al., 2010) and/or short fragments of the mitochondrial
DNA (mtDNA; Vila et al., 1999; Valiere et al., 2003; Sharma
et al., 2004; Jedrzejewksi et al., 2 005). Many of these studies
have thus suffered from limited geographical coverage and/or
insufficient resolution at the genetic level. Further, large areas
have remained scarcely sampled, such as Russia/Siberia, China,
and the Middle East; regions holding large interconnected
populations that potentially contain high genetic diversity. There
are also remote regions where wolves have not yet been studied
in terms of genetic differentiation and diversity. Among these
are the extreme north—home to phenotypically distinct arctic
subspecies, which occur in the Canadian arctic (Canis l. arctos)
and on Greenland (Canis l. orion; Pocock, 1935; Wozencraft,
2005).
By applying ancient DNA techniques to subfossil
remains, many studies have sought to reconstruct the wolf’s
phylogeographic history, often specifically in relation to its
domestication by humans (
Verginelli et al., 2005; Germonpré
et al., 2009; Skoglund et al., 2011; Thalmann et al., 2013). Like
numerous studies on modern wolves, these have generally lacked
coverage in terms of geography and/or genetic material, often
to a great extent. Samples of ancient wolf remains have mainly
been collected from sites in Europe and Alaska, le aving out much
of the historical distribution (Stiller et al., 2006; Leonard et al.,
2007; Germonpré et al., 2009), and a majority of these samples
have exclusively been targeted for a short but variable fragment
of the mitochondrial control region (CR; Verginelli et al., 2005;
Stiller et al., 2006).
Despite these limitations, ancient DNA studies have provided
important insights into the population dynamics in certain
regions. In a study on Late Pleistocene wolves from eastern
Beringia (Alaska) a population turnover was detected at the
Pleistocene-Holocene transition, where a diverse group of
haplotypes (haplogroup 2) seemed to have been replaced by a
more distinct and monophyletic group (haplogroup 1), which
also represents the modern wolves in North America (
Leonard
et al., 2007). Interestingly, the former group was morphologically
described as a robust ecomorph, possibly ad apted to large
megafaunal prey. The division into these two genetic groups
can also be applied to the larger Eurasian population, where
both groups are still present in contemporary wolves. However,
following the pattern in North America, all Late Pleistocene
European specimens have been shown to fall within haplogroup
2 (Pilot et al., 2010; Thalmann et al., 2013), which might suggest
that a population turnover occurred in Eurasia as well. Although
only present at low frequency today, haplogroup 2 exhibits a
great deal of diversity, especially when ancient specimens are
included (Pilot et al., 2010). This is also mirrored by indications
of past morphological diversity observed both in North America
and Europe, which suggests that the wolf has suffered a general
decrease, not only in genetic but also morphological diversity
across its range (Leonard et al., 2007; Germonpré et al., 2009).
In addition to the issue of the wolf’s relationship to the
modern dog, there have been several debated uncertainties
concerning specific wolf populations and their status in terms of
subspecies, hybrids, or even distinct species. It has for instance
recently been suggested that wolves from the Great Lakes region
of Ca nada and the United States (Canis l. lycaon) constitutes a
hybrid between gray wolves and coyotes or even a species on its
own (
Fain et al., 2010; vonHoldt et al., 2 011, 2013, 2016).
To provide a more comprehensive phylogeograpy of the gray
wolf, we have used CR-sequences to study t he mtDNA diversity
of wolves on a worldwide scale. We specifically aimed to include
more modern samples from previously less studied areas, as well
as to collate and analyze ancient sequences globally in order
to assess the relationships between wolf populations and how
they have changed over time. We also sampled wolves from
the Great Lakes states Michigan and Minnesota in order to
evaluate if this population shows more sharing of haplotypes with
Frontiers in Ecology and Evolution | www.frontiersin.org 2 December 2016 | Volume 4 | Article 134
Ersmark et al. Wolf Phylogeography and Demographic History
wolf or coyote. Finally, to formally test the proposed population
turnover in North America, as well as the possibility of a similar
turnover in Eurasia, we analyzed the data using serial coalescent
simulations.
MATERIALS AND METHODS
Samples
Tissue samples (n = 128) from wolves were collected from across
the range including Scandinavia, Russia/Siberia, Iran, China,
North America, and Greenland ( Table S1). With the exception
of two captive Mexican wolves, all samples were derived from
wild animals. Of the six samples from Greenland, two were dated
to the fifth and the early eighteenth century, respectively (Table
S1). Two Late Pleistocene wolf remains were also analyzed, both
taken from permafrost sites on the Taimyr Peninsula in Siberia.
After positive DNA screening, t hese samples were radioc ar bon
dated at the Oxford Radiocarbon Accelerator Unit (ORAU),
yielding approximate ages of 35 and 42 thousand calibrated
radiocarbon years before present (kyr BP). For the sake of
consistency, all subsequent radiocarbon dates are presented in
this (calibrated) form (Tables S1, S3). In summary, a total of 130
novel samples were a nalyzed (Table S1). For more information
regarding samples included in the current study, see Tables S1,
S2.
Laboratory Methods
For the modern samples, DNA was extracted from blood and
tissue using Proteinase K and organic solvents (
Sambrook et al. ,
1989), or from hairs (Hopgood et al., 1992). PCR amplification
of a 582 bp long fragment of the hypervariable domain (HVR1)
of the mitochondrial control region was performed with a set
of primers used in a previous study (Savolainen et a l., 2002).
This specific region has commonly been applied in several
previous phylogenetic studies on the wolf and was thus targeted
to facilitate inclusive comparisons. PCR products were sequenced
using the primers above with BigDye Terminator chemistry on
ABI 377 and 3700 instruments (Applied Biosystems Inc.).
Both extraction and pre-PCR preparation of the historical
and ancient samples was performed at a specifically designated
laboratory with high standards of sterility at the Swedish Museum
of Natural History in Stockholm, and the Centre for Geogenetics
at the Natural History Museum of Denmark. DNA was extracted
using a silic a-based method, where ca. 50 mg bone powder from
each sample was incubated overnight under motion at 55
◦
C in
715 µl extraction buffer (0.45 M EDTA, 0.1 M UREA, 150 µg
proteinase K). The samples were then centrifuged at 2300 rpm
for 5 min and the supernatants were collected and concentrated
using 30K MWCO Vivaspin filters (Sartorius). Purifica tion and
elution was performed following
Brace et al. (2012).
Taking the fragmented st at e of ancient DNA into account,
six partially overlapping sequences of the mitochondrial control
region were amplified to cover a total of 686 bp (including
primers), matching the fragment targeted for the modern
samples; MS_wolf1- dogdl5 (148 bp;
Leonard et al., 2005),
MS_wolf2 (205 bp), MS_wolf3 (211 bp), MS_wolf4 (175 bp),
MS_wolf5 (149 bp), and MS_wolf6 (132 bp). Primer sequences
are listed in Table S4.
Polymerase chain reactions (PCRs) were set up using; 1 mM
MgCl
2
(Qiagen), 0.2 mM dNTPs, 0.2 µM of each primer, 1X
PCR-buffer (Qiagen), 0.1 mg/ml BSA, 2 units of Hotstar Taq
(Qiagen) and 2 µl of DNA extract, making a total volume of
25 µl/reaction. Amplifications started with a 10 min denaturation
step at 95
◦
C, followed by 55 cycles of denaturation at 95
◦
C
for 30 s, annealing for 30 s at 58–62
◦
C, followed by extension
at 72
◦
C for 30 s. A final extension step at 72
◦
C for 7 min was
included at the end of the procedure. Confirming successful
amplifications was done using gel electrophoresis, by apply ing
5 µl PCR product on a 1.5% agarose gel prepared with fluorescent
GelRed (Biotium Inc.). The gel was run in 0.5X TBE buffer
(50 ml) and inspected with UV-light. The PCR-products (20 ul)
were further purified with EXO-SAP enzymes (5 ul; Thermo
Fisher Scientific). Sequencing reactions were then performed
using t he BigDye Terminator kit ver.1.1 (Applied Biosystems
Inc.), and the products were purified with the DyeEx 96 Kit
(Qiagen). An ABI 3130xl Genetic analyzer (Applied Biosystems
Inc.) was used for the final sequencing analysis.
To easily detect contamination and minimize the risk of
cross-contamination, all extractions were made in small series
with regular inclusions of blank s amples. The same procedure
was applied for the PCR preparation, and a minimum of two
PCR products were sequenced for every sample to confirm its
authenticity and detect possible discrepancies caused by post-
mortem DNA damage (
Hofreiter et al., 2001).
Alignment and Data Set Establishment
All sequences obtained from the extracted samples were aligned
and edited using the software BioEdit 7.2.3 (
Hall, 1999;
http://www.mbio.ncsu.edu/BioEdit/bioedit.html) and Geneious
7.1.3 (Kearse et al., 2012). In addition to the s a mples extracted
and sequenced by the authors, currently available sequences
from GenBank were downloaded for wolves (C. lupus; Table S2).
However, since available and published CR-sequences varied to
a great extent both in overlap and length, two alignments were
initially constructed; one covering the complete stretch of 582 bp
(alignment A, n = 314) and a second where all sequences <347
bp were excluded (alignment B, n = 335). The delimitation of
the second alignment was made in order to allow for comparison
with as many ancient wolf sequences as possible, without cutting
the alignment too short and thereby losing information. A
third alignment was finally made where all ancient sequences
available were included, but which was restricted to a mere 57 bp
(alignment C, n = 366). The number of sequences from ancient
samples within each alignment was A, n = 10; B, n = 31; and
C, n = 62. Published sequences from ancient samples younger
than 30 kyr BP that were labeled as uncertain regarding affinity
to wolf or dog or as “doglike” were not included (
Verginelli et al.,
2005; Thalmann et al., 2013). Sequences containing uncertainties
at polymorphic sites, and thereby making assignments to known
haplotypes or identification of new ones impossible, were also
excluded. Uncertainties at non-polymorphic sites, only occurring
in a few sequences were accepted. This standard was set since the
Frontiers in Ecology and Evolution | www.frontiersin.org 3 December 2016 | Volume 4 | Article 134
Ersmark et al. Wolf Phylogeography and Demographic History
latter neither provided information for haplotype identific ation
nor were phylogenetically informative.
The targeted sequence in the first two alignments included
several insertion-deletions (i ndels), which were taken into
account for the haplotype assignment, but excluded in t he
following phylogenetic analyses, rendering alignment A and B
to cover 558 and 317 bp, respectively. Within the alignments the
geographical origin of the samples were indic ated at least to the
level of continent or region: North America, Europe, south-west
Asia, northern plus eastern Asia, India, Japan (Hokkaido), and
the H imalayas (including Tibet).
Haplotype Assignment for Alignment A
During analysis, considerable confusion on haplotype
designation was revealed. This was due to the varying
sequence length used in different studies and lack of unitary
designations for new and already known haplotypes. Here,
the sequence from
Björnerfeldt et al. (2006) was chosen as a
reference, because it represents one of the longest mitochondrial
sequences (16,730 bp) in this data set (Björnerfeldt et al., 2006).
Consequently, this sequence was trimmed to match our longest
alignment (A) and named Clu1 in order to establish a new
nomenclature for the gathered wolf sequences; new haplotypes
were named accordingly and published haplotypes were renamed
(Table S2). Additionally, the novel haplotypes in the wolves were
compared to all available dog and coyote sequences in order
to identify haplotypes shared among (sub) species (Pang et al.,
2009). The program DnaSP 5.10 (Librado and Rozas, 2009) was
used to identify identical haplotypes in alignment A.
Phylogenetic Analysis
A Bayesian phylogenetic analysis was also performed on
alignment A using BEAST 1.8.0 (
Drummond and Rambaut,
2007) and applying the HKY + Ŵ model of nucleotide
substitution without partitioning the alignment, which was
selected under the AIC criterion in jModelTest (Posada,
2008). Further, we assumed the constant population size
setting as a coalescent tree prior, which is suitable for trees
describing the relationships between individuals within t he
same population/species (Drummond and Rambaut, 2 007).
The posterior distribution of nodes, divergence times, and
substitution rates were estimated by Markov chain Monte Carlo
(MCMC), where samples were drawn every 1000 MCMC steps
from a total of 30 million steps, following a discarded burn-in
of 3 million steps. Convergence to the stationary distribution
and sufficient sampling were checked by inspection of posterior
samples and ESS values. Additionally, analyses were run twice
to test for stability and convergence of MCMC chains in plots
of posterior log likelih oods in Tracer v1.5.2 (
Drummond and
Rambaut, 2007). Since radiocarbon dates were available as
internal calibration points the “Estimate” option was used with
no prior on the substitution rate.
Network Analyses
Since bifurcat ed phylogenetic trees may not accurately mirror
the multi-furcated, reticulated relationships among intraspe cifi c
haplotypes (
Posada and Crandall, 2001), network analyses were
performed: Median-joining networks were constructed using the
software PopART 1.0 (
Leigh et al., 2012) for alignment A, B,
and C excluding indels, and temporal comparisons were made
between the Late Pleistocene and modern samples in the former,
as well as using five time layers within the set of ancient samples
(alignment C). Since alignment A had a limited representation
of sequences from ancient samples (n = 10) and analyses were
subsequently focused on the shorter alignments.
Coalescent Simulations
To test the population turnover hypothesized in previous studies
(
Leonard et al., 2007; Pilot et al., 2010), coalescent simulations
were carried out on two geographically delimited datasets
representing samples from Eurasia and North America. This
division was based on the separation of the two landmasses
by the Bering Strait in the early Holocene (Hu et al., 2010).
Alignment B was chosen for testing the simulations, since the
other two alignments were either lacking enough ancient samples
(alignment A) or were too restricted in length (alignment C).
Three alternative demographic histories were compared; a
constant population size through time, a population bottleneck,
and a split model designed to test the possibility of an extinction
and replacement of divergent ecotypes (i.e., haplogroups 2 and
1;
Leonard et al., 2007) or whether the haplotypes could have
originated in situ (Leonard et al., 2007). Priors for the models
were as follows: (1) constant model with a constant population
size (Ne
mod
) drawn from a uniform prior U (50,000–1,000,000)
through time, (2) bottleneck model where a population (Ne
anc
)
drawn from a uniform prior U (50,000–1,000,000) decreased
instantaneously to a smaller size (Ne
bot
) drawn from a uniform
prior U (1000–50,000) at a time (t
bot
) drawn from a uniform prior
U (4333– 4000) generations ago and then expanded exponentially
to a population size (Ne
mod
) drawn from a uniform prior
U (50,000–1,000,000), and finally (3) split model in which a
single population (Ne
anc
= (Ne
1_mod
+ Ne
2_mod
)) was split
at a time (t
split
) U (13,333–51 6,666 ) generations ago into two
populations; the first (Ne
1_mod
) sampled independently from a
uniform prior of U (25,000–500,000) and the second (Ne
2_mod
)
sampled independently from a uniform prior of U (25,000–
500,000) with a migration rate (m
1_2
) U (0.001–0.1) between the
two populations.
The prior for the mutation rate of U [1.04577 × 10
−5
–
4.0855 × 10
−5
(mutation rate for the analyzed sequence per
generation)] was based on a rate of 1.1–4.3%/million years (
Pang
et al., 2009) and the K2P (Kimura two-parameter; Kimura,
1980) mutation model was used with 0.957% transitions and a
gamma shape parameter of 0.0860 with 6 rate classes estimated
using ModelTest 3.7 (Posada and Crandall, 1998) . The timing
of the bottleneck as well as the split time was inferred from
the turnover, mutation rate and divergence between the two
haplogroups reported in Leonard et al. (2007), calculated with
a generation time of 3 years (Mech and Seal, 1987). Coalescent
simulations (1 million iterations of each model) were performed
using BayeSSC (
Anderson et al., 2005) and segre gating sites,
nucleotide diversity, Tajima’s D, and pairwise Fst were c alculate d
for each simulation and used for model choice, cross-validation,
Bayes factors, rejection, and local linear regression a djustment,
Frontiers in Ecology and Evolution | www.frontiersin.org 4 December 2016 | Volume 4 | Article 134
Ersmark et al. Wolf Phylogeography and Demographic History
which were carried out in R using the R package abc.R (Csillery
et al., 2012) based on the 1000 closest Euclidean distances
between simulated and observed summary statistics. A PCA
was performed on simulations for each model prior to analysis
to check the appropriateness of the simulations and models
for producing summary statistics close to the obser ved and
pseudo-observed datasets (PODs) confirmed that the constant,
bottleneck, and split model could be differentiated given our
sampling and dataset.
RESULTS
Control Region Diversity
In total, we found 114 different wolf haplotypes among 314
sequences in alignment A. Thirteen of these were reported the
first time; four from North America, five from Siberia, one from
Europe (Russia), one from Iran, and two from China (Table 1 and
Table S1; GenBank accession numbers: KX898307-KY124130).
The two ancient wolf specimens from Siberia both represented
new unique haplotypes, whereas the six samples from Greenland
all belonged to a single haplotype (Clu53) previously found
among North American wolves. Three haplotypes were identified
in wolves sampled from the Great L akes region (Canis l. lycaon),
one of which was shared with coyotes (Table S1). After initial
trials of haplotype clustering, these wolves were repeatedly placed
in the most basal clade and showed a kinship closer to the coyote,
which supports its proclaimed status as a hybrid or a separate
species (
Fain et al., 2010; vonHoldt et al., 2011). Consistent
with a hybrid origin, a recent whole-genome sequence study
demonstrated extensive gray wolf and coyote introgression in
wolves from the Great lakes region (
vonHoldt et al., 2016).
After this apparent affiliation was revealed, the samples of
eastern Canadian wolves were left out from the subsequent
analyses.
In alignment B, 90 haplotypes were found among 335
sequences a nd were assigned with haplotype numbers. As a
result of th e shorter sequence length se veral haplotypes that were
distinct in alignment A collapsed into larger ones in this shorter
alignment. One was even shared by both ancient and modern
samples when indels were removed (haplotype nr. 8 and 9; Clu8,
TABLE 1 | Genetic diversity according to geographical region based on
alignment A excluding indels (558 bp); Hd = haplotype diversity, π =
nucleotide diversity.
Region n Sites Segr. Sites Haplotypes Hd π (%)
N+E Asia 95 558 49 37 0.94 1.5
Europe 94 558 37 33 0.93 1.4
N America 69 558 25 17 0.88 0.8
SW Asia 22 558 26 12 0.89 1.3
Ancient 10 558 24 10 1.0 1.3
India 5 558 2 2 0.40 0.14
Himalaya 19 558 2 2 0.11 0.04
N+E Asia = Siberia, China, Mongolia, and Korea. SW Asia = Israel, Saudi Arabia, Oman,
Iran, Afghanistan.
Clu9, Clu10, Clu22, Clu108, and Clu109). Alignment C contained
71 haplotypes among 366 sequences.
Bayesian Analysis (BEAST)
The Bayesian analysis in BEAST showed convergence of
posterior lik elihoods between runs. For all parameters of int erest,
the effective sample sizes (ESS) were higher than 2 00 (as
recommended in the BEAST manual), suggesting stabilization
and good mixing of the MCMC chains.
The phylogeny revea led an overall pattern where clades
represented samples from multiple locations, even from
geographical regions distant to one another (Figure 1: I–XIX).
There were also several recurrently supported clades, which
showed c onsiderable geographical unity (III, VII, XIII, XVI,
and XVIII). The most basal clades were represented by two
lineages of Himalayan and Indian wolves, known for their early
divergence among wolf lineages as has previously been shown
(
Sharma et al., 2004; Aggarwal et al., 2007; Meng et al., 2009).
Within the clade including the Himalayan wolves was also a
distinct subclade (Clu76) with nine wolves from China and
Mongolia (Figure 1). In the remaining phylogeny, all ancient
wolves were confined among the basal clades and the two new
samples from Siberia aligned next to two of these groups (XVIII
and XV).
Among North American wolves, three groupings (II, VI,
VIII) were discernable clearly at the upper half of the tree,
largely matching the recent findings of clusters of wolf haplotypes
in Canada and Alaska (We ck worth et al., 2011). The arctic
wolves from Greenland were grouped within the largest group
representing American wolves at the top of the tree (II). Two
American haplotypes stand out by their more basal position
in the phylogeny. One represents Mexican wolves (XI/Clu30),
which are currently only found in small captive populations
in the US, and which previously have been shown to form
a genetically distinct group (
Vila et al., 1999). The oth er
is that of a wolf from Vancouver Island in south-western
Canada. However, this latter haplotype (Clu47) is also shared
by dogs, suggesting a recent h ybridizati on, which has also been
previously reported from this region (Munoz-Fuentes et al.,
2010
).
European samples are found throughout the phylogeny, but
also form distinct groups. Two notable examples consist of
Spanish (IX/Clu35, 36) and Italian wolves (XVIII/Clu30), which
are found widely apart in the phylogeny. Furthermore, the
Spanish haplotypes (Clu35, 36) from the former group are unique
to t he Iberian Peninsula.
The Asian samples are widely spread in the phylogeny as
well, especially the ones representing northern and eastern Asia,
which are found all across both trees and networks. Three of
the new haplotypes reported from th is area originated from the
easternmost regions; Clu66 from the Chukotka Peninsula, Clu67
from Khakassia and the Amur region and Clu72 from the North
Korean border, and they all show a great diversity wit h close
affiliation to wolves on all three continents. The novel haplotype
Clu72 also included a distinct 11 bp insertion, not previously
reported.
Frontiers in Ecology and Evolution | www.frontiersin.org 5 December 2016 | Volume 4 | Article 134