Content uploaded by Terrence Demos
Author content
All content in this area was uploaded by Terrence Demos on Oct 06, 2017
Content may be subject to copyright.
Local endemism and within-island diversification of
shrews illustrate the importance of speciation in
building Sundaland mammal diversity
TERRENCE C. DEMOS,*†‡ ANANG S. ACHMADI,§THOMAS C. GIARLA,*†¶
HERU HANDIKA,**†† MAHARADATUNKAMSI,§KEVIN C. ROWE**†† and JACOB A.
ESSELSTYN*†
*Museum of Natural Science, Louisiana State University, Baton Rouge, LA 70803, USA, †Department of Biological Sciences,
Louisiana State University, Baton Rouge, LA 70803, USA, ‡Science and Education, Field Museum of Natural History, Chicago,
IL 60605, USA, §Museum Zoologicum Bogoriense, Research Center for Biology-LIPI, Cibinong, Bogor 16911, Indonesia,
¶Department of Biology, Siena College, Loudonville, NY 12211, USA, **Sciences Department, Museum Victoria, Melbourne,
Vic. 3001, Australia, ††School of Biosciences, The University of Melbourne, Melbourne, Vic. 3001, Australia
Abstract
Island systems are important models for evolutionary biology because they provide
convenient, discrete biogeographic units of study. Continental islands with a history
of intermittent dry land connections confound the discrete definitions of islands and
have led zoologists to predict (i) little differentiation of terrestrial organisms among
continental shelf islands and (ii) extinction, rather than speciation, to be the main
cause of differences in community composition among islands. However, few conti-
nental island systems have been subjected to well-sampled phylogeographic studies,
leaving these biogeographic assumptions of connectivity largely untested. We analysed
nine unlinked loci from shrews of the genus Crocidura from seven mountains and two
lowland localities on the Sundaic continental shelf islands of Sumatra and Java. Coa-
lescent species delimitation strongly supported all currently recognized Crocidura spe-
cies from Sumatra (six species) and Java (five species), as well as one undescribed
species endemic to each island. We find that nearly all species of Crocidura in the
region are endemic to a single island and several of these have their closest relative(s)
on the same island. Intra-island genetic divergence among allopatric, conspecific popu-
lations is often substantial, perhaps indicating species-level diversity remains underes-
timated. One recent (Pleistocene) speciation event generated two morphologically
distinct, syntopic species on Java, further highlighting the prevalence of within-island
diversification. Our results suggest that both between- and within-island speciation
processes generated local endemism in Sundaland, supplementing the traditional view
that the region’s fauna is relictual and primarily governed by extinction.
Keywords:Crocidura, island biogeography, Java, phylogeography, speciation, Sumatra
Received 15 September 2015; revision received 9 August 2016; accepted 16 August 2016
Introduction
Islands are appealing natural laboratories of evolution
because the surrounding oceans represent obvious bar-
riers for terrestrial species (Wallace 1876, 1881; Schluter
2000; Grant & Grant 2011). However, in continental
island systems intermittent dry land connections should
reduce isolation, and endemism is anticipated only at
the scale of the entire region (Rosenzwieg 1995; Whit-
taker & Fern
andez-Palacios 2007). This regional ende-
mism paradigm has led zoologists to predict that (i)
terrestrial organisms are widespread within continental
island systems, (ii) little evolutionary differentiation
Correspondence: Terrence C. Demos, Fax: (225) 578 3075;
E-mail: terrencedemos@gmail.com
©2016 John Wiley & Sons Ltd
Molecular Ecology (2016) doi: 10.1111/mec.13820
occurs among meta-populations within a system, and
(iii) local extinction is the main cause of differences in
faunal diversity and composition among continental
islands of the same region (MacArthur & Wilson 1967;
Heaney 1986; Okie & Brown 2009).
Biologists often view the continental island system of
Sundaland (Malay Peninsula, Java, Sumatra, and Bor-
neo) in the context of this regional endemism paradigm
(Ruedi 1996; Ruedi & Fumagalli 1996; Gorog et al. 2004;
Okie & Brown 2009). This perspective dominates
because the assumed recurrent colonization during peri-
ods of low sea level should have reduced genetic differ-
entiation among metapopulations (e.g. Heaney 2000;
Papadopoulou & Knowles 2015). Recent comparative
phylogeographic and phylogenetic studies have
explored these issues using mtDNA sequences from
limited samples of Sundaic taxa. These studies have
generally concluded that Borneo harbours more distinc-
tive lineages than other Sundaic islands (de Bruyn et al.
2014; Leonard et al. 2015; Sheldon et al. 2015). However,
many of these studies largely exclude Javan lineages
because of a lack of samples. Furthermore, Bornean
material is often dominated by specimens from Malay-
sia, with little or no material from Kalimantan, which
represents 73% of the island’s land area.
Several authors have invoked extinction to explain
the differences in diversity and composition among ver-
tebrate communities of Sundaic islands (Heaney 1986;
Okie & Brown 2009; Wilting et al. 2012; den Tex & Leo-
nard 2013). These interpretations reinforce the regional
endemism paradigm (e.g. Brown 1986) and implicate
local extinction as the primary generator of b-diversity
within continental island systems. Alternatively how-
ever, at least some of the pattern could be explained by
between-island diversification, especially if it has
occurred between the larger islands. In essence, either
extinction or speciation can generate a species distribu-
tion that covers only a portion of a continental island
system, but absences of Sundaic species on particular
islands have traditionally been interpreted as extinc-
tions. While this interpretation is certainly true for
many examples (Piper et al. 2008; Cranbrook 2010),
some studies have found relatively deep mtDNA diver-
gences among populations both between and within
large Sundaic islands (Gorog et al. 2004; Esselstyn et al.
2010; Oliveros & Moyle 2010; Roberts et al. 2011), rais-
ing the idea that speciation may generate b-diversity on
the Sunda shelf. If isolating mechanisms at the intra-
island scale are sufficient to generate speciation, then
mechanisms at the larger, between-island scale also are
almost certainly sufficient to produce the same effect.
As such, densely sampled, fine-scale phylogeographic
studies may be more informative than sparsely sam-
pled, broad-scale studies at determining the relative
importance of speciation in generating b-diversity
among continental islands.
Terrestrial vertebrate species are not fully docu-
mented for many Sundaland taxa, further obfuscating
the historical formation of the region’s biota. Even
‘well-studied’ groups such as mammals are incom-
pletely known, as demonstrated by recent discoveries of
new species on the Sunda Shelf (Achmadi et al. 2012;
Esselstyn et al. 2014) and neighbouring areas (Heaney
et al. 2011; Esselstyn et al. 2012, 2015; Rowe et al. 2016).
Tropical regions often contain a glut of data-deficient
(DD) species (IUCN 2015). Those areas that are rich in
DD taxa also tend to harbour a disproportionate num-
ber of species that were recently described (Brito 2010)
or have narrow geographic distributions (Sheth et al.
2012). Among mammals, Sundaland is a ‘hotspot’ of
DD species (Bland et al. 2015), and hence, one might
expect many species and range extensions to await dis-
covery. These deficiencies in knowledge of Sundaland
species and their distributions may have biased biogeo-
graphic inferences (e.g. Heaney 2007; Esselstyn et al.
2010; Stelbrink et al. 2012; de Bruyn et al. 2014; Leonard
et al. 2015; Merckx et al. 2015).
Shrews in the genus Crocidura are prevalent members
of small mammal communities in Sundaland. However,
because of the cryptic nature of morphological diversity
in Crocidura, and the lack of adequate comparative ser-
ies, authors have often disagreed on the number and
composition of species in Sundaland. For example, a
series of morphological revisions and faunal summaries
(Jenkins 1982; Corbet & Hill 1992; Ruedi 1995) recog-
nized 2–6 species of Sumatran Crocidura, with 0–5of
them regarded as endemic. The systematics of Javan
shrews is somewhat better resolved, having been the
subject of recent molecular and morphological investi-
gations (Esselstyn et al. 2013, 2014), but some taxonomic
issues remain (see Materials and methods).
In this study, we used DNA sequences from nine
unlinked loci to (i) estimate species boundaries and
population structure of shrews within and between the
islands of Sumatra and Java, (ii) place these species in a
broad phylogenetic context and (iii) assess the geo-
graphic scale of endemism among the shrews of Suma-
tra and Java.
Materials and methods
Species sampling and study area
We sampled all 11 currently recognized species of
Sumatran and Javan Crocidura, as well as two putative
undescribed species (224 specimens total). Geographi-
cally, our sampling is derived from inventories of
shrew species from at least one site on each of five
©2016 John Wiley & Sons Ltd
2T. C. DEMOS ET AL.
Javan mountains (from west to east: Mts. Salak, Gede,
Ciremai, Slamet and Ijen); two Sumatran mountains
(Mts. Singgalang and Tujuh); and two lowland sites on
(Mt. Leuser National Park), and adjacent (Bangka
Island) to, Sumatra (Fig. 1). All sample sites were in
forested or forest edge habitats. Our phylogenetic
analyses included an additional 17 South-East Asian
Crocidura species (1–2 samples each) and an African
out-group, Crocidura monax. Hence, our total sampling
includes 263 specimens representing 28 recognized and
two putative species. This includes 19 of 20 species
known from Sundaland and 28 of 49 species known
from South-East Asia east of the Thailand–Myanmar
border and south of the Ryukyu Islands, including the
Philippines and Sulawesi (Jenkins et al. 2009, 2010, 2013;
Esselstyn & Goodman 2010; Esselstyn et al. 2010, 2014
[and references therein]; Appendix S1, Supporting infor-
mation). Preliminary identification to species was made
using morphological characters described in Jenkins
(1982) and Ruedi (1995) and later refined by examining
mtDNA gene tree topology, with subsequent re-exami-
nation of morphology.
Taxonomy of Javan and Sumatran Crocidura
Sumatran Crocidura can be provisionally grouped on
the basis of body size. Crocidura neglecta is much smaller
(<5 g) than other Sumatran species and was until
recently (Esselstyn et al. 2013) known only from the
holotype. Ruedi (1995) included C. neglecta in the wide-
spread Sundaland C.monticola complex, but Esselstyn
et al. (2013) found it to be a distant relative of C. monti-
cola from Java (the type locality). Ruedi (1995) indicated
that the other small Sumatran species, C.beccarii
(5–8 g), may be a close relative of C.vosmaeri
(5.8–8.8 g), a possible endemic to Bangka Island (Fig. 1).
Among medium-sized species, Ruedi (1995) described
C.hutanis (10–12 g) and recognized the long-tailed
species C.paradoxura as a Sumatran endemic. Finally,
the relatively large C.lepidura (13–21 g) was assigned to
Mt. Singgalang
Mt. Tujuh
Mt. Leuser
Mt. Salak
Mt. Gede
Mt. Ciremai Mt. Slamet
Mt. Ijen
SUMATRA
JAVA
0º
10ºS
100ºE110ºE
BORNEO
MALAYSIAN PEN.
Bangka
Island
Mt. Leuser: hutanis (3)
Mt. Singgalang: beccarii (3), paradoxura (10), sp. nov. 2 (15)
Mt. Tujuh: beccarii (20), lepidura (13), neglecta (3), paradoxura (8)
Bangka Island: vosmaeri (5)
Mt. Salak: brunnea (3), monticola (3)
Mt. Gede: abscondita (1), brunnea (11), monticola (37), orientalis (2), sp. nov. 1 (20)
Mt. Ciremai: brunnea (1), orientalis (22)
Mt. Slamet: brunnea (16), monticola (11), orientalis (5)
Mt. Ijen: brunnea (4), maxi (13), monticola (1)
Species sampled
Fig. 1 Map of sampling localities on Sumatra and Java. For each location sampled, we list the species followed in parentheses by the
number of individuals of Crocidura that we sequenced. Catalog numbers are given in Appendix S1 (Supporting information).
©2016 John Wiley & Sons Ltd
SHREW DIVERSITY ON THE SUNDA SHELF 3
the geographically widespread C.fuliginosa complex by
Jenkins (1982) and Corbet & Hill (1992). Ruedi (1995),
however, considered C.lepidura a Sumatran endemic.
Among Javan shrews, Esselstyn et al. (2013, 2014) recog-
nized the large C. orientalis and C. brunnea, the small
C. monticola and C. maxi and the newly discovered
C. abscondita (mistakenly named C. absconditus, which
uses the incorrect gender). In the light of new speci-
mens of true C. maxi from East Java, we now treat the
putative C. maxi from West Java, reported by Esselstyn
et al. (2013), as an undescribed species. Genetic data
(see Results) support this conclusion and render the
taxonomy outlined in this study more consistent with
previous studies (e.g. Kitchener et al. 1994).
Molecular methods
Specimens were sequenced for a portion of the mito-
chondrial cytochrome-b (cyt-b) and eight unlinked
nuclear genes, including seven exons (ApoB,BDNF,
BRCA1,GHR10,PTGER4,RAG1 and vWF) and one
intron (MCGF). Methods of DNA extraction, PCR and
sequencing follow those of Esselstyn et al. (2009, 2013).
Chromatographs were checked manually, assembled
and edited using GENEIOUS PRO 7.1.7 (Biomatters Ltd.).
Newly generated sequences were deposited in GenBank
(KX469457–KX470389; Appendix S1, Supporting infor-
mation). Sequences from each locus were aligned inde-
pendently using the MUSCLE algorithm (Edgar 2004)
with default settings in GENEIOUS. Sequence data from
cyt-b and the seven exons were translated into amino
acids and inspected for deletions, insertions and prema-
ture stop codons to prevent inclusion of paralogous
sequences. Alignments for all data sets were inspected
visually and determined to be unambiguous. Nuclear
alleles were resolved statistically using PHASE 2.1 (Ste-
phens et al. 2001) under default parameters, except that
we adjusted the haplotype acceptance threshold to 0.70,
which has been shown to reduce the number of unre-
solved genotypes with little to no increase in false posi-
tives (Garrick et al. 2010). Input files for PHASE were
assembled using the SEQPHASE web server (Flot 2010).
PHASE was run for 1000 iterations with a burn-in of 500
and a thinning interval of 1. We tested for recombina-
tion using the Detect Recombination plugin in GENEIOUS.
Substitution models and gene tree estimation
We used the Bayesian information criterion (BIC), as
implemented in PartitionFinder (Lanfear et al. 2012), to
identify the optimal partitioning scheme and best model
of nucleotide substitution for each partition in the cyt-b
alignment. The most appropriate model of evolution for
each unpartitioned nuclear gene was determined using
the BIC on the maximum-likelihood topology estimated
for each model independently in JMODELTEST v.2.1.7 (Dar-
riba et al. 2012). We used the greedy search algorithm
and linked branch lengths for likelihood score calcula-
tions in JMODELTEST. Gene trees were inferred using max-
imum-likelihood and Bayesian methods for cyt-b and
for individual, phased nuclear genes. Maximum-likeli-
hood estimates of gene trees were made in GARLI v.2.1
(Zwickl 2006) using default settings and 1000 bootstrap
replicates. GARLI runs were replicated five times for the
cyt-b locus and 100 times for the nuclear loci to ensure
consistency. The tree that received the highest likeli-
hood was reported for each analysis, and bootstrap
scores were summarized on these ML trees using Sum-
Trees in DENDROPY (Sukumaran & Holder 2010). Baye-
sian gene tree analyses used MRBAYES v.3.2.5 (Ronquist
et al. 2012) and two replicates were run to ensure
proper mixing had occurred. Nucleotide substitution
models were unlinked across partitions and were
allowed to evolve at individual rates in the cyt-b locus.
Eight Markov chains with default heating values were
conducted for 5 910
6
generations and sampled every
1000th generation. Stationarity was assessed using TRA-
CER v.1.6 (Rambaut et al. 2014). The first 1000 samples
were discarded as burn-in and the remaining 4000 sam-
ples formed the posterior probability (PP) distributions.
Majority rule consensus trees were generated from each
analysis.
Population structure
We clustered nuclear alleles using STRUCTURE v.2.3.4
(Pritchard et al. 2000; Hubisz et al. 2009) to infer popula-
tion-level diversity. Our goal was to determine whether
assignment of individuals to populations was consistent
with (i) clade membership inferred from the mtDNA
gene tree, (ii) the geographic origin of samples and (iii)
morphology-based species identifications. Mitochondrial
DNA was excluded to avoid circularity.
We carried out a hierarchical series of STRUCTURE anal-
yses for Sumatran and Javan species groups, as well as
individual species within each group. First, analyses
were conducted independently on six subsets of taxa,
where uncertainty in species limits was suggested by
earlier morphological (Jenkins 1982; Ruedi 1995) or
molecular studies (Esselstyn et al. 2009, 2013; Omar
et al. 2013), or because recently collected specimens
could not be assigned to any currently recognized taxon.
These sets were composed of (i) Crocidura monticola
(Java), C. sp. nov. 1 [Java (C. maxi in Esselstyn et al.
2013)] and C. sp. nov. 2 (Sumatra); (ii) C. beccarii
(Sumatra), C. vosmaeri (Bangka), C. lepidura (Sumatra)
and C. hutanis (Sumatra); (iii) C. orientalis (Java); (iv)
C. brunnea (Java); (v) C. maxi (Java and Lesser Sundas);
©2016 John Wiley & Sons Ltd
4T. C. DEMOS ET AL.
and (vi) C. paradoxura (Sumatra). Next, individual
STRUCTURE analyses were employed for each of the afore-
mentioned species, excluding those sampled from only
a single location, in which case they were analysed with
their putative sister species, as inferred on the rooted
mtDNA gene tree. We used the admixture model with
correlated allele frequencies to allow for mixed ancestry
of individuals. The number of clusters (K) was inferred
using 10 replicates for each Kvalue with a burn-in of
2910
4
–10
5
and 2–5910
5
iterations. The maximum
value of K(K
max
) for each analysis was calculated as
one more than the sum of the number of localities sam-
pled per species. Two independent runs of 10 replicates
were conducted for each pooled set of individuals at
each Kbetween 1 and K
max
. The optimal value of K
was determined using the DKmethod of Evanno et al.
(2005), implemented on the CLUMPAK web server (Kopel-
man et al. 2015). However, the DKmethod may under-
estimate the optimal number of clusters in the presence
of hierarchical structure (Waples & Gaggiotti 2006), and
has been shown to fail to recover the true value of K
when subpopulation sample sizes are small (<10) and
K>2, for example (Gao et al. 2011). Therefore, we also
heuristically examined the differences in log-likelihood
values among simulations to exclude DK-supported val-
ues that were biologically unrealistic (i.e. in conflict
with a combination of phylogenetic inference, morphol-
ogy and geographic distributions sensu Meirmans 2015).
Cluster membership probabilities were provided by
CLUMPP (Jakobsson & Rosenberg 2007), and results were
visualized using DISTRUCT (Rosenberg 2004), also on the
CLUMPAK server.
Lineage delimitation
We conducted joint independent lineage delimitation
and species tree estimation using the program BPP v.3.1
(Yang & Rannala 2010, 2014). Independent BPP analyses
were carried out on the same six subsets of taxa
described above for STRUCTURE analyses. Within each of
these subsets, we designated each species from each
locality as a putative independent lineage, effectively
putting a maximum on the number of lineages that
could be delimited. While it is a widespread practice to
use BPP to explicitly delimit species under a unified lin-
eage species concept (de Queiroz 2007), we refrain from
using BPP to formally diagnose species in this study
because of our inability to assess the possible influence
of isolation by distance on delimitation analyses.
Assigning localities as putative populations is a conser-
vative approach that consistently recovers the same
number of genetically isolated populations in BPP as
using a priori population assignment based on indepen-
dent analyses (Leach
e & Fujita 2010; Camargo et al.
2012; Demos et al. 2014b). We tested the validity of our
assignment of individuals to morphospecies using both
a guide tree generated from the multilocus species tree
inferred using *BEAST and the guide-tree-free implemen-
tation of BPP. Initial analyses showed that algorithms 0
and 1 (Yang & Rannala 2010) produced similar results;
therefore, algorithm 0 was implemented for subsequent
analysis. We used initial tuning values and Γshape
parameters chosen by Giarla et al. (2014) and trial runs
showed good mixing. For each of the five data sets, we
used all eight phased nuclear loci. Two Γ-distributed
prior probability schemes were used to compare the
effectsoflargeandsmallpopulationssizes(h=Γ[1, 10]
and Γ[2, 2000], respectively) on delimitation results.
The divergence time prior, s, used a diffuse Γ-distribu-
ted probability distribution Γ(2, 2000), with a mean of
0.001, which assumes that species split one million
years ago if substitution rates are 2.2 910
9
(Kumar &
Subramanian 2002) and generation time is equal to
1 year. All BPP analyses were run for 10
6
generations,
with a burn-in of 10
5
generations and samples were
drawn every fifth generation. We carried out a repli-
cated analysis of each data set to ensure convergence
and proper mixing of the rjMCMC algorithm. Thus, 10
BPP runs were conducted for each of the six data sets.
To ensure that BPP was not arbitrarily delimiting incor-
rect groups, we randomized individual assignments to
populations once, and ran BPP analyses following the
procedure of Burbrink et al. (2011).
Species tree estimation
We estimated a species tree in *BEAST 2.1.1 (Drummond
et al. 2012) using the eight nuclear (nDNA) alignments
and all species of Javan and Sumatran Crocidura, plus a
broader sample of SE Asian Crocidura. For those Javan
and Sumatran species for which samples from more
than one disjunct population exist, we assigned samples
from separate localities as terminal taxa, resulting in 42
tips in these analyses. The nDNA loci were reduced to
three individuals per species or population when n>3
to keep analyses tractable and facilitate convergence.
During initial runs, nucleotide substitution models
selected using JMODELTEST were applied to individual
loci; however, difficulty in achieving proper MCMC
mixing necessitated the use of simpler models. We
therefore adopted HKY models that did not include
Γ-distributed rate parameters or proportion of invariant
site parameters for six loci. The substitution, clock and
tree models were unlinked across all loci. The uncorre-
lated lognormal relaxed clock was applied to each locus
with a Yule tree prior and the constant root population
size model. Four replicate analyses were conducted
with random starting seeds and chain lengths of
©2016 John Wiley & Sons Ltd
SHREW DIVERSITY ON THE SUNDA SHELF 5
2910
9
generations, with parameters sampled every
2910
5
steps. Long chains were necessary for achieving
high effective sample sizes (ESS) for parameters. Con-
vergence was assessed in TRACER v.1.6 (Rambaut et al.
2014). The first 25% of trees were removed as burn-in,
and the maximum clade credibility tree was assembled
using LOGCOMBINER v.2.1.1 and TREEANNOTATOR v.2.0.3
(Drummond et al. 2012).
Estimating interspecific gene flow
Assessing gene flow, or a lack thereof, can indicate the
strength of putative ecological barriers and whether lin-
eages should be treated as independent species. There-
fore, we used a model-testing framework implemented
in IMA2 (Hey & Nielsen 2007; Hey 2010) to compare
models of divergence history with and without gene
flow. We analysed two pairs of species from Java (i) Cro-
cidura brunnea and C. orientalis that appear to be eleva-
tionally partitioned on Mt. Slamet; and (ii) C. monticola
and an undescribed species (C. maxi in Esselstyn et al.
2013) that co-occur at mid elevations on Mt. Gede (Essel-
styn et al. 2013). We estimated the joint PP of the migra-
tion parameters m
1
and m
2
for populations of the above
species pairs using our complete phased nDNA data set
and applied the HKY model for all genes. We per-
formed extensive preliminary runs to identify appropri-
ate bounds on demographic parameter priors and to
optimize the MCMC settings for sufficient mixing. Mix-
ing was assessed by inspection of ESS, parameter trend
plots, and update rates. The recording phase for both
species pairs included 30 independent Markov chains
for 10
6
steps sampled every 10 steps with a burnin of
2910
5
.ForC. monticola versus C. sp. nov.1, the upper
prior limits were q=3, t=8, m=10. For C. brunnea vs.
C. orientalis, the upper prior limits were q=3, t=8,
m=3. Both species pair analyses used a geometric
heating scheme (-h
a
=0.96 -h
b
=0.90). Two independent
M-mode runs with different starting seeds were
performed for each species-pair analysis. We used the
L-mode analyses to compare four nested migration
models against the full migration model: (i) individual
coalescent migration rates for species 0 and species 1; (ii)
a single coalescent migration rate for both species; (iii)
no migration from species 0 to species 1; (iv) no migra-
tion from species 1 to 0; and (v) an isolation model with
no migration (cf. Kerhoulas et al. 2015). From this out-
put, nested models were ranked by relative Akaike
information criterion (AIC) differences among models
using log(P) values from the L-mode analyses as
described in Carstens et al. (2009). Following Carstens
et al. (2009), we also calculated Akaike weights (x
i
, nor-
malized relative model likelihoods) and the evidence
ratio (E
min
/i, a comparison of each model to the best
model as an objective measure of model support) as
additional measures of model support. An evidence ratio
of <10 can be considered as moderate support for a
model relative to the best model (Burnham & Anderson
2002).
Results
Loci, taxon sampling and sequence alignment
Our cyt-b (1110 bp) alignment contained 245 individuals
(full or partial coverage), 92 of which were newly gener-
ated. The alignment includes 450 variable sites, 410 of
which are parsimony informative. To aid in visualization
of phylogenies inferred from this matrix, we reduced the
matrix of 245 individuals to a set of unique sequences,
resulting in a final alignment of 113 haplotypes. Com-
plete nDNA alignments (4421 bp total) for use in individ-
ual nuclear gene tree analyses contained 476–500 alleles
for each gene (Figs S1 and S2, Appendix S1, Supporting
information; Dryad doi: 10.5061/dryad.362pt). Overall,
~4% of data were missing for the 8 nDNA loci
(Appendix S1, Supporting information). The reduced
nDNA data set for species tree inference contained 98
individuals and 176–190 phased sequences per gene,
each with 23–126 variable sites and 19–117 parsimony
informative sites. We found no evidence of intralocus
recombination from the four-gamete tests.
Phylogenetic relationships
The cyt-b gene trees generated by MrBayes and GARLI
contained strong support for many nodes, but those sur-
rounded by short branches or in deeper parts of the tree
tended to receive limited support (Fig. 2). Javan and
Sumatran species or clades are dispersed across the
topology with four clades from each island. Interisland
phylogenetic relationships include the following: Cro-
cidura maxi from Java is strongly supported as sister to
C. maxi populations from the Lesser Sunda Islands and
Bali and this multi-island C. maxi clade is strongly sup-
ported as sister to C. elongata from Sulawesi; the Javan
endemics C. orientalis and C. brunnea are sisters and
together are strongly supported as sister to four species
endemic to Sumatra; another Javan endemic, C. abscon-
dita, is poorly supported as sister to C. negligens from
peninsular Malaysia; and finally, a well-supported clade
that includes two Javan endemic species (C. monticola
and C. sp. nov. 1) is a member of a polytomy with species
from throughout South-East Asia and Sumatra. A well-
supported clade of C. maxi whose island distributions
straddle Wallace’s Line was inferred (populations from
Aru +Alor +Java Islands). Crocidura monticola is
inferred to be paraphyletic, with C. sp. nov. 1 forming a
©2016 John Wiley & Sons Ltd
6T. C. DEMOS ET AL.
0.04
substitutions/site
C. abscondita (Java-Gede) FMNH212794
C. lepidura (Sumatra-Tujuh) FMNH212861
C. sp. nov. 2 (Sumatra-Singgalang) FMNH212969
C. attenuata (China) ROM116033
C. sp. nov. 2 (Sumatra-Singgalang) FMNH212965
C. beccarii (Sumatra-Tujuh) FMNH212818
C. brunnea (Java-Slamet) MZB32083
C. orientalis (Java-Ciremai) MZB28384
C. brunnea (Java-Salak) LSUMZ37946
C. maxi (Aru) WAM37975
C. fuliginosa (Vietnam) AMCC101526
C. monticola (Java-Ijen) LSUMZ37983
C. beccarii (Sumatra-Singgalang) FMNH212953
C. cf. neglecta (Borneo) KU168063
C. paradoxura (Sumatra-Singgalang) FMNH212958
C. lepidura (Sumatra-Tujuh) FMNH212856
C. paradoxura (Sumatra-Tujuh) FMNH212881
C. ninoyi (Philippines) FMNH145685
C. lepidura (Sumatra-Tujuh) FMNH213413
C. maxi (Bali) WAM38557
C. brunnea (Java-Gede) FMNH212744
C. maxi (Java-Ijen) LSUMZ37978
C. paradoxura (Sumatra-Tujuh) FMNH212886
C. beccarii (Sumatra-Singgalang) FMNH212950
C. grayi (Philippines) FMNH167217
C. orientalis (Java-Slamet) MZB32903
C. monax
C. brunnea (Java-Ijen) LSUMZ37945
C. brunnea (Java-Gede) FMNH212743
C. wuchihensis (China) ROM116090
C. malayana (P. Malaysia) IZEA3551
C. tanakae (Taiwan) NTU970
C. cf. neglecta (Borneo) UMMZ174683
C. orientalis (Java-Ciremai) MZB28393
C. maxi (Java-Ijen) LSUMZ37975
C. orientalis (Java-Slamet) MZB32901
C. neglecta (Sumatra-Tujuh) FMNH212877
C. paradoxura (Sumatra-Singgalang) FMNH212962
C. tanakae (Taiwan) NTU971
C. foetida (Borneo) USNM590298
C. paradoxura (Sumatra-Singgalang) FMNH212954
C. hutanis (Sumatra-Leuser) MVZ192174
C. maxi (Java-Ijen) LSUMZ37980
C. cf. neglecta (P. Malaysia) JX162650
C. maxi (Alor) WAM42567
C. brunnea (Java-Gede) FMNH212738
C. sp. nov. 1 (Java-Gede) FMNH212779
C. beccarii (Sumatra-Tujuh) FMNH212820
C. negrina (Philippines) KU165047
C. paradoxura (Sumatra-Tujuh) FMNH212884
C. monticola (Java-Gede) FMNH212747
C. vosmaeri (Sumatra-Bangka) LSUMZ38077
C. paradoxura (Sumatra-Singgalang) FMNH212957
C. paradoxura (Sumatra-Tujuh) FMNH212882
C. paradoxura (Sumatra-Singgalang) FMNH212956
C. maxi (Java-Ijen) LSUMZ37977
C. palawanensis (Philippines) FMNH195214
C. sp. nov. 2 (Sumatra-Singgalang) FMNH212974
C. cf. neglecta (Borneo) UMMZ174676
C. lepidura (Sumatra-Tujuh) FMNH212858
C. monticola (Java-Gede) MZB33644
C. lepidura (Sumatra-Tujuh) FMNH212859
C.sp. nov. 1 (Java-Gede) FMNH212763
C. orientalis (Java-Gede) FMNH212778
C. orientalis(Java-Ciremai) MZB28396
C. monticola (Java-Slamet) LSUMZ37984
C. orientalis (Java-Ciremai) MZB28402
C. monticola (Java-Slamet) LSUMZ37988
C. monticola (Java-Salak) MZB31721
C. beatus (Philippines) FMNH146965
C. lepidura (Sumatra-Tujuh) FMNH212863
C. monticola (Java-Gede) MZB33648
C. negligens (P. Malaysia) IZEA3560
C. panayensis (Philippines) KU164875
C. mindorus (Philippines) CMC3582
C. brunnea (Java-Ijen) LSUMZ37947
C. vosmaeri (Sumatra-Bangka) LSUMZ38078
C. vosmaeri(Sumatra-Bangka) LSUMZ37079
C. maxi (Java-Ijen) LSUMZ37981
C. panayensis (Philippines) KU164874
C. brunnea (Java-Ciremai) MZB28409
C. kurodai (Taiwan) NTU980
C. foetida (Borneo) USNM590299
C. orientalis (Java-Slamet) MZB32145
C. beccarii (Sumatra-Singgalang) FMNH212951
C. paradoxura (Sumatra-Singgalang) FMNH212955
C. hutanis (Sumatra-Leuser) MVZ192172
C. tanakae (Philippines) KU165845
C. negrina (Philippines) KU165046
C. maxi (Java-Ijen) LSU MZ37982
C. sp. nov. 2 (Sumatra-Singgalang) FMNH212967
C. brunnea (Java-Slamet) MZB32082
C. grayi (Philippines) FMNH194718
C. paradoxura (Sumatra-Tujuh) FMNH213415
C. beccarii (Sumatra-Tujuh) FMNH212834
C. vosmaeri (Sumatra-Bangka) LSUMZ37080
C. brunnea (Java-Ijen) LSUMZ37950
C. brunnea (Java-Ijen) LSUMZ37946
C. monticola (Java-Slamet) MZB32148
C. elongata (Sulawesi) LSUMZ36907
C. sp. nov. 1 (Java-Gede) FMNH212785
C. vosmaeri (Sumatra-Bangka) LSUMZ38076
C. sp. nov. 2 (Sumatra-Singgalang) FMNH212977
C. sp. nov. 2 (Sumatra-Singgalang) FMNH212952
C. beatus (Philippines) FMNH147819
C. sp. nov. 1 (Java-Gede) FMNH213410
C. brunnea (Java-Slamet) MZB32084
C. monticola (Java-Salak) MZB31720
C. elongata (Sulawesi) LSUMZ36906
C. palawanensis (Philippines) FMNH195215
C. orientalis (Java-Ciremai) MZB28380
C. nigripes (Sulawesi) IZEA4400
C. fuliginosa (P. Malaysia) IZEA3553
Fig. 2 Cytochrome-b gene tree inferred using maximum-likelihood and Bayesian inference in the programs GARLI and MrBayes,
respectively. Clades distributed on Sumatra are highlighted with green and clades distributed on Java are highlighted with blue.
Black circles on nodes indicate ML bootstrap ≥0.70 and Bayesian posterior probability (PP) ≥0.95. Black squares indicate ML boot-
strap ≥0.50 and <0.70, and PP ≥0.75 and <0.95. Nodes with ML bootstrap <0.50 and PP <0.75 are not marked.
©2016 John Wiley & Sons Ltd
SHREW DIVERSITY ON THE SUNDA SHELF 7
polytomy with two clades of C. monticola, which together
are sister to another population of C.monticola. From
Sumatra, we also recovered four clades that are consis-
tent with multiple origins for Sumatran species. At the
interisland/island–mainland level, the following rela-
tionships are recovered: Crocidura neglecta from Sumatra
is strongly supported as sister to C. cf. neglecta from Bor-
neo (formerly C. cf. monticola; Ruedi 1995); C. sp. nov.2
(Sumatra) is weakly supported as sister to C. wuchihensis
from China; C. paradoxura is a member of a polytomy that
includes multiple species from throughout South-East
Asia; and a Sumatran clade that includes four species is
strongly supported as sister to a Javan clade consisting of
C. brunnea and C. orientalis. A within-Sumatra multi-
species clade is recovered that includes C. beccarii,
C. hutanis,C. lepidura and C. vosmaeri. In total, at least
five intra-island (in situ) speciation events are inferred,
two for Java and three for Sumatra.
Evidence of possible introgressive hybridization is
evident based on incongruence of mtDNA gene trees
and morphology for samples of Crocidura orientalis from
Mt. Slamet on Java that are more closely related to
C. brunnea than to two additional C.orientalis lineages
from Mts. Ciremai and Gede. Both of these species are
morphologically diagnosable with external characters
(Ruedi 1995; Esselstyn et al. 2014). While sympatric
populations of C. orientalis and C. brunnea are also dis-
tributed on Mts. Ciremai and Gede, those samples sort
to their respective species-level clades. In addition, one
sample of C. beccarii from Mt. Singgalang, Sumatra,
was recovered as a member of the C. hutanis lineage
from lowland forest in Mt. Leuser NP, Sumatra. There
were no other mtDNA haplotypes shared among
localities.
ESS for all but two parameters exceeded 200 in the spe-
cies tree analysis. The exceptions were the tree likeli-
hoods for ApoB and Rag1, which were each >100.
Phylogenetic relationships inferred in our *BEAST analysis
generally agreed with the mtDNA gene trees in their sup-
port for topological relationships between the 13 Javan
and Sumatran species included in our analyses (Fig. 3).
Five separate Javan and/or Sumatran clades are inferred
in the species tree while seven are inferred in the mito-
chondrial gene trees. Species tree phylogenetic estimates
support Crocidura brunnea,C. orientalis and C. maxi from
Java and C. sp. nov. 2, C. neglecta,C. paradoxura,
C. hutanis and C. lepidura from Sumatra as monophyletic.
Crocidura sp. nov. 1 from Mt. Gede, Java, is nested within
the four C. monticola populations distributed across Java
and is sister to C. monticola from Mt. Ijen, the most dis-
tant Javan sample site (Fig. 1). Crocidura vosmaeri from
Bangka Island, just off Sumatra, is nested within C. becca-
rii where it is sister to the Mt. Singgalang population.
Population structure
We carried out STRUCTURE analyses to test for differentia-
tion between (i) Crocidura beccarii and C. vosmaeri,
(ii) C. lepidura and C. hutanis and (iii) the two putative
new species and their respective sister species/clades.
All six of these putative species are supported as dis-
tinctive using the delta Kmethod (Fig. 4 and Table 1;
Evanno et al. 2005). We also tested allopatric popula-
tions of individual species for isolation. The two sam-
pled populations of C. paradoxura from Sumatra also
were distinguished by STRUCTURE with minimal evidence
of admixture. The Javan species with samples available
from more than one population (i.e. C. monticola,
C. brunnea and C. orientalis) exhibit varying degrees of
population structure. In none of these three species was
each population assigned to a separate cluster (Fig. 4
and Table 1). The potentially widespread species of
C. maxi from eastern Java and the Lesser Sunda Islands,
and C. neglecta from Sumatra and Borneo, had best-sup-
ported Kvalues of two using the Evanno method
(Fig. S3, Supporting information). However, examina-
tion of ancestry proportions using DISTRUCT revealed no
population structure and the likelihood of models for
K=1inSTRUCTURE were the highest among the models
tested (K=1–4).
Coalescent delimitation
Coalescent analyses in BPP that treated each isolated
sample location as a potential species supported delimi-
tation of 20–22 lineages among the 13 species we recog-
nized from morphology and the mtDNA gene tree
topology. These delimitation results were minimally
affected by varying the prior distributions on mutation
rate-scaled effective population sizes (h) and divergence
times (s
0
). The combination of large ancestral popula-
tion sizes and shallow divergences resulted in margin-
ally lower support values for the populations and
species to which they were assigned (Table 1). We con-
sidered any PP ≥0.99 for any guide tree or prior scheme
as strong support for a putative speciation event
(Table 1). Randomization of individuals into clades
resulted in the collapse of all nodes that previously had
PP ≥0.99, indicating BPP is not simply delimiting all lin-
eages. The joint estimation of guide trees and delimita-
tion by BPP vs. implementation of the *BEAST generated
guide tree for delimitation resulted in modest variation
in posterior probabilities for speciation and one fewer
delimited species in the latter set of analyses (i.e. Cro-
cidura brunnea from Salak). All of the previously named
Sumatran lineages tested using BPP were distinct with
posterior probabilities of 1.0. Contrary to the results
©2016 John Wiley & Sons Ltd
8T. C. DEMOS ET AL.
C. attenuata (China)
C. monticola (Java-Slamet)
C. panayensis (Philippines)
C. hutanis (Sumatra-Leuser)
C. orientalis (Java-Gede)
C. tanakae (Taiwan)
C. monax
C. monticola (Java-Salak)
C. elongata (Sulawesi)
C. nigripes (Sulawesi)
C. lepidura (Sumatra-Tujuh)
C. beccarii (Sumatra-Tujuh)
C. maxi (Java-Ijen)
C. brunnea (Java-Ciremai)
C. wuchihensis (China)
C. mindorus (Philippines)
C. beatus (Philippines)
C. grayi (Philippines)
C. monticola (Java-Ijen)
C. sp. nov. 1 (Java-Gede)
C. cf. neglecta (Borneo)
C. brunnea (Java-Ijen)
C. fuliginosa (Vietnam)
C. beccarii (Sumatra-Singgalang)
C. monticola (Java-Gede)
C. orientalis (Java-Slamet)
C. brunnea (Java-Gede)
C. sp. nov. 2 (Sumatra-Singgalang)
C. vosmaeri (Sumatra-Bangka)
C. neglecta (Sumatra-Tujuh)
C. palawanensis (Philippines)
C. foetida (Borneo)
C. paradoxura (Sumatra-Singgalang)
C. orientalis (Java-Ciremai)
C. abscondita (Java-Gede)
C. maxi (Lesser Sundas)
C. ninoy (Philippines)
C. paradoxura (Sumatra-Tujuh)
C. brunnea (Java-Slamet)
C. brunnea (Java-Salak)
C. kurodai (Taiwan)
C. negrina (Philippines)
1.0E-3
Fig. 3 Multilocus species tree inferred using *BEAST. Posterior probabilities are indicated by filled circles if ≥0.95, filled squares if
≥0.85 and <0.95, and open circles if ≥0.70 and <0.85. Nodes with posterior probability (PP) <0.70 are not marked. Terminals are
labelled with species names followed by region of origin in parentheses. Javan species and populations are highlighted in blue and
Sumatran species and populations are highlighted in green. Red rectangular bars bisecting branches indicate results from BPP species
delimitation analyses with posterior probabilities ≥0.99 for a given lineage.
©2016 John Wiley & Sons Ltd
SHREW DIVERSITY ON THE SUNDA SHELF 9
from STRUCTURE,C. neglecta from Sumatra was delimited
from its sister lineage on Borneo. The total number of
putative species on Sumatra suggested by BPP was nine,
an increase of three over our initial morphological and
mtDNA conclusions.
Among the Javan samples analysed in BPP, we recov-
ered strong support (PP =1.0) to delimit 11 lineages on
Java. The only previous molecular phylogenetic assess-
ment of Javan Crocidura diversity based on fewer sample
localities found strong support for six species using the
same method (Esselstyn et al. 2013). Despite incomplete
lineage sorting or potential introgression of mitochon-
drial and nuclear loci between Crocidura brunnea and
C. orientalis on Mt. Slamet, we delimited lineages with
strong support within each of these species (Fig. 3,
Table 1). We also delimited populations of C. maxi from
eastern Java (Mt. Ijen) from non-Sunda Shelf populations
from the Lesser Sunda Islands with strong support.
Finally, we delimited, with a probability of 1.0 in all BPP
analyses, the putative undescribed species (C. sp. nov. 1)
from Mt. Gede (incorrectly referred to C. maxi by Essel-
styn et al. 2013) from a poorly supported sister lineage of
C. monticola from Mt. Ijen.The clade that includes all
populations of C. monticola +C. sp. nov. 1 from Java is
well supported as sister to C. sp. nov. 2 from Sumatra
and all populations within this clade are delimited with
high support. There is minimal sequence divergence
between C. monticola and C. sp. nov. 1 (3.8% mtDNA
uncorrected p-distance). However, in this case, several
external phenotypic characters make the two species (i.e.
C. monticola as currently described, and the putative
C. sp. nov. 1) diagnosable and support their status as
distinct species (Esselstyn et al. 2014). We did not use BPP
to test one recently described species from Java,
C. abscondita, as it is distantly related to other species in
both gene trees and the species tree. Based on available
samples, it has no close relatives and its known distribu-
tion is restricted to Mt. Gede.
Gene flow
Results from IMA2 were nearly identical between inde-
pendent runs, and therefore, we present results from
only one run. For the species pair on Mt. Slamet, the best-
supported model was unidirectional migration from
C. brunnea into C. orientalis based on the AIC scores of
ranked models from the L-mode analysis in IMA2
(Table S1, Supporting information). The model of no
migration (pure isolation model) was the least supported.
Our analyses indicated significant unidirectional gene
flow from C.brunnea into C. orientalis on Mt. Slamet
[Table 2; log-likelihood ratio test (LLR) =123.250,
P<0.001; posterior distribution peak at 0.21 migrants
per generation].There were also indications of hybridiza-
tion in one C. orientalis specimen based on shared alleles
at three nDNA loci (Figs S1 and S2, Supporting informa-
tion). Gene flow was near zero and not significant from
C. orientalis into C. brunnea (Table 2; LRR =0, n.s.). For
the species pair on Mt. Gede, the best-supported model
based on ranked AIC scores was unidirectional migration
from C. sp. nov.1 into C. monticola based on the ranked
models from the L-mode analysis (Table S2, Supporting
information). The model of no migration was the least
supported model. Analyses of these putative sibling spe-
cies indicated significant unidirectional gene flow from
C. sp. nov. 1 into C. monticola (Table 2; LLR =33.334,
Fig. 4 DISTRUCT visualization of STRUCTURE analyses assigning individuals to major populations for Sumatran and Javan Crocidura.
©2016 John Wiley & Sons Ltd
10 T. C. DEMOS ET AL.
P<0.001; 0.13 migrants per generation), but was not sig-
nificant from C. monticola into C. sp. nov. 1 (Table 2;
LLR =0, n.s.). The model of no migration was the least
supported model. These results suggest that violations of
the assumption of no gene flow among BPP delimited taxa
(Yang & Rannala 2014) are minimal because BPP will
infer one species when the migration rate is very high
(e.g. ≫1 immigrant per generation), while moderate
amounts of immigration (e.g. 1 immigrant per gen-
eration) had little impact on BPP performance (Zhang
et al. 2011).
Discussion
Our documentation of previously unrecognized shrew
diversity and relationships on Java and Sumatra
demonstrates substantial within-island endemism and
diversification in a continental island system. These
results, combined with those of other studies (e.g. Gorog
et al. 2004; Esselstyn et al. 2013), contradict a priori
expectations for low levels of interisland diversification
and large species ranges. Our phylogenetic inferences
suggest at least five intra-island speciation events in
Crocidura on Java and Sumatra. If we take our BPP
results literally, they support up to eight more intra-
island speciation events. These inferences suggest the
existence of isolating mechanisms that operate within
islands and produce species. As we argue above, if bar-
riers within islands are sufficient to generate speciation,
then between-island mechanisms also should be suffi-
cient. Therefore, past interpretations of faunal differ-
ences between islands of the Sunda shelf (e.g. Heaney
Table 1 Summary of results from BPP3 and STRUCTURE analyses. Prior schemes tested in BPP indicated by BPPsm and BPPlr. In both
schemes, divergence depths assuming a mutation rate of 10
9
substitutions per site per year (Kumar & Subramanian 2002) indicate a
divergence depth prior with a mean of ~1.0 Ma. BPPlr indicates a hprior of ~0.1 and BPPsm indicates a hprior of ~0.001. The guide
tree was based on the *BEAST species tree for the first pair of analyses (species tree guide tree). Species tree inference and species
delimitation were jointly inferred for the second pair of analyses without a priori provision of a guide tree (BPP guide tree free).
STRUCTURE admixture indicates that DISTRUCT plots for populations in which more than one individual had a partial assignment to
more than one group (q<0.90) were considered admixed. STRUCTURE assignment indicates the population or combined populations
that individuals were assigned to (numbers assigned to populations in left of Table 1). Agree indicates that BPP3 and STRUCTURE analy-
ses supported the same populations as genetically isolated. Monophyly indicates the cyt-b tree supports reciprocal monophyly for a
clade
Species tree
guide tree
BPP guide
tree free
STRUCTURE
admixture
STRUCTURE
assignment Agree MonophylyBPPlr BPPsm BPPlr BPPsm
Java 1 C. brunnea (Ciremai) 0.87 0.99 0.77 0.55 Admixed 1 +2+5—No
2C. brunnea (Gede) 0.82 0.99 0.77 0.55 Admixed 1 +2+5—No
3C. brunnea (Ijen) 1.0 1.0 1.0 1.0 Admixed 3 —No
4C. brunnea (Salak) 0.82 0.99 0.99 0.99 —4—No
5C. brunnea (Slamet) 1.0 1.0 1.0 1.0 Admixed 1 +2+5—No
6C. orientalis (Ciremai) 1.0 1.0 0.98 1.0 —6 Yes Yes
7C. orientalis (Gede) 1.0 1.0 0.98 1.0 —7+8—Yes
8C. orientalis (Slamet) 1.0 1.0 1.0 1.0 —7+8—Yes
9C. monticola (Gede) 1.0 1.0 1.0 1.0 Admixed 9 —Yes
10 C. monticola (Ijen) 1.0 1.0 1.0 1.0 —10 +11 +12 —Yes
11 C. monticola (Salak) 1.0 1.0 1.0 1.0 Admixed 10 +11 +12 —Yes
12 C. monticola (Slamet) 1.0 1.0 1.0 1.0 —10 +11 +12 —Yes
13 C. sp. nov.1 (Gede) 1.0 1.0 1.0 1.0 —13 Yes Yes
14 C. maxi (Ijen) 1.0 1.0 1.0 1.0 na na na Yes
15 C. maxi (Lesser Sundas) 1.0 1.0 1.0 1.0 na na na No
Sumatra 16 C. beccarii (Singgalang) 1.0 1.0 1.0 1.0 —16 Yes No
17 C. beccarii (Tujuh) 1.0 1.0 1.0 1.0 —17 Yes Yes
18 C. hutanis (Leuser) 1.0 1.0 1.0 1.0 —18 Yes Yes
19 C. lepidura (Tujuh) 1.0 1.0 1.0 1.0 —19 Yes Yes
20 C. vosmaeri (Bangka) 1.0 1.0 1.0 1.0 —20 Yes Yes
21 C. neglecta (Tujuh) 1.0 1.0 1.0 1.0 na na na Yes
22 C. cf. neglecta (Borneo) 1.0 1.0 1.0 1.0 na na na Yes
23 C. paradoxura (Singgalang) 1.0 1.0 1.0 1.0 —23 Yes Yes
24 C. paradoxura (Tujuh) 1.0 1.0 1.0 1.0 —24 Yes Yes
25 C. sp. nov.2 (Singgalang) 1.0 1.0 1.0 1.0 —25 Yes Yes
©2016 John Wiley & Sons Ltd
SHREW DIVERSITY ON THE SUNDA SHELF 11
1986; Okie & Brown 2009) may have overemphasized
the importance of local extinction of wide-range species
while discounting the importance of intra- and interis-
land diversification. If between-island speciation were
rampant, it would generate the same patterns (a species
is present on one island, but not another) that have
been interpreted as extinction of local populations.
Although the fossil record clearly demonstrates that
extinction has happened on Sundaic islands (Cranbrook
2010), we suggest that speciation is also an important
factor contributing to patterns of b-diversity.
Knowledge of the timing and extent of possible inter-
and intra-island barriers has been greatly expanded in
the last few years. For instance, paleoecological data
support the presence of continuous lowland dipterocarp
rainforest between Sumatra and Borneo, but not Java,
when the Sunda Shelf was exposed during glacial max-
ima (Raes et al. 2014). That study contradicted an earlier
supposition that extensive savannah habitats isolated
both Sumatra and Java from Borneo during glacial peri-
ods (Heaney 1991). Rather, data from Raes et al. (2014)
are consistent with recurrent interisland barriers to for-
est-dependent species between Java and other Sundaic
regions, but not between Sumatra and Borneo. Also,
new tectonic reconstructions support recurrent isolation
of Javan and Sumatran montane blocks as a result of
cyclical marine inundation (above present sea levels)
and volcanic activity up to the Pliocene–Pleistocene
boundary (Hall 2009; de Bruyn et al. 2014), providing a
possible mechanism for past isolation within modern
islands. Finally, reconstruction of Sundaland rainforest
coverage at the LGM suggests persistent and extensive
forest coverage in Borneo, but highly diminished and
fragmented forest coverage in Java, and an intermediate
level of forest contraction in Sumatra (Cannon et al.
2009; de Bruyn et al. 2014). These processes may have
isolated forest fragments both between and within
islands to varying extents, thereby producing idiosyn-
cratic patterns of genetic diversity in extant lineages
(Sheldon et al. 2015).
Distinct patterns of micro-endemism are apparent at
the species and population levels on both Java and
Sumatra. We recovered evidence for species-level diver-
gence between populations of Crocidura paradoxura and
C. beccarii from Mts. Singgalang and Tujuh. These peaks
are separated by ~190 km and are connected by contin-
uous montane or lower montane forest. Although the
divergences between populations of the two shrew spe-
cies may simply reflect isolation by distance, they
potentially represent divergence between closely related
allopatric species. However, isolation on mountains sep-
arated by a matrix of lowland forest or nonforest habi-
tats does not necessarily generate such levels of genetic
divergence. For example, analyses of shrew and rodent
populations distributed among disjunct montane forests
in Kenya, at similar distances between mountains, did
not support independent evolutionary lineages (Demos
et al. 2014a,b, 2015).
Environmentally, Java is very different from Sumatra.
Javan mountains are volcanic and typically separated by
wide expanses of drier lowland habitats, while Suma-
tran peaks are better connected by mountainous habi-
tats, providing a more obvious potential explanation for
genetic divergence between Javan populations. Thus,
although we anticipated detecting distinct genetic popu-
lations on isolated Javan peaks, the divergence observed
between Sumatran populations was more surprising.
Nevertheless, the phylogenetic relationships reflecting
within-island speciation and the small species ranges we
find on both islands are not consistent with the expecta-
tions of large species ranges and limited isolation in the
supposedly well-connected Sundaic system.
Diversity in Sundaland Crocidura
Through a combination of improved sampling from
new fieldwork and multilocus molecular analyses, we
found that shrew species diversity on the South-East
Asian continental shelf islands of Java and Sumatra is
underestimated, indicating the need for additional bio-
diversity surveys and taxonomic revisions. Estimates of
Crocidura diversity based on the most recent compre-
hensive morphological revision (Ruedi 1995) reported
only three Crocidura species on Java, two of which were
proposed as endemic. Our study, which expanded the
data set from Esselstyn et al. (2013), identified at least
six species on Java (C. abscondita,C. brunnea,C. maxi,
C. monticola,C. orientalis and an undescribed species),
including five endemic species. Using coalescent analy-
ses in BPP, we delimited an additional six lineages on
Java that may represent distinct species (three lineages
in C. brunnea, four in C. monticola and two in C. orien-
talis). Thus, at least six species are present on Java, but
as many as 12 may be represented in our current
Table 2 Mean number of migrants per generation between
geographically sympatric Crocidura populations using IMA2
From population To population
Migrants
per generation
C. brunnea (Slamet) C. orientalis (Slamet)
†
0.2069 (0.08–0.45)
C. orientalis (Slamet) C. brunnea (Slamet) 0.0007 (0.00–0.12)
C. monticola (Gede) C. sp. nov.1 (Gede) 0.0003 (0.00–0.08)
C. sp. nov. 1 (Gede) C. monticola (Gede)
†
0.1290 (0.05–0.27)
Results are based on eight nuclear loci. The 95% highest poste-
rior density is shown in parentheses.
†
Migration rates that are significantly different from zero at the
P<0.001 level in LLR tests (Nielsen and Wakeley 2001).
©2016 John Wiley & Sons Ltd
12 T. C. DEMOS ET AL.
sampling. We came to a similar conclusion for Sumatra,
where Ruedi (1995) diagnosed six species, five of which
he considered endemic. Our results recovered at least
seven species on Sumatra and a small neighbouring
island (C. beccarii,C. hutanis,C. lepidura,C. neglecta,
C. paradoxura,C. vosmaeri and an undescribed species).
All of these species are single-island endemics. Our BPP
analyses also delimited two lineages each of C. paradoxura
and C. beccarii, which suggests that nine or more
species of Crocidura may be endemic to Sumatra. The
large number of lineages delimited by BPP analyses with
a geographically sparse set of samples emphasizes the
need for additional specimen collection.
Our expanded geographic sampling that included
specimens from the eastern extremity of Java (Mt. Ijen)
also made clear the status of Crocidura maxi. This spe-
cies was previously recognized from East Java and the
Lesser Sunda Islands (Kitchener et al. 1994). Esselstyn
et al. (2013, 2014), however, identified specimens from
Mt. Gede (the first West Javan record) as C. maxi and
reported that they were not closely related to animals
from the Lesser Sundas. In this study, we obtained new
specimens of C. maxi from Mt. Ijen, which are closely
related to the Lesser Sunda shrews, but not the Mt.
Gede series. This clarifies that C. maxi is indeed present
in East Java and the Lesser Sundas, as Kitchener et al.
(1994) indicated, while the Mt. Gede series from Essel-
styn et al. (2013) is a new species (C. sp. nov. 1).
Syntopic sister species
On Mt. Gede, Java, Crocidura monticola was collected
together with a genetically and phenotypically distin-
guishable (see figs 3 and 4 in Esselstyn et al. 2014), but
as yet undescribed species, C. sp. nov. 1. We found an
apparent pattern of partial elevational overlap between
these species, with 16 C. monticola sampled at 1377 m,
32 C. monticola and 22 C. sp. nov.1 sampled at 1611 m,
and 13 C. sp. nov. 1 sampled at 1950 m. These two spe-
cies have an uncorrected pairwise cyt-b distance of 3.8%.
Median fossil calibrated multilocus divergence estimates
from Esselstyn et al. (2013) ranged from 178 000 to
515 000 generations ago, placing divergence in the Pleis-
tocene. Where the two species are syntopic, our results
from IMA2 analyses support moderate gene flow from C.
sp. nov.1 into C. monticola and a very low level of gene
flow in the other direction (Table 2). A migration rate
LLR test in IMA2 was only significant for gene flow from
C. sp. nov. 1 into C. monticola. These results suggest that
population divergence may have occurred with gene
flow, possibly along a single elevational gradient. It is
surprising that these two species evolved diagnosable
morphological differences since their very recent diver-
gence (Esselstyn et al. 2014). The fact that morphological
disparity has evolved in such a short time in an other-
wise morphologically conservative group suggests that
selection is involved. We suggest this species pair repre-
sents a plausible example of ecological speciation (e.g.
Nosil 2012) that could be tested with more data.
Conclusions
Our phylogenetic and phylogeographic analyses found
high levels of previously unrecognized inter- and intra-
island diversity. Inferences from multiple analyses
strongly support at least seven Sumatran and six Javan
Crocidura lineages as valid species. All but one of these
species is endemic to a single island, and several species
are only known from a single mountain. Two pairs of sis-
ter taxa on each of these islands suggest that at least five
within-island speciation events have occurred. The most
recent of these events generated two morphologically dis-
tinct species, with current populations occurring syntopi-
cally in at least one site. The newly recognized patterns of
endemism, in which no species of Crocidura is widespread
on the Sunda shelf, indicate that evolutionary processes
on these islands may be more similar to those reported
for oceanic archipelagos (Heaney et al. 2011; Justiniano
et al. 2015), where species ranges are often smaller than
the islands themselves. This is in stark contrast to the tra-
ditional expectation that species should be widespread in
continental island systems such as Sundaland, and war-
rants reconsideration of speciation as part of the processes
that generated b-diversity across this island system.
Acknowledgments
We thank P. Putri, R. Robi, R. Kurnia, N. Supriatna and
Apandi for their assistance with fieldwork. The National
Science Foundation (OISE-0965856 and DEB-1343517) provided
financial support. Staff at the Field Museum of Natural History
(R. Banasiak, A. Goldman, L. Heaney, A. Niedzielski and the
late W. Stanley), Museum Zoologicum Bogoriense (N. Supri-
atna and Apandi) and LSUMZ (S. Cardiff) provided invaluable
curatorial support. We thank N. Kerhoulas for helpful advice
on the analyses. Kementerian Riset dan Teknologi, Kerinci
Seblat National Park and Balai Konservasi dan Sumber Daya
Alam (BKSDA) Sumatera Barat provided research permits. We
thank the Ambrose Monell Cryo Collection, American Museum
of Natural History (AMCC); Cincinnati Museum Center
(CMC); Field Museum of Natural History, Chicago (FMNH);
Louisiana State University Museum of Natural Science, Baton
Rouge (LSUMZ); Museum of Vertebrate Zoology, University of
California, Berkeley (MVZ); Museum Zoologicum Bogoriense,
Bogor, Indonesia (MZB); National University of Taiwan (NTU);
Royal Ontario Museum (ROM); University of Kansas Natural
History Museum (KU); University of Lausanne (IZEA); and
Western Australian Museum (WAM) for providing access to
voucher specimens. This material is based upon work sup-
ported by HPC@LSU computing resources.
©2016 John Wiley & Sons Ltd
SHREW DIVERSITY ON THE SUNDA SHELF 13
References
Achmadi AS, Maryanto I, Maharadatunkamsi (2012)
Systematic and descriptions of new species within genus
Maxomys from East Kalimantan, Borneo Island. Treubia,39,
1–28.
Bland LM, Collen B, Orme CDL, Bielb J (2015) Predicting the
conservation status of data-deficient species. Conservation
Biology,29, 250–259.
Brito D (2010) Overcoming the Linnean shortfall: data defi-
ciency and biological survey priorities. Basic and Applied Ecol-
ogy,11, 709–713.
Brown JH (1986) Two decades of interaction between the
MacArthur-Wilson model and the complexities of mam-
malian distributions. Biological Journal of the Linnean Society,
28, 231–251.
de Bruyn M, Stelbrink B, Morley RJ et al. (2014) Borneo and
Indochina are major evolutionary hotspots for Southeast
Asian biodiversity. Systematic Biology,63, 879–901.
Burbrink FT, Yao H, Ingrasci M et al. (2011) Speciation at the
Mogollon Rim in the Arizona Mountain Kingsnake (Lampro-
peltis pyromelana). Molecular Phylogenetics and Evolution,60,
445–454.
Burnham KP, Anderson DR (2002) Model Selection and Multi-
model Inference: A Practical Information-Theoretic Approach.
Springer-Verlag, New York.
Camargo A, Morando M, Avila LJ, Sites JW (2012) Species
delimitation with ABC and other coalescent-based methods:
a test of accuracy with simulations and an empirical example
with lizards of the Liolaemus darwinii complex (Squamata:
Liolaemidae). Evolution,66, 2834–2849.
Cannon CH, Morley RJ, Bush ABG (2009) The current refugial
rainforests of Sundaland are unrepresentative of their
biogeographic past and highly vulnerable to disturbance.
Proceedings of the National Academy of Sciences of the United
States of America,106, 11188–11193.
Carstens BC, Stoute HN, Reid NM (2009) An information-theo-
retical approach to phylogeography. Molecular Ecology,18,
4270–4282.
Corbet GB, Hill JE (1992) The Mammals of the Indomalayan
Region: A Systematic Review. Oxford University Press, Oxford.
Cranbrook E (2010) Later quaternary turnover of mammals in
Borneo: the zooarchaeological record. Biodiversity Conserva-
tion,19, 373–391.
Darriba D, Taboada GL, Doallo R, Posada D (2012) jModelTest
2: more models, new heuristics and parallel computing. Nat-
ure Methods,9, 772.
Demos TC, Agwanda B, Hickerson MJ (2014a) Integrative tax-
onomy of Hylomyscus (Rodentia: Muridae) and description of
a new species from western Kenya. Journal of Mammalogy,95,
E1–E15.
Demos TC, Kerbis Peterhans JC, Agwanda B, Hickerson MJ
(2014b) Uncovering cryptic diversity and refugial persistence
among small mammal lineages across the Eastern Afromon-
tane biodiversity hotspot. Molecular Phylogenetics and Evolu-
tion,71,41–54.
Demos TC, Kerbis Peterhans JC, Joseph TA et al. (2015) Com-
parative population genomics of African montane forest
mammals support population persistence across a climatic
gradient and Quaternary climatic cycles. PLoS ONE,10,
e0131800.
Drummond AJ, Suchard MA, Xie D, Rambaut A (2012) Baye-
sian phylogenetics with BEAUti and the BEAST 1.7. Molecu-
lar Biology and Evolution,29, 1969–19673.
Edgar RC (2004) MUSCLE: multiple sequence alignment with
high accuracy and high throughput. Nucleic Acids Research,
32, 1792–1797.
Esselstyn JA, Goodman SM (2010) New species of shrew (Sori-
cidae: Crocidura) from Sibuyan Island, Philippines. Journal of
Mammalogy,91, 1467–1472.
Esselstyn JA, Timm RM, Brown RM (2009) Do geological or cli-
matic processes drive speciation in dynamic archipelagos?
The tempo and mode of diversification in Southeast Asian
shrews. Evolution,63, 2595–2610.
Esselstyn JA, Oliveros CH, Moyle RG et al. (2010) Integrating
phylogenetic and taxonomic evidence illuminates complex
biogeographic patterns along Huxley’s modification of Wal-
lace’s Line. Journal of Biogeography,37, 2054–2066.
Esselstyn JA, Achmadi AS, Rowe KC (2012) Evolutionary nov-
elty in a rat with no molars. Biology Letters,8, 990–993.
Esselstyn JA, Maharadatunkamsi, Achmadi AS et al. (2013)
Carving out turf in a biodiversity hotspot: multiple, previ-
ously unrecognized shrew species co-occur on Java Island,
Indonesia. Molecular Ecology,22, 4972–4987.
Esselstyn JA, Achmadi AS, Maharadatunkamsi (2014) A new
species of shrew (Soricomorpha: Crocidura) from West Java,
Indonesia. Journal of Mammalogy,95, 216–224.
Esselstyn JA, Achmadi AS, Handika H, Rowe KC (2015) A
hog-nosed shrew rat (Rodentia: Muridae) from Sulawesi
Island, Indonesia. Journal of Mammalogy,96, 895–907.
Evanno G, Regnaut S, Goudet J (2005) Detecting the number of
clusters of individuals using the software STRUCTURE: a
simulation study. Molecular Ecology,14, 2611–2620.
Flot JF (2010) SEQPHASE: a web tool for interconverting phase
input/output files and fasta sequence alignments. Molecular
Ecology Resources,10, 162–166.
Gao H, Bryc K, Bustamante CD (2011) On identifying the opti-
mal number of population clusters via the deviance informa-
tion criterion. PLoS ONE,6, e21014.
Garrick RC, Sunnucks P, Dyer RJ (2010) Nuclear gene phylo-
geography using PHASE: dealing with unresolved geno-
types, lost alleles, and systematic bias in parameter
estimation. BMC Evolutionary Biology,10, 118.
Giarla TC, Voss RS, Jansa SA (2014) Hidden diversity in the
Andes: comparison of species delimitation methods in mon-
tane marsupials. Molecular Phylogenetics and Evolution,70,
137–151.
Gorog AJ, Sinaga MH, Engstrom MD (2004) Vicariance or
dispersal? Historical biogeography of three Sunda shelf mar-
ine rodents (Maxomys surifer,Leopoldamys sabanus and Max-
omys whiteheadi). Biological Journal of the Linnean Society,81,
91–109.
Grant PR, Grant BR (2011) How and Why Species Multiply: The
Radiation of Darwin’s Finches. Princeton University Press,
Princeton, New Jersey.
Hall R (2009) Southeast Asia’s changing palaeogeography. Blumea,
54, 148–161.
Heaney LR (1986) Biogeography of mammals in SE Asia: esti-
mates of rates of colonization, extinction and speciation. Bio-
logical Journal of the Linnean Society,28, 127–165.
Heaney LR (1991) A synopsis of climate and vegetational
changes in Southeast Asia. Climatic Change,19,53–61.
©2016 John Wiley & Sons Ltd
14 T. C. DEMOS ET AL.
Heaney LR (2000) Dynamic disequilibrium: a long-term, large-
scale perspective on the equilibrium model of island bio-
geography. Global Ecology and Biogeography,9,59–74.
Heaney LR (2007) Is a new paradigm emerging for oceanic
island biogeography? Journal of Biogeography,34, 753–757.
Heaney LR, Balete DS, Rickart EA et al. (2011) Seven new species
and a new subgenus of forest mice (Rodentia: Muridae:
Apomys) from Luzon Island. Fieldiana Life and Earth Sciences,2,
1–60.
Hey J (2010) Isolation with migration models for more than
two populations. Molecular Biological Evolution,27, 905–920.
Hey J, Nielsen R (2007) Integration within the Felsenstein equa-
tion for improved Markov chain Monte Carlo methods in
population genetics. Proceedings of the National Academy of
Sciences of the United States of America,104, 2785–2790.
Hubisz MJ, Falush D, Stephens M, Pritchard JK (2009)
Inferring weak population structure with the assistance of
sample group information. Molecular Ecology Resources,9,
1322–1332.
IUCN (2015) The IUCN Red List of Threatened Species. Version
2015-4. Available from http://www.iucnredlist.org.
Jakobsson M, Rosenberg NA (2007) CLUMPP: a cluster match-
ing and permutation program for dealing with label switch-
ing and multimodality in analysis of population structure.
Bioinformatics,23, 1801–1806.
Jenkins PD (1982) A discussion of Malayan and Indonesian
shrews of the genus Crocidura (Insectivora: Soricidae). Zoolo-
gische Mededelingen,56, 267–279.
Jenkins PD, Lunde DP, Moncrieff CB (2009) Descriptions of
new species of Crocidura (Soricomorpha: Soricidae) from
mainland Southeast Asia, with synopses of previously
described species and remarks on biogeography. Bulletin of
the American Museum of Natural History,331, 356–405.
Jenkins PD, Abramov AV, Rozhnov VV, Olsson A (2010) A new
species of Crocidura (Soricomorpha: Soricidae) from southern
Vietnam and north-eastern Cambodia. Zootaxa,2345,60–68.
Jenkins PD, Abramov AV, Bannikova AA, Rozhnov VV (2013)
Bones and genes: resolution problems in three Vietnamese
species of Crocidura (Mammalia, Soricomorpha, Soricidae)
and the description of an additional new species. ZooKeys,
313,61–79.
Justiniano R, Schenk JJ, Balete DS et al. (2015) Testing diversifi-
cation models of endemic Philippine forest mice (Apomys)
with nuclear phylogenies across elevational gradients reveals
repeated colonization of isolated mountain ranges. Journal of
Biogeography,42,51–64.
Kerhoulas NJ, Gunderson AM, Olson LE (2015) Complex his-
tory of isolation and gene flow in hoary, Olympic, and
endangered Vancouver Island marmots. Journal of Mammal-
ogy,96, 810–826.
Kitchener DJ, Hisheh S, Schmitt LH, Suyanto A (1994) Shrews
(Soricidae: Crocidura) from the Lesser Sunda Islands, and
South-East Maluku, Eastern Indonesia. Australian Mammal-
ogy,17,7–17.
Kopelman NM, Mayzel J, Jakobsson M et al. (2015) CLUMPAK
: a program for identifying clustering modes and packaging
population structure inferences across K.Molecular Ecology
Resources,15, 1179–1191.
Kumar S, Subramanian S (2002) Mutation rates in mammalian
genomes. Proceedings of the National Academy of Sciences of the
United States of America,99, 803–808.
Lanfear R, Calcott B, Ho SYW, Guindon S (2012) PartitionFin-
der: combined selection of partitioning schemes and substi-
tution models for phylogenetic analyses. Molecular Biology
and Evolution,29, 1695–1701.
Leach
e AD, Fujita MK (2010) Bayesian species delimitation
in West African forest geckos (Hemidactylus fasciatus).
Proceedings of the Royal Society B-Biological Sciences,277,
3071–3077.
Leonard JA, den Tex R-J, Hawkins MTR et al. (2015) Phylo-
geography of vertebrates on the Sunda Shelf: a multi-species
comparison. Journal of Biogeography,42, 871–879.
MacArthur RH, Wilson EO (1967) The Theory of Island
Biogeography. Princeton University Press, Princeton, New
Jersey.
Meirmans PG (2015) Seven common mistakes in population
genetics and how to avoid them. Molecular Ecology,13, 3223–
3231.
Merckx VSFT, Hendriks KP, Beentjes KK et al. (2015) Evolution
of endemism on a young tropical mountain. Nature,524,
347–350.
Nielsen R, Wakeley J (2001) Distinguishing migration from iso-
lation: a Markov chain Monte Carlo approach. Genetics,158,
885–896.
Nosil P (2012) Ecological Speciation. Oxford University Press,
New York.
Okie JG, Brown JH (2009) Niches, body sizes, and the disas-
sembly of mammal communities on the Sunda Shelf islands.
Proceedings of the National Academy of Sciences of the United
States of America,106(Suppl.), 19679–19684.
Oliveros CH, Moyle RG (2010) Origin and diversification of
Philippine bulbuls. Molecular Phylogenetics and Evolution,54,
822–832.
Omar H, Hashim R, Bhassu S, Ruedi M (2013) Morphological
and genetic relationships of the Crocidura monticola species
complex (Soricidae: Crocidurinae) in Sundaland. Mammalian
Biology,78, 446–454.
Papadopoulou A, Knowles LL (2015) Genomic tests of the
species-pump hypothesis: recent island connectivity cycles
drive population divergence but not speciation in
Caribbean crickets across the Virgin Islands. Nature,69,
1501–1517.
Piper PJ, Ochoa J, Lewis H, Paz V, Ronquillo WP (2008) The
first evidence for the past presence of the tiger Panthera tigris
(L.) on the island of Palawan, Philippines: extinction in an
island population. Palaeogeography, Palaeoclimatology, Palaeoe-
cology,264, 123–127.
Pritchard JK, Stephens M, Donnelly P (2000) Inference of popu-
lation structure using multilocus genotype data. Genetics,
155, 945–959.
de Queiroz K (2007) Species concepts and species delimitation.
Systematic Biology,56, 879–886.
Raes N, Cannon CH, Hijmans RJ et al. (2014) Historical distri-
bution of Sundaland’s Dipterocarp rainforests at Quaternary
glacial maxima. Proceedings of the National Academy of Sciences
of the United States of America,111, 16790–16795.
Rambaut A, Suchard MA, Xie D, Drummond AJ (2014) Tracer
v1.6, Available from http://beast.bio.ed.ac.uk/Tracer.
Roberts TE, Lanier HC, Sargis EJ, Olson LE (2011) Molecular
phylogeny of treeshrews (Mammalia: Scandentia) and the
timescale of diversification in Southeast Asia. Molecular
Phylogenetics and Evolution,60, 358–372.
©2016 John Wiley & Sons Ltd
SHREW DIVERSITY ON THE SUNDA SHELF 15
Ronquist F, Teslenko M, van der Mark P et al. (2012) MrBayes
3.2: efficient Bayesian phylogenetic inference and model choice
across a large model space. Systematic Biology,61,539–542.
Rosenberg NA (2004) DISTRUCT: a program for the graphical
display of population structure. Molecular Ecology Notes,4,
137–138.
Rosenzwieg ML (1995) Species Diversity in Space and Time. Cam-
bridge University Press, Cambridge.
Rowe KC, Achmadi AS, Esselstyn JA (2016) A new genus and
species of omnivorous rodent (Muridae: Murinae) from Sula-
wesi, nested within a clade of endemic carnivores. Journal of
Mammaolgy,97, 978–991.
Ruedi M (1995) Taxonomic revision of shrews of the genus
Crocidura from the Sunda Shelf and Sulawesi with a descrip-
tion of two new species (Mammalia: Soricidae). Zoological
Journal of the Linnean Society,115, 211–265.
Ruedi M (1996) Phylogenetic evolution and biogeography of
Southeast Asian shrews (genus Crocidura: Soricidae). Biologi-
cal Journal of the Linnean Society,58, 197–219.
Ruedi M, Fumagalli L (1996) Genetic structure of Gymnures
(genus Hylomys; Erinaceidae) on continental islands of South-
east Asia: historical effects of fragmentation. Journal of Zoo-
logical Systematics and Evolutionary Research,34, 153–162.
Schluter D (2000) The Ecology of Adaptive Radiation. Oxford
University Press, Oxford.
Sheldon FH, Lim HC, Moyle RG (2015) Return to the Malay
Archipelago: the biogeography of Sundaic rainforest birds.
Journal of Ornithology,156, S91–S113.
Sheth SN, Lohmann LG, Distler T, Jim
enez I (2012) Under-
standing bias in geographic range size estimates. Global Ecol-
ogy and Biogeography,21, 732–742.
Stelbrink B, Albrect C, Hall R, von Ritelen T (2012) The bio-
geography of Sulawesi revisited: is there evidence for a
vicariant origin of taxa on Wallace’s “anomalous island”?
Evolution,66, 2252–2271.
Stephens M, Smith NJ, Donnelly P (2001) A new statistical
method for haplotype reconstruction from population data.
American Journal of Human Genetics,68, 978–989.
Sukumaran J, Holder MT (2010) DendroPy: a Python library
for phylogenetic computing. Bioinformatics,26, 1569–1571.
den Tex R-J, Leonard JA (2013) A molecular phylogeny of
Asian barbets: speciation and extinction in the tropics. Molec-
ular Phylogenetics and Evolution,68,1–13.
Wallace AR (1876) The Geographical Distribution of Animals, With
a Study of the Relations of Living and Extinct Faunas as Elucidat-
ing the Past Changes of the Earth’s Surface. Harper & Brothers,
New York.
Wallace AR (1881) Island Life. Harper & Brothers, New York.
Waples RS, Gaggiotti O (2006) What is a population? An
empirical evaluation of some genetic methods for identifying
the number of gene pools and their degree of connectivity.
Molecular Ecology,15, 1419–1439.
Whittaker RJ, Fern
andez-Palacios JM (2007) Island Biogeography:
Ecology, Evolution and Conservation. Oxford University Press,
Oxford.
Wilting A, Sollmann R, Meijaard E et al. (2012) Mentawai’s
endemic, relictual fauna: is it evidence for Pleistocene extinc-
tions on Sumatra? Journal of Biogeography,39, 1608–1620.
Yang Z, Rannala B (2010) Bayesian species delimitation using
multilocus sequence data. Proceedings of the National Academy
of Sciences of the United States of America,107, 9264–9269.
Yang Z, Rannala B (2014) Unguided species delimitation using
DNA sequence data from multiple loci. Molecular Biology and
Evolution,31, 3125–3135.
Zhang C, Zhang D-X, Zhu T, Yang Z (2011) Evaluation of a
Bayesian coalescent method of species delimitation. System-
atic Biology,60, 747–761.
Zwickl DJ (2006) Genetic algorithm approaches for the phylogenetic
analysis of large biological sequence datasets under the maximum
likelihood criterion. PhD Dissertation, The University of Texas
at Austin, Austin, Texas.
Data accessibility
DNA sequence data: GenBank, Accession numbers
KX469457-KX470389. DNA sequence alignments: Dryad
Accession doi: 10.5061/dryad.362pt.
T.C.D., T.C.G. and J.A.E. designed the study; J.A.E.,
A.S.A., H.H., M. and K.C.R. conducted fieldwork and
identified specimens; T.C.D., T.C.G. and J.A.E.
sequenced DNA; T.C.D. and T.C.G. analysed the data;
T.C.D., J.A.E. and K.C.R. wrote the manuscript with
editorial contributions from T.C.G., A.S.A. and H.H.
Supporting information
Additional supporting information may be found in the online ver-
sion of this article.
Appendix S1 List of the museum voucher numbers, localities,
elevations, and GenBank accession numbers for all specimens
used in this study. NA indicates samples intentionally not
included in the study and blank cells indicate failure of poly-
merase chain reaction amplification.
Table S1 Results of pairwise IMA2 L-mode analyses using
ranked nested-models of migration for Crocidura brunnea and
C. orientalis.
Table S2 Results of pairwise IMA2 L-mode analyses using
ranked nested-models of migration for Crocidura monticola and
C. sp. nov.1.
Fig. S1 Maximum likelihood gene tree estimates of phased
alleles from Southeast Asian shrews (genus Crocidura) for (A)
ApoB, (B) BDNF, (C) BRCA1, (D) GHR10, (E) MCGF, (F)
PTGER4, (G) RAG1, and (H) vWF.
Fig. S2 Bayesian gene tree estimates of phased alleles from
Southeast Asian shrews (genus Crocidura) for (A) ApoB, (B)
BDNF, (C) BRCA1, (D) GHR10, (E) MCGF, (F) PTGER4, (G)
RAG1, and (H) vWF.
Fig. S3 Results from STRUCTURE analyses of eight nuclear loci
from 12 Crocidura species (A–G) with the number of popula-
tions (K) varying from 1–7.
©2016 John Wiley & Sons Ltd
16 T. C. DEMOS ET AL.
