Page 1
s
es
g
os
a Department of Biology, Indiana University, Bloomington, IN 47405, USA
b Department of Ecology, Evolution, and Environmental Biology, Columbia University, New York, NY 10027, USA
c New York Consortium in Evolutionary Primatology (NYCEP), USA
like a DNA barcode, with potential applications to conservation.
the crested gibbons, genus Nomascus (four species), and
the genus Hylobates, formerly known as the lar group
(six species) (Brandon-Jones et al., 2004; Mootnick and
‘‘lar group’’). Past analyses have disagreed on the Kloss’s
gibbon’s placement in this genus; however, we agree on
its inclusion in this group based on a variety of shared
characters, including cranial shape, intermembral index,
genital features, and especially chromosome number
(2n = 44) (Groves, 2001).
* Corresponding author. Fax: +1 812 855 6785.
E-mail address: djwhitta@indiana.edu (D.J. Whittaker).
Molecular Phylogenetics and Evolut
ARTICLE IN PRESS� 2007 Elsevier Inc. All rights reserved.
Keywords: Hylobatidae; Southeast Asia; mtDNA; Partition homogeneity test
1. Introduction
The gibbons (Family Hylobatidae) are the small apes,
endemic to the forests of South and Southeast Asia, known
for their musical morning calls and their brachiating loco-
motion. There are currently 12 recognized species in four
genera: the siamang, Symphalangus syndactylus, the hoo-
lock gibbon, Hoolock hoolock (formerly Bunopithecus),
Groves, 2005). The relationships among these four genera
are not yet fully agreed upon; however, most analyses show
Hylobates to be the most derived taxon within this
radiation.
The genus Hylobates includes six species: lar, agilis,
moloch, muelleri, pileatus, and klossii. The first five have
long been considered closely related, and are virtually
indistinguishable on the basis of cranial characters (hence,d Center for Environmental Research and Conservation (CERC), Columbia University, New York, NY 10027, USA
Received 5 February 2007; revised 25 June 2007; accepted 10 August 2007
Abstract
Gibbons of the genus Hylobates likely speciated very rapidly following isolation by rising sea levels during the Pleistocene. We
sequenced the hypervariable region I (HV-I) of the mitochondrial D-loop to reconstruct the phylogeny of this group. Although the
results clearly supported monophyly of each of the six species, the relationships among them were not clearly resolved by these data
alone. A homogeneity test against published data sets of a coding mitochondrial locus (ND3–ND4 region), behavioral characters (vocal-
izations), and morphological traits (including skeletal and soft tissue anatomy) revealed no significant incongruence, and combining them
resulted in a phylogenetic tree with much stronger support. The Kloss’s gibbon (H. klossii), long considered a primitive taxon based on
morphology, shares many molecular and vocal characteristics with the Javan gibbon (H. moloch), and appear as the most recently
derived species. The northernmost species (H. lar and H. pileatus) are the most basal taxa. These data suggest that ancestral gibbons
radiated from north to south. Unlike other markers, the HV-I region can accurately identify members of different gibbon species muchResolution of the Hylobate
mitochondrial D-loop sequenc
and morpholo
Danielle J. Whittaker a,*, Juan Carl1055-7903/$ - see front matter � 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.ympev.2007.08.009
Please cite this article in press as: Whittaker, D.J. et al., Resolution o
Evol. (2007), doi:10.1016/j.ympev.2007.08.009phylogeny: Congruence of
with molecular, behavioral,
ical data sets
Morales b,c,d, Don J. Melnick b,c,d
www.elsevier.com/locate/ympev
ion xxx (2007) xxx–xxxf the Hylobates phylogeny: Congruence of ..., Mol. Phylogenet.
Page 2
ene
ARTICLE IN PRESSThe Kloss’s gibbon (Hylobates klossii) was first
described as a ‘‘dwarf siamang’’ due to its invariant black
pelage and small size (relative to the siamang), and was
placed in the genus Symphalangus (Miller, 1903) and later
in the subgenus Symphalangus within the genus Hylobates
(Groves, 1968). Later observations and morphological
studies led to the conclusion that the Kloss’s gibbon is nei-
ther a dwarf nor a siamang (Groves, 1972; Tenaza and
Hamilton, 1971), but related to the so-called lar group
(Chivers, 1977; Creel and Preuschoft, 1984; Haimoff
et al., 1982; Marshall and Sugardjito, 1986). For instance,
in a multiple discriminant analysis of 90 cranial and dental
variables by Creel and Preuschoft (1984), H. klossii clusters
with the lar group, far from the siamang, the hoolock, or
the crested (Nomascus) gibbons. These later studies still
note the features of the Kloss’s gibbon identified as primi-
tive (such as reduced hair density, lack of facial markings,
higher average number of vertebrae) in earlier studies and
still place this species as basal to the rest of the group.
However, Geissmann’s (1993, 2002a) cladistic analysis of
morphological and karyological ‘‘non-communicatory’’
characters (which excludes pelage characteristics) disagrees
with this conclusion, and places the Kloss’s gibbon as a sis-
ter taxon to H. agilis at an internal node.
Vocalizations have been considered reliable taxonomic
indicators of closely related species (Gautier, 1988; Oates
et al., 2000; Snowdon et al., 1986; Zimmermann, 1990).
Geissmann (1993, 2002a) has analyzed gibbon phylogenetic
relationships using vocal traits. Based on these characters,
H. klossii is considered the sister taxon of the Javan silvery
gibbon (H. moloch) because these two species, unlike all
other gibbon taxa, do not sing duets; the males and females
have separate songs (Geissmann, 1993, 2002a). Geissmann
suggests that duet-splitting (partners singing at different
times of the day) is a derived characteristic, evolving after
song-splitting (partners singing different parts of a duet)
(Geissmann, 2002b). A more detailed study on Javan gib-
bon calling behavior has demonstrated that like Kloss’s
gibbons, all Javan gibbon males in an area chorus before
dawn, while the females chorus after dawn (Geissmann
and Nijman, 2006; Tenaza, 1976).
A number of molecular studies have included the
Kloss’s gibbon, but all produce different results (reviewed
in Takacs et al., 2005). Most recently, Takacs et al.
(2005) constructed a molecular phylogeny of all 12 recog-
nized species using the ND3–ND4 region of the mitochon-
drial genome. In this study, the relationships among
members of the genus Hylobates remain largely unresolved,
though H. klossii and H. moloch appear as sister taxa. The
lack of resolution in this and other molecular studies is
likely due to the fact that while more slowly evolving genes
such as the mitochondrial cytochrome b gene are often
appropriate for phylogenetic analysis of temporally deep
branches, gibbons radiated over a relatively short time
span, and thus a more quickly evolving locus is necessary
2 D.J. Whittaker et al. / Molecular Phylogto identify intrageneric species relationships. The hyper-
variable region I (HV-I) of the displacement loop, or D-
Please cite this article in press as: Whittaker, D.J. et al., Resolution o
Evol. (2007), doi:10.1016/j.ympev.2007.08.009loop, has a faster mutation rate than any other part of
the primate mitochondrial genome. This region is useful
for examining intraspecific relationships and evolutionary
relationships between closely related species (Avise, 2000)
and may be more appropriate than slower-evolving genes
for the phylogenetic analysis of gibbons (Chatterjee,
2001; Roos and Geissmann, 2001). This locus was used
to attempt to resolve the previously unclear relationships
between the four subgenera or genera of gibbons (Hylo-
bates, Hoolock, Nomascus, and Symphalangus) (Roos and
Geissmann, 2001), as well as relationships among popula-
tions of H. lar (Woodruff, 1993; Woodruff et al., 2005),
H. moloch (Andayani et al., 2001), and H. klossii (Whittak-
er, 2005).
Alternatively, the difficulty of resolving the gibbon phy-
logeny may suggest that gibbons may not have speciated in
a strict bifurcating or branching pattern, as assumed by
phylogenetic methods. Instead, gibbon phylogenetic rela-
tionships may in fact represent a ‘‘hard’’ polytomy, an
actual simultaneous or nearly simultaneous multiple speci-
ation event (as opposed to a ‘‘soft’’ polytomy, which is a
result of insufficient data to resolve speciation patterns
within an analysis). One scenario is that populations of a
single ancestral species were simultaneously isolated on
islands by rising sea levels (vicariance) and subsequently
differentiated into multiple species. Marshall and Sugardjito
(1986) have observed that with the exception of the sia-
mang, all gibbons are ecologically and behaviorally similar,
and because of their non-overlapping distributions have
not needed to adapt to different niches. Thus, the primary
differences between the gibbon species are characters that
aid in species identification, including coloration and
vocalizations.
The present study aims to resolve the phylogeny of the
genus Hylobates. We included samples from wild Kloss’s
gibbons throughout their range in the Mentawai Islands
of Indonesia (Whittaker, 2005), as well as samples collected
from pet H. moloch and H. agilis of known origin (Anday-
ani et al., 2001) to eliminate errors due to species misidenti-
fication. We chose to sequence the HV-I region of the
mitochondrial D-loop to increase the likelihood of picking
up a phylogenetic signal from a relatively fast evolutionary
event. We then tested the incongruence of our D-loop data
set with three other data sets (molecular, vocal, and mor-
phological) using the partition homogeneity test. Data sets
that were not significantly incongruent were combined to
produce a combined evidence phylogeny, which provided
much stronger support for the resulting phylogenetic
reconstruction.
2. Methods
2.1. D-loop sequencing
2.1.1. H. klossii samples
tics and Evolution xxx (2007) xxx–xxxFrom January to May 2001 and August to Decem-
ber 2003, one of us (DJW) collected fecal samples
f the Hylobates phylogeny: Congruence of ..., Mol. Phylogenet.
Page 3
2.2. Phylogenetic analysis
Sequences were aligned using the CLUSTAL X Multiple
Sequence Alignment Program, version 1.81 (Jeanmougin
et al., 1998). We used PAUP*4.0b10 (Swofford, 2002) to
perform maximum parsimony and maximum likelihood
phylogenetic inference analyses.
We used unweighted maximum parsimony analysis to
find the most parsimonious tree. A heuristic search was
performed with 1000 bootstrap replications. To determine
which model of evolutionary change best fit the data for
the maximum likelihood analysis, the program MODEL-
TEST 3.6 was employed (Posada and Crandall, 1998),
AF338897,
AF338906
H. moloch
pongoalsoni
NAN06, NAN07,
NAN10, NAN13,
NAN28, NAN30,
NAN33, NAN35
AF338878,
AF338879,
AF338882,
AF338885,
AF338899,
AF338900,
AF338902,
AF338904
Andayani
et al.
(2001)
B. hoolock Bunopithecus AF311725 Roos and
Geissmann
(2001)
S. syndactylus Symphalangus AF311722 Roos and
Geissmann
(2001)
N. gabriellae Nomascus AF193804 Roos and
Geissmann
(2001)
Table 2
Significance values from pairwise partition homogeneity tests
D-loop ND3–ND4 Vocalizations Morphology
D-loop — All four together: p = 0.16
ND3–ND4 p = 0.29 —
Vocalizations p = 0.06 p = 0.80 —
Morphology p = 0.14 p = 0.18 p = 0.10 —
ene
ARTICLE IN PRESSnon-invasively from unhabituated wild Kloss’s gibbons
at each of seven sites across the four Mentawai islands.
These samples were also used to test whether geo-
graphically isolated populations of H. klossii have
genetically diverged (Whittaker, 2005). Samples were
collected from 2–8 groups for each site, for a total of
32 gibbon groups sampled, and stored in either lysis
buffer or RNAlater� RNA Stabilization Solution
(Ambion, Inc.). DNA was extracted using standard
phenol–chloroform procedures (Sambrook et al.,
1989) or Qiagen Stool Kits�.
The HV-I region of the D-loop was amplified and
sequenced using the gibbon-specific primers GIBDLF3
(50-CTT CAC CCT CAG CAC CCA AAG C-30) and GIB
DLR4 (50-GGG TGA TAG GCC TGT GAT C-30)
(Andayani et al., 2001) which correspond to the human
primers L15996 (Vigilant et al., 1989) and H16498 (Kocher
et al., 1989). The loci were amplified using optimized Poly-
merase Chain Reaction (PCR) protocols (Palumbi, 1996)
in 50 lL reactions and processed on Perkin-Elmer� ther-
mocyclers with an annealing temperature of 55 �C. Because
of low concentration of DNA in each extraction, large
quantities of template DNA (up to 8 lL) were used in each
reaction. (Not all samples could be successfully amplified
due to extremely low DNA concentration, resulting in a
final set of 21 H. klossii sequences.) Bovine Serum Albumin
was added to each reaction to overcome any remaining
PCR inhibitors. PCR products were purified with Qiagen
PCR purification kits and cycle-sequenced using Perkin-
Elmer’s ABI Prism� BigDye� Terminator Cycle Sequenc-
ing Ready Reaction Kits. The ABI Prism� 377 Automated
Sequencer and ABI 3730 XL 96-well Capillary Sequencer
were used for sequencing. Consensus sequences for each
individual were generated using the ABI software package
AutoAssembler� as well as the Sequencher 3.1 program.
All sequences were deposited in GenBank (Accession
Nos. EF363486 through EF363506).
2.1.2. Other Hylobatid species
HV-I sequences from other gibbon species were
obtained from GenBank (H. agilis: 2; H. lar: 2; H. moloch
moloch: 5; H. moloch pongoalsoni: 8) (Table 1). The H.
moloch and H. agilis samples were obtained from captive
animals of known origin (Andayani et al., 2001), while
the H. lar and outgroup specimens were from carefully
identified zoo animals (Roos and Geissmann, 2001).
Sequences for H. muelleri and H. pileatus, as well as H. agi-
lis albibarbis, were not available on GenBank. We
sequenced DNA from blood samples from zoo specimens
(H. muelleri: JP92, JP93 [GenBank accession nos.
EF363507 and EF363508]; H. pileatus: JP99 [EF363509];
and a putative H. agilis albibarbis: JP90 [EF363485]) fol-
lowing the same protocols as above. One sequence from
each of the other three gibbon genera (Hoolock, Nomascus
gabriellae, and Symphalangus) were also obtained from
D.J. Whittaker et al. / Molecular PhylogGenBank and used as outgroups for the phylogenetic anal-
yses (Table 2).
Please cite this article in press as: Whittaker, D.J. et al., Resolution o
Evol. (2007), doi:10.1016/j.ympev.2007.08.009Table 1
List of sequences retrieved from GenBank
Taxon Sample ID GenBank
acquisition
number
Reference
H. agilis NAN04, NAN39 AF338876,
AF338905
Andayani
et al.
(2001)
H. lar lar2, lar3 AF311724,
AF311723
Roos and
Geissmann
(2001)
H. moloch
moloch
NAN08, NAN12,
NAN14, NAN26,
NAN41
AF338880,
AF338884,
AF338886,
Andayani
et al.
(2001)
tics and Evolution xxx (2007) xxx–xxx 3using the hierarchical likelihood ratio test (hLRT) to
choose which of 56 models best fit the data. A heuristic
f the Hylobates phylogeny: Congruence of ..., Mol. Phylogenet.
Page 4
ene
ARTICLE IN PRESSmaximum likelihood search was performed with 100 boot-
strap replications.
Finally, because the data matrix is large, and because
the sampling was uneven across taxa, we chose a single rep-
resentative for each species and performed an exhaustive
search using the maximum parsimony criterion. For H.
klossii, the most common haplotype was chosen (five indi-
viduals [PL04, SB04, SB06, SB19, and SP08] had the same
haplotype). For taxa with smaller sample sizes (H. agilis,
H. lar, H. muelleri, H. moloch), a single representative
was chosen at random. Only one sequence was available
for H. pileatus.
2.3. Congruence with other data sets
A ‘‘total evidence’’ approach, in which different kinds of
data (e.g., morphological, molecular, and behavioral; or
mitochondrial and nuclear DNA sequences) are combined
into a single analysis, can be employed to increase explan-
atory power (DeSalle and Brower, 1997; Eernisse and
Kluge, 1993). To overcome disparities among data sets
due to different mutation rates or selection pressures, the
data can be partitioned and statistically compared to look
for conflicting phylogenetic signals. The Incongruence
Length Difference test (ILD, also known as the partition
homogeneity test) examines whether the difference between
data partitions is greater than that expected by chance, and
can be particularly useful for determining whether data can
be combined into a single analysis (Cunningham, 1997a,b;
Farris et al., 1995).
We tested the D-loop sequences from this study for con-
gruence with three other data sets: (1) the more slowly
evolving, protein-coding mitochondrial ND3–ND4 region
(Takacs et al., 2005), with a total of 2274 basepairs; (2)
29 vocalization characters (Geissmann (1993, 2002a));
and (3) Geissmann’s (1993, 2002a) ‘‘non-communicatory’’
data set which includes 26 skeletal, postcranial, soft parts
anatomy, and karyological characters (Geissmann, 1993;
Groves, 1972; Marshall and Sugardjito, 1986; Prouty
et al., 1983; Stanyon et al., 1987; van Tuinen and Ledbet-
ter, 1983). All four data sets were combined into a single
data set and partitioned. We used the Partition Homogene-
ity Test in PAUP* 4.0b10 to examine whether the data sets
displayed significantly incongruent phylogenetic signals
(Farris et al., 1995; Swofford, 2002).
A single representative of each species was used in the
analysis. Takacs et al. (2005) did not include H. agilis albi-
barbis as a separate taxon, so it was excluded from the con-
gruence analysis. The morphological and vocalization
characters were ordered (Wagner parsimony) (Geissmann,
2002a), while the genetic characters were unordered (Fitch
parsimony). Uninformative characters were excluded (Lee,
2001), leaving 125 characters in the D-loop data set, 258 in
the ND3–ND4 set, 21 vocal characters, and 19 morpholog-
ical characters. The partition homogeneity test was run
4 D.J. Whittaker et al. / Molecular Phylogusing heuristic parsimony (100 replicates) with random
stepwise addition of sequences, and the Goloboff fit crite-
Please cite this article in press as: Whittaker, D.J. et al., Resolution o
Evol. (2007), doi:10.1016/j.ympev.2007.08.009rion (K = 2) was employed to reduce the effect of homo-
plasy on the tree (Goloboff, 1993). We conducted the
partition homogeneity test both as a simultaneous four-
way test and as each possible pairwise comparison.
Several authors agree that if no significant conflict is
observed among different data sets, they can be combined
into a single data set and analyzed as one (Cunningham,
1997a; DeSalle and Brower, 1997; Flynn and Nedbal,
1998; Omland, 1994; Remsen and DeSalle, 1998; Vogler
and Pearson, 1996). When no significant incongruence
was found among the four data sets, we produced a com-
bined evidence phylogenetic tree using maximum parsi-
mony analysis.
3. Results
3.1. D-loop phylogeny
The length of the amplified sequence was 487–491 bp in
Hylobates species, longer in outgroup species (Hoolock:
502, Nomascus: 507, Symphalangus: 520); the alignment
including gaps was 528 bp long, including 404 uninforma-
tive sites. Average within-species pairwise sequence diver-
gence estimates were as follows (range in parentheses): H.
agilis, 7.6% (5.3–8.9%); H. klossii, 2.5% (0.2–4.5%); H.
lar, 2.9%; and H. moloch, 3.3% (1.2–5.7%). Average
between-species sequence divergence within the genus
Hylobates was 10.9% (range 7.1–17.2%), and average inter-
generic divergence was 21.5% (range 17.0–26.3%). The
highest interspecific divergences within the genus Hylo-
bates, overlapping with intrageneric divergence, were
observed between H. pileatus and the other species (13.1–
17.2%), with the greatest distance found between H. pilea-
tus and H. agilis.
Using D-loop sequences, the genus Hylobates is mono-
phyletic with 98–100% bootstrap support in all analyses,
and species with multiple samples also show monophyly
with high support. Hylobates klossii clusters with H.
moloch and H. agilis, inside the lar group, and not basal
to the lar group. H. pileatus and H. lar appear as the basal
taxa in most analyses.
3.1.1. Maximum parsimony
The heuristic search found 167 equally parsimonious
trees. H. klossii clusters with H. moloch, with H. agilis as
the next most closely related species. The bootstrap maxi-
mum parsimony analysis (1000 replications) resulted in
an unresolved polytomy of the entire genus Hylobates (data
not shown).
3.1.2. Maximum likelihood
Using the hLRT, the best fitting nucleotide substitution
model was HKY + G (Hasegawa et al., 1985). This model
assumes that transitions are more likely than transversions,
that purine and pyrimidine transitions are equally likely,
tics and Evolution xxx (2007) xxx–xxxand that the substitution rate follows a gamma distribu-
tion. Using a heuristic search in PAUP*, two equally likely
f the Hylobates phylogeny: Congruence of ..., Mol. Phylogenet.
Page 5
trees were produced. In a strict consensus tree, H. klossii
and H. agilis are shown as sister taxa, with H. moloch as
the next most closely related (not shown), however, in a
bootstrap tree, H. klossii is part of an unresolved four-
way polytomy with H. moloch, H. agilis, and H. muelleri
(Fig. 1).
3.1.3. Exhaustive maximum parsimony search
The exhaustive maximum parsimony search using one
representative of each taxon found a tree with two clades:
on one branch, H. klossii and H. moloch are sister taxa,
with H. agilis and H. agilis albibarbis as a sister clade,
and on the other branch, H. muelleri, H. pileatus and H.
lar form a clade with H. lar as the basal taxon (tree not
shown).
3.2. Partition homogeneity test and combined evidence tree
No significant incongruence was found between the
D-loop data and the other data sets (p > 0.05 in all compar-
isons, Table 2), so we combined them in a maximum parsi-
mony analysis. The maximum parsimony tree (1000
bootstrap replications) is shown in Fig. 2. The pattern in
this analysis closely follows that of the D-loop analyses,
but with much stronger support. H. klossii and H. moloch
are sister taxa (91% bootstrap support), with H. agilis as
the next most closely related species. H. pileatus is the most
basal taxon in the Hylobates radiation, followed by H. lar.
4. Discussion
The mitochondrial D-loop appears to be an appropriate
locus for assisting in the reconstruction of the phylogeny of
the genus Hylobates. These gibbon species are closely
related and probably diverged in a short period of time,
and by using this fast-evolving locus we were able to dis-
cern a phylogenetic signal. However, on its own the D-loop
does not provide strong enough resolution, as evident in
the extremely low bootstrap support values in the maxi-
mum parsimony and maximum likelihood analyses. Com-
bining the D-loop sequences with other types of
phylogenetically informative data resulted in a much more
strongly supported tree. Importantly, the signal from the
D-loop did not significantly conflict with information from
other sources. The data presented here suggest that the
Kloss’s gibbon clusters with Javan silvery gibbon and the
agile gibbon within the Hylobates genus, with the pileated
gibbon in the most basal position of this radiation. The
Fig. 2. Combined evidence tree produced by maximum parsimony, 1000
bootstrap replications. Heavy lines numbered 1–4 correspond to biogeo-
graphic breaks.
D.J. Whittaker et al. / Molecular Phylogene
ARTICLE IN PRESSFig. 1. Tree produced by maximum likelihood analysis of HV-I sequences
(100 bootstrap replications).
Please cite this article in press as: Whittaker, D.J. et al., Resolution o
Evol. (2007), doi:10.1016/j.ympev.2007.08.009tics and Evolution xxx (2007) xxx–xxx 5lar gibbon and the Bornean gibbon occupy intermediate
positions.
f the Hylobates phylogeny: Congruence of ..., Mol. Phylogenet.
End of preview.
