Coalescence patterns of endemic Tibetan species of stream ...labs.eeb.utoronto.ca/murphy/PDFs of...
Transcript of Coalescence patterns of endemic Tibetan species of stream ...labs.eeb.utoronto.ca/murphy/PDFs of...
Molecular Ecology (2012) 21, 3308–3324 doi: 10.1111/j.1365-294X.2012.05606.x
Coalescence patterns of endemic Tibetan speciesof stream salamanders (Hynobiidae: Batrachuperus)
BIN LU,*† YUCHI ZHENG,* ROBERT W. MURPHY‡§ and XIAOMAO ZENG*
*Department of Herpetology, Chengdu Institute of Biology, Chinese Academy of Sciences, Chengdu, Sichuan 610041, China,
†Graduate School of the Chinese Academy of Sciences, Beijing 100049, China, ‡Department of Natural History, Royal Ontario
Museum, 100 Queen’s Park, Toronto, ON M5S 2C6, Canada, §State Key Laboratory of Genetic Resources and Evolution,
Kunming Institute of Zoology, the Chinese Academy of Sciences, Kunming 650223, China
Corresponde
E-mail: zengx
Abstract
Orogenesis of topographically diverse montane regions often drives complex evolution-
ary histories of species. The extensive biodiversity of the eastern edge of the Tibetan
Plateau, which gradually decreases eastwardly, facilitates a comparison of historical
patterns. We use coalescence methods to compare species of stream salamanders
(Batrachuperus) that occur at high and low elevations. Coalescent simulations reveal that
closely related species are likely to have been influenced by different drivers of
diversification. Species living in the western high-elevation region with its northsouth
extending mountains appear to have experienced colonization via dispersal followed by
isolation and divergence. In contrast, species on the eastern low-elevation region, which
has many discontinuous mountain ranges, appear to have experienced fragmentation,
sometimes staged, of wide-ranging ancestral populations. The two groups of species
appear to have been affected differently by glaciation. High-elevation species, which are
more resistant to cooler temperatures, appear to have experienced population declines as
recently as the last glaciation (0.016–0.032 Ma). In contrast, salamanders dwelling in the
warmer and wetter habitats at low-elevation environs appear to have been affected less
by the relatively recent, milder glaciation, and more so by harsher, extensive glaciations
(0.5–0.175 Ma). Thus, elevation, topography and cold tolerance appear to drive evolu-
tionary patterns of diversification and demography even among closely related taxa. The
comparison of multiple species in genealogical analyses can lead to an understanding of
the evolutionary drivers.
Keywords: coalescent simulation, diversification drivers, ecological niche modelling, elevation
effects, historical demography, Tibetan uplift
Received 9 June 2011; revision received 28 February 2012; accepted 12 March 2012
Introduction
Tectonic movement, orogenesis and climatic change can
drive species diversification especially via allopatric
speciation. For terrestrial species, this can occur by the
formation of sky islands in which dry, hot, deep valleys
serve as barriers to gene flow (Knowles 2000), as well
as the height of mountains forming a barrier to dis-
persal for species that live in the valleys (Craw et al.
2008). For aquatic species, orogenesis can divide and
nce: Xiaomao Zeng, Fax: +86 028 85222753;
change the courses of rivers (Haffer 2008). Such drivers
are exemplified by one region, the Tibetan Plateau (Che
et al. 2010; Zhang et al. 2010). This region’s extensive
biodiversity and well-studied complex geology facilitate
investigations on the genetic consequence of orogenesis.
For example, the eastern edge of the Tibetan Plateau
contains major features that might have played impor-
tant roles in the divergence of some endemic verte-
brates (Yu et al. 2000; Luo et al. 2004; Peng et al. 2006).
However, the processes that shaped the patterns remain
to be confirmed. Are all extant populations derived
from the fragmentation of a widely distributed common
ancestral population, sequentially staged isolation
� 2012 Blackwell Publishing Ltd
PHYL OGEOGRAPHY OF TIBETAN SALAMANDERS 3309
events or colonization via a series of dispersals followed
by isolation and divergence?
Genetic patterns can identify historical events and the
associated driving processes. For example, the Euro-
pean alpine system is topographically complex, and the
resident species have correspondingly complex evolu-
tionary histories and patterns of divergence (Despres
et al. 2002; Kropf et al. 2003). Topography can drive
divergence patterns (Smith & Farrell 2005) and Pleisto-
cene climatic fluctuations associated with cyclical glacia-
tion events can also be involved, even in lower
latitudes (Bryson et al. 2011a,b). The eastern edge of the
Tibetan Plateau is extremely complicated geologically. It
is composed of several nonuniform landform assem-
blages, yet it can be defined as having two main topo-
graphical areas: the western highland and eastern
lowland. The highland area includes various northsouth
extending mountain ranges, such as the Shaluli and Da-
xue mountains. High-elevation montane taxa might
have dispersed either from north to south or vice versa
along the crest of the mountain ranges during orogene-
sis of the Tibetan Plateau and in conjunction with asso-
ciated climatic changes. In contrast, the lowland area is
comprised of many isolated mountains, such as Emei,
Luoji and Bailing mountains, that are separated by deep
valleys. The physiographic structure would probably
inhibit dispersal and promote fragmentation, especially
for taxa that have limited dispersal abilities. The more
continuous high-elevation habitat and ongoing orogene-
sis in the west would be expected to lead to a pattern
of sequential isolation when compared with a more uni-
form early fragmentation pattern in the east. Therefore,
the two topographical developments may be the drivers
of contrasting patterns of genetic divergence.
The effects of Pleistocene glacial cycling on popula-
tions depend on latitude and topography. Responses to
climatic oscillations appear to vary among species as a
function of cold tolerance (Hewitt 1996, 2004). Whereas
heavy ice is known to have covered high-latitude conti-
nental Europe and North America, glaciations occurred
discontinuously on high mountains of the eastern edge
of the Tibetan Plateau; low-elevation habitats are
thought to have been ice free (Shi et al. 1998; Zhou
et al. 2006). Climatic oscillations appear to have miti-
gated demographic stresses for species at lower relative
to higher elevations (Qu et al. 2010).
Orogenesis of the Tibetan Plateau is known to have
continued during the Late Pleistocene. A combination
of uplifted terrain and lowered snowlines during this
period may be responsible for previously low, glacier-
free regions becoming glaciated (Shi et al. 1997; Liu
et al. 1999; Zheng et al. 2002). Whereas climatic shifts in
montane regions can affect the distributions of temper-
ate species, some cold-tolerant taxa may persist in ice-
� 2012 Blackwell Publishing Ltd
free areas (Crespi et al. 2003; Smith & Farrell 2005;
Shepard & Burbrink 2008). Variation in lineage-specific
cold tolerance complicates the analyses of elevational
patterns because all taxa do not respond equally to cli-
matic changes.
Species of Batrachuperus (family Hynobiidae) are pri-
marily restricted to isolated montane regions of the
Tibetan Plateau’s eastern edge. These salamanders have
weak dispersal abilities, and they are likely to have
experienced several orogenic events and glacial cycles
(Zhang et al. 2006). The well-established phylogeny (Fu
& Zeng 2008) provides a solid foundation for testing
hypotheses regarding the ecological and physiographic
drivers of genetic divergence. The species appear to be
parapatric, yet they are separated by elevation. Eleva-
tional differences between sister species provide an
opportunity to investigate responses to topographical
variation and Pleistocene glaciation, the two most
important factors influencing the current spatial distri-
butions of species occurring in the Tibetan Plateau and
places of similar topography (Hewitt 2000; Zhang &
Jiang 2006).
Recent methodological developments facilitate the
ability to infer historical processes. In particular,
hypothesis tests employing coalescent-based methods
allow for the statistical evaluation of the fit of diver-
gence timing to various historical scenarios while taking
into account the stochasticity of genealogical relation-
ships (Knowles & Maddison 2002; Knowles 2004, 2009).
Further, ecological modelling allows for the identifica-
tion of significant differences in the environments of
species and enables the prediction of suitable environ-
mental conditions (Carstens & Richards 2007; Shepard
& Burbrink 2008).
Herein, we compare the patterns of diversification
and demography in species of Batrachuperus living at
high and low elevations using coalescent-based meth-
ods and a combination of historical demography and
ecological niche modelling. Specifically, we test hypoth-
eses involved in the following questions: (i) How was
divergence driven in closely related, high- and low-
elevation species? (ii) How did these species respond to
glacial cycles as a result of elevation and cold-tolerant
physiology?
Material and methods
Species sampling and lab protocols
We selected five species of Batrachuperus distributed in
the montane eastern edge of the Tibetan Plateau (Fu &
Zeng 2008) including B. karlschmidti, B. londongensis,
B. pinchonii, B. yenyuanensis and an undescribed spe-
cies; the undescribed species, mainly distributed in the
3310 B . LU ET AL.
central and southern Shaluli Mountains, was referred to
herein as ‘Daocheng species’. It corresponded to Fu &
Zeng’s (2008) B. sp. 2. Batrachuperus karlschmidti and
Daocheng species occurred at western high elevations
(3500–4300 m) such as Shaluli and Daxue mountains,
while B. londongensis, B. pinchonii and B. yenyuanensis
were mainly in eastern lower reaches (1800–3200 m),
such as Emei, Luoji and Bailing mountains. A total of
418 specimens of Batrachuperus from 54 locations were
sampled between 1999 and 2009 from throughout the
species’ distributions (Fig. 1; Appendix S1, Supporting
information). We also downloaded previously pub-
lished sequences of Batrachuperus tibetanus (GenBank:
DQ333817), Pseudohynobius shuichengensis (GenBank:
FJ532060), Hynobius quelpaertensis (EF201847), Liua shihi
(DQ333810), Pachyhynobius shangchengensis (DQ333812),
Paradactylodon mustersi (DQ333821) and Ranodon sibiricus
(AJ419960) for the subsequent phylogenetic analysis
and age estimation.
Total genomic DNA was extracted using the high-salt
method (Aljanabi & Martinez 1997). Liver or muscle tis-
sue of alcohol-preserved specimens was minced, 20 lL
20 mg ⁄ mL Proteinase-K was added and then the mix-
ture was incubated at 56 �C for 6 h. Three mitochon-
drial gene fragments were sequenced: cytochrome b
(Cytb), cytochrome c oxidase subunit I (COI) and
NADH dehydrogenase 5 (ND5). Slightly modified prim-
ers Bacytb1F (5¢-GAACTAATGGCCCACCCAATTCG
AA-3¢) and Bacytb1R (5¢-AAYARAAAATATCAYTCYG
GTTGRAT-3¢) based on MVZ15 and MVZ16 were used
for amplifying Cytb (Mueller et al. 2004). Using con-
served portions of complete mitochondrial genomes of
stream salamanders (Zhang et al. 2006), we designed
two pairs of primers. Bacox1-F9 (5¢-CACTCGATGAC
TATTYTCAACTAATCATA-3¢) and Bacox1-R943 (5¢-GGATRGCAATAATTATTGTGGCGGA-3¢) were used
for amplifying COI, and Band5-F1 (5¢-ATGAATTC
AGTATTAATTTTTAATTC-3¢) and Band5-R939 (5¢-TAAYCCTAATTGGCTTGAAGTTGA-3¢) were used for
amplifying ND5. A reaction volume of 50 lL contained
2 mM MgCl2, 0.2 mM of each dNTP, 0.2 lM of each primer
and 2.5 U of Taq DNA polymerase (TianGen Biotech)
with 20–100 ng template DNA. PCR involved 5 min at
94 �C, then 35 cycles with 94 �C for 30 s, 50 �C (Cytb) or
55 �C (COI and ND5) for 40 s and 72 �C for 1 min, fol-
lowed by incubation for 10 min at 72 �C. The amplified
DNA products were purified, and automated DNA
sequencing was performed on an ABI3730 with an ABI
PRISM BigDye terminator Cycle Sequencing Ready Reac-
tion Kit (PerkinElmer Biosystems). The three fragments
were concatenated into one data set. All sequences were
aligned using BIOEDIT 7.0.5 with the CLUSTAL W option
and default multiple alignment parameters. Haplotype
identification, nucleotide diversity (p) and haplotype
diversity (Hd) per lineage were estimated in DNASP
5.10.01 (Librado & Rozas 2009) (Table 1).
Genealogical reconstruction
Bayesian inference analyses (BI) were performed using
MRBAYES 3.1.2 (Huelsenbeck & Ronquist 2001; Ronquist &
Huelsenbeck 2003). Three partitioning strategies were
applied to our data set to determine the optimal strategy.
First, we defined the concatenated data set as one parti-
tion. Second, we defined a separate partition for each
gene (three partitions). Third, we divided the data set into
nine partitions according to each codon position in each
gene. The multiple partitions controlled for heterogeneity
Fig. 1 Map showing collecting localities
for salamanders of the genus Batrachupe-
rus on the eastern edge of the Tibetan
Plateau, southwestern China (see
Appendix S1, Supporting information
for details).
� 2012 Blackwell Publishing Ltd
Table 1 Genetic diversity and demographic statistics on salamanders of the genus Batrachuperus
Species Lineage n h p Hd SSD r
B. karlschmidti Daxue Mtn C 19 12 0.00507 0.924 0.03210 0.05475
Gan Zi 50 23 0.00529 0.932 0.01171 0.01581
Xin Long 18 7 0.00354 0.569 0.37703*** 0.12986
Shaluli Mtn C 7 6 0.00444 0.952 0.04567 0.05669
Shaluli Mtn S 11 6 0.00695 0.891 0.12104 0.12959
Total 105 54 0.01850 0.969 0.01070 0.00509
B. londongensis Emei Mtn 9 6 0.00136 0.722 0.07184 0.17284
Wawu Mtn 21 5 0.00039 0.724 0.01032 0.12000
Han Yuan 14 7 0.00162 0.670 0.16500* 0.40370**
Total 44 18 0.00952 0.871 0.07209** 0.06890**
B. pinchonii Qionglai Mtn NW 7 4 0.00154 0.810 0.07424 0.24490
Qionglai Mtn SE 39 17 0.01630 0.889 0.05308* 0.03988***
Emei Mtn 6 4 0.00108 0.600 0.50667*** 0.20444
Daxue Mtn S 25 15 0.00454 0.910 0.02718 0.06106*
Total 77 40 0.02893 0.959 0.00938 0.00911***
Daocheng species Shaluli Mtn SE 15 10 0.00201 0.933 0.00947 0.03619
Shaluli Mtn S 7 5 0.00629 0.857 0.87982*** 0.31066
Yulongxue Mtn 37 26 0.00938 0.979 0.01766 0.00919
Shaluli Mtn C 71 9 0.00073 0.560 0.08795 0.23995
Total 130 50 0.02320 0.867 0.04047* 0.02748***
B. yenyuanensis Daxue Mtn S 11 4 0.00184 0.745 0.15509 0.34116*
Bailing Mtn 22 18 0.00445 0.983 0.00651 0.01008
Luoji Mtn 29 17 0.00273 0.938 0.03614 0.03351
Total 62 39 0.01675 0.977 0.01440** 0.00432
n, number of individuals; h, number of haplotypes; p, nucleotide diversity; Hd, haplotype diversity; SSD, sum of square deviation
(goodness of fit to a simulated population expansion); r, raggedness index (*P < 0.05, **P < 0.01, ***P < 0.001).
PHYL OGEOGRAPHY OF TIBETAN SALAMANDERS 3311
across data set, such as variation in substitution rates. The
best-fitting substitution model for each partition was
selected using JMODELTEST 0.1.1 (Posada 2008) (Table 2).
Bayesian information criterion (BIC) was used to select a
model because of its high accuracy and precision (Luo
et al. 2010). For each partition strategy, Markov chains
were run for 10 million generations with two parallel
searches using one cold and three heated chains, each
starting with a random tree. Trees were sampled every
1000 generations using split frequencies <0.01 to indicate
convergence. TRACER 1.5 (Drummond & Rambaut 2007b)
was used to determine when the log likelihood (lnL) of
sampled trees reached a stationary distribution. In all
Bayesian analyses, apparent stationarity of Markov Chain
Monte Carlo (MCMC) runs was reached within 1 million
generations; we conservatively discarded the first 5 mil-
lion generations from each run as ‘burn-in’ and used the
sampled trees from the remaining 5 million generations
(5001 trees) to calculate the frequency of nodal resolution
in a 50% majority-rule consensus tree, termed Bayesian
posterior probabilities (BPPs). Three replicate analyses
were conducted to assess whether or not individual runs
failed to converge upon the optimal posterior distribution
and if likelihood values, branch lengths, tree topology
and posterior probabilities differed between runs. The rel-
ative fit of different partitioning strategies was evaluated
� 2012 Blackwell Publishing Ltd
using Bayes factors (BF) in TRACER. A BF > 10 was consid-
ered strong evidence for using a more-partitioned model
(Kass & Raftery 1993). Maximum-likelihood (ML) analy-
ses were performed under the BF-preferred partitioning
strategy using the GTR+G model in the program RAXML
7.0.3 (Stamatakis 2006). Additional ML analyses were run
using the alternative partitioning schemes to check for
sensitivity to BF. A likelihood ratio test was performed to
verify the BF-preferred partitioning strategy. For estimat-
ing nodal support, nonparametric bootstrap proportions
(Felsenstein 1985) with 500 replicates were used. Maxi-
mum-likelihood bootstrap proportions (MLBs) ‡70% and
BPPs ‡0.95 were considered strong support (Hillis & Bull
1993; Erixon et al. 2003; Huelsenbeck & Rannala 2004).
The approach potentially assessed discordance among
trees based on the implemented method of inference and
provided a preferred topology for divergence estimates.
Divergence time estimation
Topology, divergence dates and substitution rates were
simultaneously estimated using BEAST 1.6.1 (Drummond
et al. 2002; Drummond & Rambaut 2007a). The best-
fitting substitution model for each data partition under
the BF was chosen as in the BI analyses. Application of
the optimal model avoided over-parameterization of the
Table 2 Nucleotide substitution models used in genealogical reconstruction, divergence date estimation (intra- and interspecific),
demographic history estimation and coalescent simulations. The estimation of h and Ne was also presented
Partition
Population level Species level
Best model Pinvar Gamma Best model Pinvar Gamma
ND5_1 GTR+G — 0.241 GTR+I+G 0.560 1.806
ND5_2 GTR+I+G 0.453 0.249 HKY+I 0.823 —
ND5_3 HKY+G — 1.427 HKY+G — 1.523
Cytb_1 SYM+G — 0.157 K80 + G — 0.088
Cytb_2 HKY+I 0.817 — HKY+G — 0.010
Cytb_3 GTR+G — 2.497 HKY+G — 2.095
COI_1 K80 + I+G 0.622 0.239 K80 + I+G 0.627 0.229
COI_2 F81 — — F81 — —
COI_3 GTR+G — 1.737 HKY+G — 1.945
BSPs and coalescent simulations
Best model Pinvar Gamma h (95% CI) Ne (95% CI)
Batrachuperus karlschmidti GTR+I+G 0.645 0.305 0.0240 (0.0126–0.0512) 993 789 (521 739–2 120 083)
Batrachuperus londongensis HKY — — 0.0039 (0.0014–0.0094) 161 491 (57 971–389 234)
Batrachuperus pinchonii GTR+G — 0.011 0.0181 (0.0104–0.0341) 749 482 (430 642–1 412 008)
Daocheng species HKY+I+G 0.536 0.305 0.0239 (0.0141–0.0452) 989 648 (583 851–1 871 636)
Batrachuperus yenyuanensis HKY+I+G 0.319 0.018 0.0150 (0.0085–0.0268) 621 118 (351 967–1 109 731)
BSP, Bayesian skyline plots; COI, cytochrome c oxidase subunit I; Pinvar, proportion of invariable sites; Gamma, gamma shape
parameter; CI, confidence interval.
3312 B . LU ET AL.
model and reduced the trade-off between bias and
unnecessary variance (Kelchner & Thomas 2007). Prior
to the BEAST analysis, we ran analysis using different
clock models (non-clock, strict clock and relaxed clock)
in MRBAYES to determine whether our data set was
evolving in a clocklike fashion and which clock model
was more appropriate for our data set. For each clock
model, Markov chains were run for 10 million genera-
tions. Trees were sampled every 1000 generations. BF
was used to select among the models and to test the
null hypothesis of a linear rate of change (Weisrock
et al. 2001). For the BEAST analysis, an initial MCMC was
run 20 million generations with an uncorrelated lognor-
mal clock model and a constant population size prior to
determine the optimal parameter in BEAST. TRACER was
used to examine the parameter ucld.stdev to test for
departures from the molecular clock. A parameter value
near zero suggested that our data evolved in a clocklike
manner. An additional BEAST analysis with an enforced
strict clock was performed to compare the result using
the relaxed clock approach. The final MCMC was run
for 20 million generations while sampling every 1000
generations. Five independent runs achieved the same
result. Burn-in and convergence of the chains were
determined with TRACER. Further, effective sample sizes
(ESS) were required to have values greater than 200.
Divergence times for Batrachuperus were estimated
using two different methods: a calibration method and
a mean substitution rate method. First, we used the fos-
sil record and previously established divergence times
among hynobiid genera (Zheng et al. 2011) to calibrate
the clock. The split between Ranodon and Paradactylodon
was constrained to be at least 2.6 Ma based on the fossil
Ranodon cf. sibiricus reported from the Upper Pliocene
(Averianov & Tjutkova 1995). We did not download
data for Andrias davidianus because the commonly used
cryptobranchid fossil (c. 145 Ma) (Gao & Shubin 2003;
Marjanovic & Laurin 2007; Zhang & Wake 2009; Roe-
lants et al. 2011; Zheng et al. 2011) was probably too
old to be directly applicable to our younger taxa (Zhang
et al. 2006), and it might have caused an overestimation
of the divergence dates (Hugall et al. 2007; Zheng et al.
2011). Slowly evolving nuclear exons have been sug-
gested to be appropriate to correct potential bias in esti-
mating divergence dates, especially for a large-timescale
phylogeny (Diego 2010; Zheng et al. 2011). Conse-
quently, we used the divergence dates recently reported
by Zheng et al. (2011) based on nuclear exons. Conser-
vatively, we estimated priors for the divergence
between Hynobius and Pseudohynobius-Liua-Batrachuperus
to be 30 ± 14 Ma (mean ± 95% CI), and among Batra-
chuperus to 13 ± 8 Ma, which followed normal distribu-
tions. We applied all three constraints in this analysis.
Both intra- and interspecific hierarchical levels were
included in our divergence time analysis. Rates of long-
term (species level) substitutions have been reported to
� 2012 Blackwell Publishing Ltd
PHYL OGEOGRAPHY OF TIBETAN SALAMANDERS 3313
often strongly conflict with short-term (population
level) substitutions. Thus, results based on implied rates
of change across different hierarchical timescales have
been suggested to be invalid and often overestimated
(Ho et al. 2005). To investigate the possible effect this
phenomenon might have had on our results, we also
performed BEAST analyses using only species-level sam-
pling with calibration. The Yule process was used to
describe speciation under relaxed clock model. The
results were used as interspecific times of divergences
for Batrachuperus. The divergence dates and substitution
rates estimated using this calibration method were
applied to subsequent coalescent simulations.
To compare the results from the calibration methods,
we calculated divergence dates using an estimated mean
substitution rate (0.64% per MY per lineage) for our
combined sequence data. Yoshikawa et al. (2008)
reported a substitution rate of 0.68% per MY per lineage
for Cytb in another hynobiid (Onychodactylus japonicas).
This rate was similar to that estimated for other salaman-
ders, for example an average rate of 0.62% per MY per
lineage for plethodontids (Mueller 2006). The average
net p-distance for the combined sequences vs. Cytb
alone was 0.947. Therefore, the estimated mean substitu-
tion rate for the combined data was obtained by multi-
plying the Cytb rate of O. japonicas (0.68%) by the
p-distance ratio for Cytb alone (0.947).
Coalescent simulations
Coalescent simulations were conducted in MESQUITE
2.7.2 (Maddison & Maddison 2010). These were used to
both investigate the possible drivers of diversification
and test whether or not the timing of diversification
within each species of Batrachuperus was consistent with
the most recent uplift of the Tibetan Plateau and to test
the climatic shift hypotheses.
Three possible models of diversification were tested:
fragmented ancestor, staged fragmentation and coloni-
zation. Dates for hypotheses testing corresponded to the
approximate age estimated for each species in the BEAST
analysis, the most intensive and frequent uplift of the
Tibetan Plateau (Li et al. 1996; Li & Fang 1999; Wu
et al. 2001), and the initial Pleistocene climatic changes
(2.6–3.6 Ma). These absolute times (years) were con-
verted to coalescent times (generations) assuming a gen-
eration time of 3.5 years for the species of Batrachuperus
(Han & Lu 1999). ML estimations of Ne (inbreeding
effective population size) used h-values were calculated
in the program MIGRATE-N 3.2.1 (Beerli & Felsenstein
1999, 2001). Using the 0.69% per lineage substitution
rate produced in the BEAST analysis and a generation
time of 3.5 years, the h-value was then converted to Ne
using the formula h = Ne*l for mitochondrial DNA,
� 2012 Blackwell Publishing Ltd
with l = 2.42 · 10)8 (3.5 · 0.69 · 10)8) substitutions per
site per generation (Table 2). For the coalescent simula-
tions, the overall Ne was set to empirically estimated
values. The Ne of the putative regional population was
constrained to a size proportional to the overall Ne at
any single point on a branch. DNA sequences were sim-
ulated under the same evolutionary model of the
empirical data, as selected by JMODELTEST 0.1.1 (Posada
2008) (Table 2). For each simulation, 100 coalescent
genealogies constrained within different hypotheses of
population history were simulated under a neutral coa-
lescent process. PAUP* 4.0b10 (Swofford 2002) was used
to construct trees from the simulated matrices and to
generate one consensus tree representing genes for each
simulation. We fit the simulated gene trees to each pos-
sible divergence model of population trees with differ-
ent date hypotheses and calculated the amount of
discordance between the gene trees and population
trees as measured by number of deep coalescences (DC)
(Maddison 1997). We then compared DC values from
the empirically reconstructed gene trees to those gener-
ated from the reconstructed gene trees from simulated
data. The null model was rejected when the observed
DC value fell below 95% of the distribution of the
expected values.
The area of origin for each species of Batrachuperus
was estimated from reconstructed ancestral states using
MESQUITE 2.7.2 (Maddison & Maddison 2010). Each indi-
vidual was assigned to its collecting location. Because
no prior knowledge of dispersal rate existed, the Mar-
kov k-state 1 parameter (Mk1) model with equal likeli-
hood the rate of change among different regions was
selected for estimating ancestral areas.
Historical demography
Past population dynamics of all lineages were estimated
using coalescent-based Bayesian skyline plots (BSP;
Drummond et al. 2005) as well as mismatch distribu-
tions (Rogers & Harpending 1992). BSPs were imple-
mented in BEAST. They were used to estimate the
dynamics of past populations through time without
requiring a prespecified parametric model of demo-
graphic history. Uncertainty in the genealogy was con-
trolled using a Bayesian approach under a coalescence
model. The best-fitting model (Table 2) with a relaxed
lognormal clock was selected to construct the BSP in
BEAST for each lineage. Chains were run for 10 million
generations, sampled every 1000 generations, and the
first 10% of the trees were discarded as burn-in. The
results were summarized using TRACER. Mismatch dis-
tributions were calculated using ARLEQUIN 3.5 (Excoffier
& Lischer 2010). Multimodal distributions were
expected in populations at demographic equilibrium or
3314 B . LU ET AL.
in decline, and unimodal distributions were anticipated
in populations having experienced a recent demo-
graphic expansion (Slatkin & Hudson 1991; Rogers &
Harpending 1992). The expected distributions were gen-
erated by bootstrap resampling (10 000 replicates) using
a model of sudden demographic expansion. The sum of
square deviations and raggedness index between the
observed and the expected mismatch were used as a
test statistic. P-values were calculated as the probability
of simulations producing a greater value than the
observed value.
Species habitat modelling
The relationship between species occurrences and the
corresponding environment was estimated from a dis-
tribution model using MAXENT 3.3.3a with default set-
tings (Phillips et al. 2006; Elith et al. 2011). The 19
bioenvironment-associated variables with 30 arc-sec res-
olution (�1 km2) were downloaded from the high-reso-
lution interpolated WORLDCLIM database (http://
www.worldclim.org/) (Hijmans et al. 2005). The vari-
ables were used in habitat modelling to examine for
potential factors that limited the species’ distributions.
Principal components analysis was used to convert a
set of possibly correlated environmental variables into a
set of values of independent variables. Principal compo-
nents with eigenvalues >1 that explained >10% of the
variation were retained. These factor scores were used
in a multivariate analysis of variance (MANOVA) as
implemented in SPSS version 19 to test for differences
among the five species of Batrachuperus, and the multi-
variate tests for each principal component were per-
formed to determine significant differences in
environmental conditions between species because these
may have been the drivers responsible for demographic
and divergence patterns.
Results
Topology and divergence dates
We resolved 764 bp of Cytb, 848 bp of COI and 856 bp
of ND5. The concatenated data set comprised 2468
aligned sites. Sequences were deposited in GenBank
under accession nos JQ303644–JQ304246. Both the BF
(>509) and likelihood ratio test (P < 0.001) strongly
favoured the nine-partitioning strategy against the one-
and three-partitioning strategies. Therefore, the nine-
partitioning strategy was used for all analyses of the
concatenated data using MRBAYES, RAXML and BEAST.
Bayesian inference and ML produced identical overall
topologies. Thus, only the ML tree [Fig. 2a] with support
values was presented. However, the topology among lin-
eages of B. pinchonii in the BEAST tree differed from those
in the MRBAYES and ML trees. In the BEAST tree, the lineage
Qionglai Mtn NW clustered with all other lineages of
B. pinchonii (BPPs, 1.0). In contrast, the lineage Qionglai
Mtn NW clustered with Qionglai Mtn SE 2 [Fig. 2a;
BPP = 0.96; MLB = 75%]. We then used BF comparisons
to test alternative topologies among lineages of B. pincho-
nii. BI incorporating a topological constraint was per-
formed with the same model. Markov chains were run
for 10 million generations. Trees were sampled every
1000 generations. BFs provided strong evidence
(BF > 14) for Qionglai Mtn NW and Qionglai Mtn SE 2
being sister lineages. Thus, the BI and ML topologies
were preferred, and all subsequent BEAST analyses used a
topological constraint. This disorder may have been
caused by being trapped in local optimum of MCMC
(Huelsenbeck & Ronquist 2001).
BF strongly favoured the relaxed clock model over
the nonclock model (BF > 268) and the strict clock
model (BF > 144), suggesting that the relaxed clock
model was more appropriate for our data. Thus, an un-
correlated lognormal clock model was imposed for
BEAST analyses. These estimated divergence times were
then employed to get a general idea of the timescale of
diversification as a reference for the subsequent coales-
cent simulations. However, a nearly zero ucld.stdev
parameter was estimated, and similar age estimations
were obtained for most nodes after running BEAST with
both the relaxed and strict clock models (Table 3).
Therefore, the analyses failed to reject the hypothesis of
a linear rate of change.
Time estimations based on calibration points indi-
cated that the earliest intraspecific divergences within
Batrachuperus occurred from the Late Miocene c. 7.2 Ma
(95% CI, 5.4–9.1 Ma) to the Early Pleistocene c. 2.0 Ma
(95% CI, 1.3–2.8 Ma; Fig. 2b; Table 3). These estimates
were younger than those obtained using a mean substi-
tution rate of 0.64% per MY per lineage. This rate was
inferred from Cytb in the Japanese clawed salamander,
and this application could not be validated indepen-
dently. Thus, we did not use dates estimated by the
mean rate method for subsequent coalescent simula-
tions. The BEAST analyses using only species-level sam-
pling estimated divergence times ranging from the
Middle Miocene c. 14.9 Ma (95% CI, 11.9–18.0 Ma) to
the Late Miocene c. 8.5 Ma (95% CI, 6.3–10.9 Ma). This
timescale was similar to that using both intra- and
interspecific hierarchical levels (Table 3). Thus, the dif-
ference between long- and short-term substitution rates
probably did not heavily affect our data set (Ho et al.
2005). The mean substitution rate produced by the cali-
bration method was 0.69% (95% CI, 0.58%–0.82%) per
MY per lineage. This rate was then used in subsequent
coalescent simulations.
� 2012 Blackwell Publishing Ltd
(a)
(b)
62.46
98
88
99
95
98
46.46
98
69.87
75.96
65
83.99
Daxue Mtn C
Xin Long
Bt
Bk
Bp
Bl
By
DC
Gan Zi
Sll Mtn CSll Mtn S
Daxue Mtn SEmei MtnQl Mtn SE 1
Ql Mtn SE 2
Ql Mtn NWEmei MtnHan Yuan
Wawu Mtn
Daxue Mtn S
Bailing Mtn
Luoji Mtn
Sll Mtn SESll Mtn S
Ylx Mtn
Sll Mtn C
0.09
0.010.020.030.040.050.0
a
b
c
d
e
f
g
h
i
j
k
l
mno
pq
r
s
tu
v
w
xy
z
Bk
Bp
Bl
By
DC
RsPmPasHqPssLsBtDaxue Mtn C
Xin Long
Gan ZiSll Mtn C
Sll Mtn SDaxue Mtn SEmei Mtn
Ql Mtn SE 1
Ql Mtn SE 2Ql Mtn NW
Emei MtnHan Yuan
Wawu MtnDaxue Mtn S
Luoji MtnBailing Mtn
Sll Mtn SE
Sll Mtn S
Ylx Mtn
Sll Mtn C
Fig. 2 Topology and divergence dates for salamanders of the
genus Batrachuperus. (a) RAXML phylogram. Names of the non-
Batrachuperus out-group species are removed for clarity. The
branching order is the same as that in the BEAST chronogram.
Maximum-likelihood bootstrap proportions (MLBs) support is
shown above the node, with Bayesian posterior probabilities
(BPPs) shown below. Unmarked nodes have MLBs = 100%
and BPPs = 1.0. (b) Simplified BEAST chronogram produced by
calibration method under relaxed clock model. Blue bars repre-
sent 95% highest probability densities (HPD) interval. Num-
bers on scale bar are millions of years before present. The
mean age estimates and actual 95% HPD are shown in
Table 3. Species name abbreviations: Rs, Ranodon sibiricus; Pm,
Paradactylodon mustersi; Pas, Pachyhynobius shangchengensis; Hq,
Hynobius quelpaertensis; Pss, Pseudohynobius shuichengensis; Ls,
Liua shihi; Bt, B. tibetanus; Bk, B. karlschmidti; DC, Daocheng
species; Bp, B. pinchonii; Bl, B. londongensis; By, B. yenyuanen-
sis. Region name abbreviations: Sll, Shaluli; Ql, Qionglai; Ylx,
Yulongxue.
PHYL OGEOGRAPHY OF TIBETAN SALAMANDERS 3315
� 2012 Blackwell Publishing Ltd
Hypotheses testing using coalescent simulations
Results from coalescent simulations with the corre-
sponding Ne failed to reject the colonization models for
B. karlschmidti (P = 0.188) and Daocheng species
(P = 0.190), the fragmented ancestor models for B. lon-
dongensis (P = 0.136) and B. yenyuanensis (P = 0.155),
and the staged fragmentation model for B. pinchonii
(P = 0.109). Thus, the diversification of these closely
related species seemed to be driven by three different
causes (Fig. 3).
The coalescent framework was also used to evaluate
whether or not the level of divergence within each
species of Batrachuperus corresponded with the most
recent orogenesis of the Tibetan Plateau and the corre-
sponding glacial-induced climatic shift hypothesis.
Thus, we adjusted the first divergence dates (tree
depth) to 2.6 and 3.6 Ma. These two time-points were
also close to or fell within the 95% CI obtained from
the BEAST dating estimates for most species (Table 3).
Coalescent simulations for corresponding diversifica-
tion models failed to reject the null hypotheses at
2.6 Ma for B. karlschmidti (P = 0.117), B. londongensis
(P = 0.159) and B. yenyuanensis (P = 0.101); diversifica-
tion appeared to be driven by the most recent orogen-
esis of the Tibetan Plateau and its concomitant glacial-
induced climatic shift. This hypothesis was rejected
when tested for B. pinchonii and Daocheng species
(P < 0.01), which had estimated divergence dates of
c. 5.0 Ma (95% CI, 3.7–6.3 Ma) and c. 7.2 Ma (95% CI,
5.4–9.1 Ma), respectively. Consequently, we assumed
the divergences occurred in the last 3.6 Ma, the begin-
ning of the most recent uplift of the Tibetan Plateau.
Coalescent simulations were then used to test these
null hypotheses. The DC values from reconstructed
Table 3 Age estimates using different methods. Letters for nodes correspond to Fig. 2b. The age unit is millions years ago
Node
Calibration method (95% HPD)
Mean rate method (95% HPD)Relaxed clock Strict clock Species-level sampling
a 40.8 (32.4–49.4) 37.9 (29.1–46.5) 38.7 (31.6–45.8) 44.2 (25.5–69.3)
b* 26.4 (17.9–35.9) 25.5 (20.5–31.2) 27.4 (20.9–34.0) 29.1 (13.9–49.4)
c 38.2 (30.6–46.5) 37.1 (28.6–45.9) 37.5 (30.7–44.2) 41.1 (23.3–64.3)
d* 34.7 (27.6–42.0) 35.2 (27.8–44.0) 35.3 (29.1–42.1) 37.3 (20.7–58.9)
e 28.7 (22.4–35.5) 30.8 (24.6–39.5) 30.5 (24.7–36.6) 30.9 (16.9–49.2)
f 23.4 (18.0–29.0) 26.0 (20.4–33.0) 25.7 (20.4–31.1) 25.1 (13.3–39.5)
g* 14.8 (11.7–18.0) 14.2 (11.3–17.9) 14.9 (11.9–18.0) 15.8 (9.0–24.5)
h 12.4 (9.5–15.4) 11.6 (8.9–15) 11.8 (9.1–14.9) 13.2 (7.5–21)
i 11.7 (8.9–14.4) 11.9 (9.3–15.2) 12.2 (9.5–14.9) 12.5 (7.2–19.7)
j 7.8 (5.6–10.1) 8.1 (6.5–10.4) 8.5 (6.3–10.9) 8.5 (4.5–13.3)
k 10.3 (7.7–13.0) 10.6 (8.0–13.2) 10.8 (8.4–13.7) 11.0 (6.1–17.3)
l 3.9 (2.8–5.1) 3.6 (2.7–4.8) — 4.2 (2.4–6.7)
m 2.2 (1.5–2.8) 1.9 (1.4–2.5) — 2.3 (1.3–3.7)
n 1.7 (1.2–2.2) 1.4 (1.1–1.9) — 1.8 (1.0–2.8)
o 1.3 (0.9–1.7) 1.2 (0.8–1.5) — 1.4 (0.8–2.2)
p 1.5 (1.0–2.1) 1.4 (0.9–1.9) — 1.6 (0.8–2.7)
q 3.8 (2.8–5.1) 3.6 (2.7–4.7) — 4.2 (2.3–6.8)
r 5.0 (3.7–6.3) 4.8 (3.6–6.2) — 5.5 (3.0–8.6)
s 4.5 (3.3–5.8) 4.3 (3.1–5.5) — 4.9 (2.7–7.8)
t 2.0 (1.3–2.8) 2.1 (1.6–2.6) — 2.1 (1.1–3.5)
u 1.0 (0.6–1.4) 1.0 (0.6–1.3) — 1.0 (0.4–1.7)
v 2.2 (1.4–2.9) 1.9 (1.4–2.5) — 2.4 (1.3–3.9)
w 3.0 (2.1–4.0) 2.6 (1.8–3.6) — 3.3 (1.8–5.3)
x 7.2 (5.4–9.1) 6.5 (5.0–8.2) — 7.6 (4.2–12.2)
y 5.7 (4.2–7.4) 5.4 (4.4–7.0) — 6.1 (3.3–9.8)
z 2.1 (1.4–2.8) 1.7 (1.3–2.1) — 2.2 (1.2–3.4)
HPD, highest probability densities.
*Calibration points.
The first divergence occurred within each species of Batrachuperus.
3316 B . LU ET AL.
gene trees for B. pinchonii (P = 0.352) and Daocheng
species (P = 0.236) were not significantly lower than
those values from the gene trees simulated by neutral
coalescence. Again, the null hypotheses of two main
divergence events could not be rejected.
Ancestral area reconstructions assigned the highest
probability for the origin of B. karlschmidti to the central
region of Daxue Mountains (proportional likelihoods,
PL = 0.61; Table 4), for Daocheng species, the south-
eastern region of Shaluli Mountains (PL = 0.60), and for
B. pinchonii, the southeastern region of Qionglai Moun-
tains (PL = 0.72). The probabilities of the three species’
origins for alternative regions were considerably lower
(highest PL = 0.23). Thus, a single ancestral area of ori-
gin was identified. These results were consistent with
the colonization models of B. karlschmidti and Daocheng
species and the staged fragmentation model of
B. pinchonii. In contrast, probabilities for all regions of
B. londongensis and B. yenyuanensis were low (highest
PL = 0.43). It seemed highly likely that multiple ances-
tral areas existed at the deeper nodes on the tree. This
discovery supported results from coalescent simulations
that diversification occurred via the fragmentation of a
wide-ranging common ancestor.
Historical demography
The mismatch distributions revealed distinct demo-
graphic histories for each species. Distributions of the
pairwise mutation differences were unimodal for B. kar-
lschmidti and B. yenyuanensis (Fig. S1, Supporting infor-
mation), and the sum of square deviations and
raggedness index suggested that the curves did not sig-
nificantly differ from the distributions expected under a
model of sudden demographic expansion (Table 1). In
contrast, the mismatch analyses for the remaining three
species of Batrachuperus showed a bimodal profile,
which might have resulted from either a constant or
declining demography. Further, the BSP analysis pro-
vided additional and sometimes different details. The
population size of the two high-elevation species
(B. karlschmidti and Daocheng species) appeared to
� 2012 Blackwell Publishing Ltd
Table 4 Proportional likelihoods for the area of origin for the oldest ancestral node in the five species of Batrachuperus
Species The area of origin
B. karlschmidti Daxue Mtn C Gan Zi Xin Long Shaluli Mtn C Shaluli Mtn S
0.6118 0.0896 0.2341 0.0323 0.0323
B. londongensis Emei Mtn Wawu Mtn Han Yuan — —
0.3941 0.2812 0.3248 — —
B. pinchonii Qionglai Mtn NW Qionglai Mtn SE Emei Mtn Daxue Mtn S —
0.0838 0.7244 0.0904 0.1013 —
Daocheng species Shaluli Mtn SE Shaluli Mtn S Yulongxue Mtn Shaluli Mtn C —
0.5976 0.2008 0.1985 0.0031 —
B. yenyuanensis Daxue Mtn S Bailing Mtn Luoji Mtn — —
0.2554 0.3100 0.4346 — —
0.1792 0.1208 0.2583 0.3208 0.1208
0.4872 0.1282 0.3846
0.3400 0.5867 0.0733
0.1757 0.2008 0.5690 0.0544
0.5193 0.0883 0.0221 0.3702Daxue Mtn C Xin Long Gan Zi Shaluli Mtn C Shaluli Mtn S
Qionglai Mtn SE2 Qionglai Mtn NW Qionglai Mtn SE1 Emei&Daxue
Time in generations
(a) B. karlschmidti
Colonization
(c) B. pinchonii
(d) B. londongensis
(e) B. yenyuanensis
Staged Fragmentation
Fragmentation
Colonization
(b) Daocheng species
Han Yuan Wawu Mtn Emei Mtn
Luoji Mtn Bailing Mtn Daxue Mtn S
Fragmentation
Shaluli Mtn SE Shaluli Mtn S Yulongxue Mtn Shaluli Mtn C
Time in generations
Time in generations
Time in generations
Time in generations
Fig. 3 Illustration of the three biogeographic hypotheses of diversification within each species of Batrachuperus tested using coales-
cent simulations. Branch lengths are time in generations based on a 3.5-year generation time. Branch widths (Ne) are scaled based on
the proportion of the Total Ne that each mountain comprised (listed below mountain name). Internal branches on models were scaled
such that all branch widths summed to Total Ne at any single point in time. (c) Emei&Daxue, Emei Mtn and Daxue Mtn S.
PHYL OGEOGRAPHY OF TIBETAN SALAMANDERS 3317
remain relatively stable during the extensive glaciation
period (0.5–0.175 Ma), rapidly declined in the Late
Pleistocene, but increased sharply around the end of
the last glacial maximum (LGM) in the eastern edge of
the Tibetan Plateau (0.016–0.032 Ma; Fig. 4). The sud-
den expansion model for Daocheng species contradicted
the results from the mismatch distributions. The popu-
lation size of two low-elevation species (B. pinchonii
� 2012 Blackwell Publishing Ltd
and B. londongensis) appeared to gradually decrease at
the beginning of the extensive glacial period (0.5–
0.175 Ma). Substantial population growth of B. yenyuan-
ensis started approximately 0.15 Ma, which coincided
with the ending of the extensive glaciation period. The
historical dynamics of the five species of Batrachuperus
were also supported by the BSP analyses for each regio-
nal population within each species.
Table 5 Nineteen biological environmental variables used in
habitat modelling and the principal component analysis on these
variables used in a comparison of climatic conditions between
occurrence locations for the five species of Batrachuperus
Bio-variable
Component
1 2
BIO1 = Annual Mean Temperature 0.786 0.612
BIO2 = Mean Diurnal Range )0.902 0.247
BIO3 = Isothermality )0.666 0.229
BIO4 = Temperature Seasonality )0.300 0.086
BIO5 = Max Temperature of Warmest Month 0.654 0.720
BIO6 = Min Temperature of Coldest Month 0.896 0.421
BIO7 = Temperature Annual Range )0.809 0.174
BIO8 = Mean Temperature of Wettest Quarter 0.768 0.629
BIO9 = Mean Temperature of Driest Quarter 0.771 0.599
BIO10 = Mean Temperature of Warmest Quarter 0.769 0.625
BIO11 = Mean Temperature of Coldest Quarter 0.796 0.562
BIO12 = Annual Precipitation 0.829 )0.514
BIO13 = Precipitation of Wettest Month 0.737 )0.520
BIO14 = Precipitation of Driest Month 0.850 )0.402
BIO15 = Precipitation Seasonality )0.502 0.051
BIO16 = Precipitation of Wettest Quarter 0.756 )0.548
BIO17 = Precipitation of Driest Quarter 0.849 )0.391
BIO18 = Precipitation of Warmest Quarter 0.742 )0.563
(a)
(c)
(d)
(e)
(b)
Fig. 4 Bayesian skyline plots showing
the demographic history of five species
of Batrachuperus. (a) B. karlschmidti, (b)
Daocheng species, (c) B. pinchonii, (d)
B. londongensis and (e) B. yenyuanensis.
Dark lines represent median values for
the log10 of the population size (Ne*s) (s,
generation time), blue lines mark the
95% highest probability density (HPD)
intervals in all panels.
3318 B . LU ET AL.
Ecological niche modelling
Ecological niche modelling projected that the distribu-
tion of each species of Batrachuperus was now restricted
to a specific montane habitat. The region of highly suit-
able habitat was limited to and consistent with the dis-
tributions of each species (Fig. S2, Supporting
information). For each distributional model, the AUC
value approached one (‡0.997) indicating a better than
random prediction (0.5 = random). Two principal com-
ponents among the 19 possibly correlated environmen-
tal variables explained 80.64% of the total variation
(Table 5). The first principal component represented
precipitation and temperature (the lower the value, the
colder and drier), whereas the second principal compo-
nent represented climatic variability (the lower the
value, the less variable). Habitats of high-elevation spe-
cies appeared to be cooler, drier and more variable than
those of low-elevation species (Fig. 5). The MANOVA
analysis indicated that the habitats of the five species of
Batrachuperus differed significantly (Wilk’s k = 0.281,
F = 15.07, P < 0.001). This result provided ecological
and physiographic explanations for the distinct demo-
graphic and divergence histories of each species.
BIO19 = Precipitation of Coldest Quarter 0.847 )0.378
Initial eigenvalues 11.028 4.293
% of variance 58.042 22.595
DiscussionThe various drivers of species divergence associated with
orogenesis seem to play roles in the evolution of Batra-
chuperus. Closely related species of Batrachuperus seem to
have experienced different patterns of diversification dri-
ven by topography. They appear to have been affected
by different glacial periods as a result of variation in ele-
vation distributions and cold-tolerant physiology.
� 2012 Blackwell Publishing Ltd
Fig. 5 Results from the comparison of environmental (means ± 95% CIs) conditions at sampling sites for Batrachuperus. Habitats
where high-elevation species (Bk, B. karlschmidti; DC, Daocheng species) occur are significantly cooler, drier and more variable than
those occupied by low-elevation species (Bp, B. pinchonii; Bl, B. londongensis; By, B. yenyuanensis) occupied.
PHYL OGEOGRAPHY OF TIBETAN SALAMANDERS 3319
Distinct divergence patterns
Topographically diverse montane regions usually have
a suite of species with differing and complex evolution-
ary histories and divergence patterns (Despres et al.
2002; Kropf et al. 2003). Closely related species of sala-
manders endemic to the eastern edge of the Tibetan
Plateau also have distinct patterns, and these are associ-
ated with high- and low-elevation topographies.
Coalescent simulations find that diversification in
western high-elevation species is more consistent with
a colonization model involving a series of dispersals
from one mountain to another followed closely by iso-
lation and divergence. Fragmentation of the range of a
widely distributed common ancestor is not suggested.
In contrast, diversification in eastern low-elevation spe-
cies appears to have been driven by either fragmenta-
tion or staged fragmentation of a once wide-ranging
species. Species of Batrachuperus generally occupy moist
habitats in forested areas. Consequently, their distribu-
tions can be strongly influenced by the occurrence of
mesic forests. Orogenesis of the Tibetan Plateau and
Pleistocene glacial cycling are responsible for climatic
and ecological shifts. The climate appears to have
become gradually drier and colder; the development of
glaciers in this region is associated with the contraction
of broadleaf deciduous forests and a dominance of
grasslands (Wu et al. 2001). Most temperate species
appear to have dispersed to lower latitudes in response
to Pleistocene climatic oscillations (Hewitt 2000; Rowe
et al. 2004).
Under a coalescent framework, the null hypotheses
states that the initial divergence within each species
occurred at c. 2.6–3.6 Ma, and it cannot be rejected. This
dating corresponds to the most intense uplift of the
Tibetan Plateau (Zheng et al. 2000; An et al. 2001) and
the beginning of the Pleistocene. Consequently, climatic
changes triggered by the two events may have driven
the divergence, although more evidence is needed.
� 2012 Blackwell Publishing Ltd
The region’s northsouth tending mountain ranges
facilitate lower-latitude dispersal by providing favour-
able climate and habitats. The levels of population dif-
ferentiation reflect historical isolation events. Multiple
montane occurrences and drift, and not ongoing gene
flow, are suggested to be the driving force. The precise
colonization routes are likely to have been obscured by
the accumulation of genetic differences and coalescence
times. Eastern low-elevational species are highly likely
to have been restricted to mesic sky island forests iso-
lated by deep valleys with unfavourable environments.
Unlike contiguous mountains in the western area, val-
leys appear to serve as barriers to dispersal because lin-
eages are separated by unsuitable conditions (Wiens
2004; Wiens & Graham 2005; Wiens et al. 2006).
Ancestral area reconstructions provide additional
support for the distinction between divergence patterns
in high- and low-elevation species. High-elevation spe-
cies are associated with a single area of origin. Their
ancestors are from a relatively small area. Batrachuperus
karlschmidti seems to be derived from the central region
of Daxue Mountains (1–4), and subsequent colonization
appears to have occurred mainly along the Shaluli
Mountain range eventually reaching its southern region
(18, 19). Daocheng species appears to have its origin in
the southeastern Shaluli Mountains (19); a series of
short-distance colonizations restrict it to the central and
southern Shaluli Mountains. Interestingly, the common
ancestor of Daocheng species appears to have been
present in the southeastern Shaluli Mountains before
B. karlschmidti colonized that area. Thus, B. karlschmidti
seems to have invaded suitable habitats previously
dominated by Daocheng species. Niche modelling pre-
dicts that the two species occupy similar habitats, which
may facilitate competition, as is documented to occur in
other closely related species of salamanders (Hairston
1951; Crespi et al. 2003).
Our analyses cannot identify single ancestral areas of
origin for most low-elevation species that exhibit
3320 B . LU ET AL.
fragmentation patterns, especially B. londongensis and
B. yenyuanensis. This finding suggests the rapid frag-
mentation of wide-ranging ancestral populations into
several parts. The ancestral area reconstruction for
B. pinchonii assigns the highest probability to southeast-
ern Qionglai Mountains (28–32). The staged fragmenta-
tion pattern of this low-elevation species infers that its
common ancestor initially fragmented into two ances-
tral populations (northern and southern), and subse-
quently both of them fragmented. The first stage may
be associated with orogenesis of the Qionglai Moun-
tains, although more powerful evidence is needed to
support this scenario. Gongga Mountain, the highest
(7556 m) in the region, is surrounded by more than 20
other peaks that exceed 6000 m in the central Daxue
Mountains. The palaeomagnetic age of Gongga Moun-
tain is c. 2.48 Ma (Li et al. 1991), and the first coloniza-
tion event (c. 2.6 Ma) in B. karlschmidti is closely
associated with this origin. Batrachuperus karlschmidti
does not occur south of Gongga Mountain. Orogenesis
in the region may have restricted ancestral populations
to the north. The Late Mesozoic Yanshan movement is
responsible for the initial creation of Mount Emei, and
the most recent Tibetan Plateau movement (c. 3 Ma)
uplifted the mountain to its present height. Concor-
dance between dates for the fragmentation of B. lon-
dongensis and the uplifting of Mount Emei suggests that
the creation of deep valleys produced barriers to disper-
sion, thus reducing gene flow and promoting allopatric
divergence.
Contrasting demographic histories
Species at high and low elevations appear to have been
affected by different glacial periods. Responses to Pleis-
tocene glacial cycles are expected to vary among species
and geographical regions, in part because of differential
cold tolerance (Hewitt 1996, 2004).
Bayesian skyline plots show that high-elevation spe-
cies were suppressed by the last glaciation (0.016–
0.032 Ma) but not by the earlier, more extensive glacia-
tions (0.5–0.175 Ma; Shi 2002; Shi et al. 1995; Zhang
et al. 2000; Zheng et al. 2002), although the mismatch
distribution rejects the model of sudden demographic
expansion for Daocheng species. Methods estimating
population parameters based on the distribution of
pairwise differences do not make full use of DNA
sequence data, while methods based on coalescent the-
ory, by incorporating information from the genealogy,
may obtain better estimates of the demographic histo-
ries of populations (Felsenstein 1992; Pybus et al. 2000).
Only a few mountain heights extend above the Middle
Pleistocene snowline, and glaciation appears to have
been limited to areas such as Gongga Mountain (Shi
et al. 1990; Li et al. 1991). Glaciation is controlled by cli-
mate and geomorphology. Although the climate is doc-
umented to have been colder during the extensive
glacial period than the LGM (Shi et al. 1995), mountain
heights below the snowline could not be glaciated. Dur-
ing the LGM, if mountain heights extended above the
snowline, glaciers could form and became cold sources
acting as positive-feedback mechanisms through reflect-
ing more of the sun’s energy and absorbing less; oro-
genesis raised mountain heights and glaciers
contributed to cooler climates (Zheng et al. 2002). Cur-
rent snowlines in the eastern edge of the Tibetan Pla-
teau extend from 4200 to 5200 m (Shi et al. 1997; Liu
et al. 1999; Li et al. 2009). During the LGM, the snow-
line descended to approximately 3300 m in many
mountain ranges (Shi et al. 1997; Liu et al. 1999). The
two high-elevation species currently occur above
3500 m. Their suitable habitat seems to have been heav-
ily covered with ice during the LGM. If true, then sala-
manders must have dispersed to ice-free refugia, such
as lower elevations or adjacent areas, only to recolonize
the region via a population expansion during postgla-
cial times.
The population sizes of the three low-elevation spe-
cies seem to have been affected from the beginning of
the extensive glacial period. The low-elevation species
dwell below the snow line (1800–3200 m). Suitable hab-
itat might have not been covered by ice during both
the LGM and the extensive glacial period. Ecological
niche modelling suggests that the low-elevation species
are less cold-tolerant than the high-elevation species. If
true, then low-elevation species are more likely to suf-
fer from climatic cooling than high-elevation species.
From the beginning of the extensive glaciation period,
low-elevation species are predicted to have lowered
population sizes even in ice-free regions. In contrast,
high-elevation, cold-tolerant species are expected to be
affected only if their suitable habitat is heavily covered
with ice during the LGM. These predictions are consis-
tent with the hypothesis that variation in ecological
adaptations affects geographical patterns of genetic var-
iation (Gavrilets 2003; Hewitt 2004). Glacial advances
decrease temperatures while increasing aridity. These
changes may have contributed to declines in popula-
tion size in the less cold-tolerant species and resulted
in the use of refugia by many species during the LGM
(Hewitt 1996, 2000, 2004; Rowe et al. 2004). However,
B. yenyuanensis, which occupies a low-elevation and
low-latitude region, seems to have experienced a sub-
stantial amount of population growth after the exten-
sive glaciation period yet did not suffer losses during
the LGM. The climate during the extensive glaciation
period might have been much colder and drier than
that during the LGM (Shi et al. 1995; Zheng et al.
� 2012 Blackwell Publishing Ltd
PHYL OGEOGRAPHY OF TIBETAN SALAMANDERS 3321
2002). This may be an important reason why glaciation
affects low-latitude species. In contrast, a cold climate
without ice cover is likely to have little negative effect
on more cold-tolerant species, such as B. karlschmidti
and Daocheng species. It remains unclear why the pop-
ulations of B. pinchonii and B. londongensis did not
expand after the glaciers retreated. Other factors appear
to be involved.
Our study on Tibetan Batrachuperus underscores the
various roles played by elevation, geological history
and animal physiology in shaping the patterns of
genetic diversity and demography among closely
related species. Further studies using the rigorous test-
ing of hypotheses are likely to yield an understanding
the drivers of evolutionary change.
Acknowledgements
We thank Jinzhong Fu for constructive comments on earlier
versions of this manuscript. We also thank Xianguang Guo, Li
Ding and Jiatang Li for valuable discussion during the revi-
sion. This project was supported by a National Natural Science
Foundation of China (NSFC grant no. 30870287) and a Chinese
Academy of Sciences (KSCX2-EW-J-22) to X.M. Zeng, and a
National Natural Sciences Foundation of China (NSFC-
30900134) and a Chinese Academy of Sciences (09C3011100,
KSCX2-YW-Z-0906) to Y.C. Zheng, and by a Visiting Professor-
ship for Senior International Scientists from the Chinese Acad-
emy of Sciences and a Natural Sciences and Engineering
Research Council Discovery Grant (A3148) to R.W. Murphy.
References
Aljanabi SM, Martinez I (1997) Universal and rapid salt-
extraction of high quality genomic DNA for PCR-based
techniques. Nucleic Acids Research, 25, 4692–4693.
An Z, Kutzbach JE, Prell WL, Porter SC (2001) Evolution of
Asian monsoons and phased uplift of the Himalaya-Tibetan
plateau since Late Miocene times. Nature, 411, 62–66.
Averianov A, Tjutkova L (1995) Ranodon cf. sibiriens
(Amphibia, Caudata) from the Upper Pliocene of southern
Kazakhstan: the first fossil record of the family Hynobiidae.
Palaontologische Zeitschrift, 69, 257–264.
Beerli P, Felsenstein J (1999) Maximum-likelihood estimation of
migration rates and effective population numbers in two
populations using a coalescent approach. Genetics, 152, 763–
773.
Beerli P, Felsenstein J (2001) Maximum likelihood estimation of
a migration matrix and effective population sizes in n
subpopulations by using a coalescent approach. Proceedings
of the National Academy of Sciences, USA, 98, 4563–4568.
Bryson RW, Murphy RW, Graham MR, Lathrop A, Lazcano-
Villareal D (2011a) Ephemeral Pleistocene woodlands
connect the dots for highland rattlesnakes of the Crotalus
intermedius group. Journal of Biogeography, 38, 2299–2310.
Bryson RW, Murphy RW, Lathrop A, Lazcano-Villareal D
(2011b) Evolutionary drivers of phylogeographical diversity
in the highlands of Mexico: a case study of the Crotalus
� 2012 Blackwell Publishing Ltd
triseriatus species group of montane rattlesnakes. Journal of
Biogeography, 38, 697–710.
Carstens BC, Richards CL (2007) Integrating coalescent and
ecological niche modeling in comparative phylogeography.
Evolution, 61, 1439–1454.
Che J, Zhou W-W, Hu J-S et al. (2010) Spiny frogs (Paini)
illuminate the history of the Himalayan region and
Southeast Asia. Proceedings of the National Academy of Sciences,
USA, 107, 13765–13770.
Craw D, Burridge CP, Upton P, Rowe DL, Waters JM (2008)
Evolution of biological dispersal corridors through a
tectonically active mountain range in New Zealand. Journal
of Biogeography, 35, 1790–1802.
Crespi EJ, Rissler LJ, Browne RA (2003) Testing Pleistocene
refugia theory: phylogeographical analysis of Desmognathus
wrighti, a high-elevation salamander in the southern
Appalachians. Molecular Ecology, 12, 969–984.
Despres L, Loriot S, Gaudeul M (2002) Geographic pattern of
genetic variation in the European globeflower Trollius
europaeus L. (Ranunculaceae) inferred from amplified
fragment length polymorphism markers. Molecular Ecology,
11, 2337–2347.
Diego SM (2010) A multilocus timescale for the origin of extant
amphibians. Molecular Phylogenetics and Evolution, 56, 554–561.
Drummond AJ, Rambaut A (2007a) BEAST: Bayesian
evolutionary analysis by sampling trees. BMC Evolutionary
Biology, 7, 214.
Drummond AJ, Rambaut A (2007b) Tracer v1.4, Available from
http://beast.bio.ed.ac.uk/Tracer.
Drummond AJ, Nicholls GK, Rodrigo AG, Solomon W (2002)
Estimating mutation parameters, population history and
genealogy simultaneously from temporally spaced sequence
data. Genetics, 161, 1307–1320.
Drummond AJ, Rambaut A, Shapiro B, Pybus OG (2005)
Bayesian coalescent inference of past population dynamics
from molecular sequences. Molecular Biology and Evolution,
22, 1185–1192.
Elith J, Phillips SJ, Hastie T et al. (2011) A statistical
explanation of MaxEnt for ecologists. Diversity and
Distributions, 17, 43–57.
Erixon P, Svennblad B, Britton T, Oxelman B (2003) Reliability
of Bayesian posterior probabilities and bootstrap frequencies
in phylogenetics. Systematic Biology, 52, 665–673.
Excoffier L, Lischer HEL (2010) Arlequin suite ver 3.5: a new
series of programs to perform population genetics analyses
under Linux and Windows. Molecular Ecology Resources, 10,
564–567.
Felsenstein J (1985) Confidence limits on phylogenies: an
approach using the bootstrap. Evolution, 39, 783–791.
Felsenstein J (1992) Estimating effective population size from
samples of sequences: inefficiency of pairwise and
segregating sites as compared to phylogenetic estimates.
Genetical Research, 59, 139–147.
Fu J, Zeng X (2008) How many species are in the genus
Batrachuperus? A phylogeographical analysis of the stream
salamanders (Family Hynobiidae) from southwestern China
Molecular Ecology, 17, 1469–1488.
Gao K-Q, Shubin NH (2003) Earliest known crown-group
salamanders. Nature, 422, 424–428.
Gavrilets S (2003) Perspective: models of speciation: what have
we learned in 40 years? Evolution, 57, 2197–2215.
3322 B . LU ET AL.
Haffer J (2008) Hypotheses to explain the origin of species in
Amazonia. Brazilian Journal of Biology, 68, 917–947.
Hairston NG (1951) Interspecies competition and its probable
influence upon the vertical distribution of Appalachian
salamanders of the genus Pethodon. Ecology, 32, 266–274.
Han Y, Lu X (1999) Age estimation of Batrachuperus tibetanus.
Journal of Changshu College, 13, 73–77 (In Chinese with
English abstract).
Hewitt GM (1996) Some genetic consequences of ice ages, and
their role, in divergence and speciation. Biological Journal of
the Linnean Society, 58, 247–276.
Hewitt GM (2000) The genetic legacy of the Quaternary ice
ages. Nature, 405, 907–913.
Hewitt GM (2004) Genetic consequences of climatic
oscillations in the Quaternary. Philosophical Transactions of
the Royal Society of London. Series B: Biological Sciences, 359,
183–195.
Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005)
Very high resolution interpolated climate surfaces for
global land areas. International Journal of Climatology, 25,
1965–1978.
Hillis DM, Bull JJ (1993) An empirical test of bootstrapping as
a method for assessing confidence in phylogenetic analysis.
Systematic Biology, 42, 182–192.
Ho SYW, Phillips MJ, Cooper A, Drummond AJ (2005) Time
dependency of molecular rate estimates and systematic
overestimation of recent divergence times. Molecular Biology
and Evolution, 22, 1561–1568.
Huelsenbeck JP, Rannala B (2004) Frequentist properties of
Bayesian posterior probabilities of phylogenetic trees under
simple and complex substitution models. Systematic Biology,
53, 904–913.
Huelsenbeck JP, Ronquist F (2001) MrBayes: Bayesian inference
of phylogeny. Bioinformatics, 17, 754–755.
Hugall AF, Foster R, Lee MSY (2007) Calibration choice, rate
smoothing, and the pattern of tetrapod diversification
according to the long nuclear gene RAG-1. Systematic Biology,
56, 543–563.
Kass RE, Raftery AE (1993) Bayes factors and model
uncertainty. Journal of the American Statistical Association, 90,
773–795.
Kelchner SA, Thomas MA (2007) Model use in phylogenetics:
nine key questions. Trends in Ecology and Evolution, 22, 87–94.
Knowles LL (2000) Tests of Pleistocene speciation in montane
grasshoppers (genus melanoplus) from the sky islands of
western North America. Evolution, 54, 1337–1348.
Knowles LL (2004) The burgeoning field of statistical
phylogeography. Journal of Evolutionary Biology, 17, 1–10.
Knowles LL (2009) Statistical phylogeography. Annual Review
of Ecology, Evolution, and Systematics, 40, 593–612.
Knowles L, Maddison W (2002) Statistical phylogeography.
Molecular Ecology, 11, 2623–2635.
Kropf M, Kadereit JW, Comes HP (2003) Differential cycles of
range contraction and expansion in European high
mountain plants during the Late Quaternary: insights from
Pritzelago alpina (L.) O. Kuntze (Brassicaceae). Molecular Ecology,
12, 931–949.
Li J, Fang X (1999) Uplift of the Tibetan Plateau and
environmental changes. Chinese Science Bulletin, 44, 2117–2124.
Li Z, Chen Z, Wan M (1991) Classification and correlation of
the Quaternary glacial epoch in the Hengduan (transverse)
mountains. Geological Review, 37, 126–132 (In Chinese with
English abstract).
Li J, Fang X, Ma H (1996) Geomorphological and
environmental evolution in the upper reaches of Huanghe
River during the Late Cenozoic. Science in China Series D
(English version), 39, 380.
Li Z, He Y, Wang S et al. (2009) Changes of some monsoonal
temperate glaciers in Hengduan Mountains region during
1900–2007. Acta Ggeographica Sinica, 64, 1319–1330 (In
Chinese with English abstract).
Librado P, Rozas J (2009) DnaSP v5: a software for
comprehensive analysis of DNA polymorphism data.
Bioinformatics, 25, 1451–1452.
Liu T, Zhang X, Xiong S, Qin X (1999) Qinghai-Xizang Plateau
glacial environment and global cooling. Quaternary Sciences,
5, 385–396 (In Chinese with English Abstract).
Luo J, Yang D, Suzuki H et al. (2004) Molecular phylogeny
and biogeography of Oriental voles: genus Eothenomys
(Muridae, Mammalia). Molecular Phylogenetics and Evolution,
33, 349–362.
Luo A, Qiao H, Zhang Y et al. (2010) Performance of criteria
for selecting evolutionary models in phylogenetics: a
comprehensive study based on simulated datasets. BMC
Evolutionary Biology, 10, 242.
Maddison WP (1997) Gene trees in species trees. Systematic
Biology, 46, 523–536.
Maddison WP, Maddison DR (2010) Mesquite: a modular
system for evolutionary analysis. Version 2.73 http://
mesquiteproject.org.
Marjanovic D, Laurin M (2007) Fossils, molecules, divergence
times, and the origin of lissamphibians. Systematic Biology,
56, 369–388.
Mueller RL (2006) Evolutionary rates, divergence dates, and
the performance of mitochondrial genes in Bayesian
phylogenetic analysis. Systematic Biology, 55, 289–300.
Mueller RL, Macey JR, Jaekel M et al. (2004) Morphological
homoplasy, life history evolution, and historical
biogeography of plethodontid salamanders inferred from
complete mitochondrial genomes. Proceedings of the National
Academy of Sciences, USA, 101, 13820–13825.
Peng Z, Ho SYW, Zhang Y, He S (2006) Uplift of the Tibetan
Plateau: evidence from divergence times of glyptosternoid
catfishes. Molecular Phylogenetics and Evolution, 39, 568–572.
Phillips SJ, Anderson RP, Schapire RE (2006) Maximum
entropy modeling of species geographic distributions.
Ecological Modelling, 190, 231–259.
Posada D (2008) jModelTest: phylogenetic model averaging.
Molecular Biology and Evolution, 25, 1253–1256.
Pybus OG, Rambaut A, Harvey PH (2000) An integrated
framework for the inference of viral population history from
reconstructed genealogies. Genetics, 155, 1429–1437.
Qu Y, Lei F, Zhang R, Lu X (2010) Comparative
phylogeography of five avian species: implications for
Pleistocene evolutionary history in the Qinghai-Tibetan
Plateau. Molecular Ecology, 19, 338–351.
Roelants K, Haas A, Bossuyt F (2011) Anuran radiations and
the evolution of tadpole morphospace. Proceedings of the
National Academy of Sciences, USA, 108, 8731–8736.
Rogers AR, Harpending H (1992) Population growth makes
waves in the distribution of pairwise genetic differences.
Molecular Biology and Evolution, 9, 552–569.
� 2012 Blackwell Publishing Ltd
PHYL OGEOGRAPHY OF TIBETAN SALAMANDERS 3323
Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian
phylogenetic inference under mixed models. Bioinformatics,
19, 1572–1574.
Rowe KC, Heske EJ, Brown PW, Paige KN (2004) Surviving the
ice: northern refugia and postglacial colonization. Proceedings
of the National Academy of Sciences, USA, 101, 10355–10359.
Shepard DB, Burbrink FT (2008) Lineage diversification and
historical demography of a sky island salamander, Plethodon
ouachitae, from the interior highlands. Molecular Ecology, 17,
5315–5335.
Shi Y (2002) Characteristics of Late Quaternary monsoonal
glaciation on the Tibetan Plateau and in East Asia.
Quaternary International, 97–98, 79–91.
Shi Y, Zheng B, Li S (1990) Last glaciation and maximum glaciation
in Qinghai-Xizang (Tibet) Plateau. Journal of Glaciology and
Geocryology, 12, 1–15 (In Chinese with English abstract).
Shi Y, Zheng B, Li S, Ye B (1995) Studies on altitude and
climatic environment in the middle and east part of Tibetan
Plateau during Quaternary maximum glaciation. Journal of
Glaciology and Geocryology, 17, 97–112 (In Chinese with
English abstract).
Shi Y, Zheng B, Yao T (1997) Glaciers and environments
during the last glacial maximum (LGM) on the Tibetan
Plateau. Journal of Glaciology and Geocryology, 19, 97–113 (In
Chinese with English abstract).
Shi Y, Li J, Li B et al. (1998) Uplift and environmental
evolution of Qinghai Xizang (Tibet) Plateau. In: Formation,
Evolution and Development of Qinghai-Xizang (eds Sun H,
Zheng D), pp. 75–137. Guangdong Science and Technology
Press, Guangdong. (In Chinese).
Slatkin M, Hudson RR (1991) Pairwise comparisons of
mitochondrial DNA sequences in stable and exponentially
growing populations. Genetics, 129, 555–562.
Smith CI, Farrell BD (2005) Phylogeography of the longhorn
cactus beetle Moneilema appressum LeConte (Coleoptera:
Cerambycidae): was the differentiation of the Madrean sky
islands driven by Pleistocene climate changes? Molecular
Ecology, 14, 3049–3065.
Stamatakis A (2006) RAxML-VI-HPC: maximum likelihood-
based phylogenetic analyses with thousands of taxa and
mixed models. Bioinformatics, 22, 2688–2690.
Swofford D (2002) PAUP*: Phylogenetic Analysis Using
Parsimony (*and Other Methods), Version 4.0b10. Sinauer
Associates, Sunderland, Massachusetts.
Weisrock DW, Macey JR, Ugurtas IH, Larson A, Papenfuss TJ
(2001) Molecular phylogenetics and historical biogeography
among salamandrids of the ‘‘true’’ salamander clade: rapid
branching of numerous highly divergent lineages in
Mertensiella luschani associated with the rise of Anatolia.
Molecular Phylogenetics and Evolution, 18, 434–448.
Wiens JJ (2004) Speciation and ecology revisited: phylogenetic
niche conservatism and the origin of species. Evolution, 58,
193–197.
Wiens JJ, Graham CH (2005) Niche conservatism: integrating
evolution, ecology, and conservation biology. Annual Review
of Ecology, Evolution, and Systematics, 36, 519–539.
Wiens JJ, Engstrom TN, Chippindale PT (2006) Rapid
diversification, incomplete isolation, and the ‘‘species clock’’
in North American salamanders (genus Plethodon): testing
the hybrid swarm hypothesis of rapid radiation. Evolution,
60, 2585–2603.
� 2012 Blackwell Publishing Ltd
Wu Y, Cui Z, Liu G et al. (2001) Quaternary geomorphological
evolution of the Kunlun Pass area and uplift of the Qinghai-
Xizang (Tibet) Plateau. Geomorphology, 36, 203–216.
Yoshikawa N, Matsui M, Nishikawa K, Kim J-B, Kryukov A
(2008) Phylogenetic relationships and biogeography of the
Japanese clawed salamander, Onychodactylus japonicus
(Amphibia: Caudata: Hynobiidae), and its congener inferred
from the mitochondrial cytochrome b gene. Molecular
Phylogenetics and Evolution, 49, 249–259.
Yu N, Zheng C, Zhang Y-P, Li W-H (2000) Molecular
systematics of pikas (genus Ochotona) inferred from
mitochondrial DNA sequences. Molecular Phylogenetics and
Evolution, 16, 85–95.
Zhang F, Jiang Z (2006) Mitochondrial phylogeography and
genetic diversity of Tibetan gazelle (Procapra picticaudata):
implications for conservation. Molecular Phylogenetics and
Evolution, 41, 313–321.
Zhang P, Wake DB (2009) Higher-level salamander
relationships and divergence dates inferred from complete
mitochondrial genomes. Molecular Phylogenetics and Evolution,
53, 492–508.
Zhang D, Fengquan L, Jianmin B (2000) Eco-environmental
effects of the Qinghai-Tibet Plateau uplift during the
Quaternary in China. Environmental Geology, 39, 1352–1358.
Zhang P, Chen Y-Q, Zhou H et al. (2006) Phylogeny, evolution,
and biogeography of Asiatic salamanders (Hynobiidae).
Proceedings of the National Academy of Sciences, USA, 103,
7360–7365.
Zhang D-R, Chen M-Y, Murphy RW et al. (2010) Genealogy and
palaeodrainage basins in Yunnan Province: phylogeography
of the Yunnan spiny frog, Nanorana yunnanensis
(Dicroglossidae). Molecular Ecology, 19, 3406–3420.
Zheng H, Powell CM, An Z, Zhou J, Dong G (2000)
Pliocene uplift of the northern Tibetan Plateau. Geology, 28,
715–718.
Zheng B, Xu Q, Shen Y (2002) The relationship between
climate change and Quaternary glacial cycles on the
Qinghai-Tibetan Plateau: review and speculation. Quaternary
International, 9798, 93–101.
Zheng Y, Peng R, Kuro-o M, Zeng X (2011) Exploring patterns
and extent of bias in estimating divergence time from
mitochondrial DNA sequence data in a particular lineage: a
case study of salamanders (order Caudata). Molecular Biology
and Evolution, 28, 2521–2535.
Zhou S, Wang X, Wang J, Xu L (2006) A preliminary study on
timing of the oldest Pleistocene glaciation in Qinghai-Tibetan
Plateau. Quaternary International, 154–155, 44–51.
B.L is interested in understanding the genetic and evolutionary
mechanisms that generate and maintain patterning variation
within and between species. Y.C.Z‘s research focuses on the
systematics, correlated evolution, and biogeography of amphi-
bians. R.W.M. is broadly interested in genetics, genomics, and
evolution. X.M.Z‘s general areas of interest are systematics and
evolutionary biology of amphibians and reptiles. Current
research activities focus on the mechanisms of speciation and
more specifically to understand the role of chromosomal rear-
rangements in speciation.
3324 B . LU ET AL.
Data accessibility
DNA sequences: GenBank accessions JQ303644–JQ304246.
Sampling information uploaded as online supplemental mate-
rial.
Phylogenetic data and DNA sequence alignment for all indi-
viduals: TreeBASE Study accession no. S12425.
Supporting information
Additional supporting information may be found in the online
version of this article.
Fig. S1 Mismatch distributions for five species of Batrachuperus.
(a) B. karlschmidti, (b) Daocheng species, (c) B. pinchonii, (d)
B. londongensis and (e) B. yenyuanensis.
Fig. S2 The ecological niche models of the current distribution
for five species of Batrachuperus. (a) B. karlschmidti, (b) Daoch-
eng species, (c) B. pinchonii, (d) B. londongensis and (e) B. yeny-
uanensis. Colours indicate predicted probability that conditions
are suitable, with values ranging from 0 to 1. High values indi-
cate high probability of suitable conditions for the species,
whereas low values indicate low predicted probability of suit-
able conditions.
Appendix S1 Specimens of Batrachuperus used in genealogical
and coalescent analyses. All vouchers are deposited in the
Chengdu Institute of Biology (CIB). Locality numbers corre-
spond to those on the map in Fig. 1.
Please note: Wiley-Blackwell is not responsible for the con-
tent or functionality of any supporting information supplied
by the authors. Any queries (other than missing material)
should be directed to the corresponding author for the
article.
� 2012 Blackwell Publishing Ltd