Content uploaded by Xianguang Guo
Author content
All content in this area was uploaded by Xianguang Guo on Nov 13, 2017
Content may be subject to copyright.
Phylogeny and biogeography of Chinese sisorid catfishes
re-examined using mitochondrial cytochrome b
and 16S rRNA gene sequences
Xianguang Guo
a,b
, Shunping He
a,*
, Yaoguang Zhang
c
a
Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan 430072, PR China
b
Graduate school of Chinese Academy of Sciences, Beijing 100039, PR China
c
School of Life Science, Southwest China Normal University, Chongqing 400715, PR China
Received 25 April 2004; revised 29 November 2004
Abstract
The family Sisoridae is one of the largest and most diverse Asiatic catfish families, most species occurring in the water systems of
the Qinhai-Tibetan Plateau and East Himalayas. To date published morphological and molecular phylogenetics hypotheses of siso-
rid catfishes are part congruent, and there are some areas of significant disagreement with respect to intergeneric relationships. We
used mitochondrial cytochrome band 16S rRNA gene sequences to clarify existing gaps in phylogenetics and to test conflicting
vicariant and dispersal biogeographical hypotheses of Chinese sisorids using dispersal–vicariance analysis and weighted ancestral
area analysis in combination with palaeogeographical data as well as molecular clock calibration. Our results suggest that: (1) Chi-
nese sisorid catfishes form a monophyletic group with two distinct clades, one represented by (Gagata (Bagarius,Glyptothorax)) and
the other by (glyptosternoids, Pseudecheneis); (2) the glyptosternoid is a monophyletic group and Glyptosternum,Glaridoglanis, and
Exostoma are three basal species having a primitive position among it; (3) a hypothesis referring to Pseudecheneis as the sister group
of the glyptosternoids, based on morphological evidence, is supported; (4) the genus Pareuchiloglanis, as presently defined, is not
monophyletic; (5) congruent with previous hypotheses, the uplift of Qinghai-Tibetan Plateau played a primary role in the speciation
and radiation of the Chinese sisorids; and (6) an evolutionary scenario combining aspects of both vicariance and dispersal theory is
necessary to explain the distribution pattern of the glyptosternoids. In addition, using a cytochrome bsubstitution rate of 0.91% per
million years and 0.23% for 16S rRNA, we tentatively date that the glyptosternoids most possibly originated in Oligocene–Miocene
boundary (19–24Myr), and radiated from Miocene to Pleistocene, along with a center of origin in the Irrawaddy–Tsangpo drainages
and several rapid speciation in a relatively short time.
2005 Elsevier Inc. All rights reserved.
Keywords: Sisoridae; Glyptosternoids; Cytochrome b; 16S rRNA; Phylogeny; Biogeography
1. Introduction
The family Sisoridae, established by Regan (1911),is
one of the largest and most diverse Asiatic families,
quite a few sisorid species inhabiting in basins around
the Qinghai-Tibetan Plateau and East Himalayas. The
main basins in which these fishes live include: Yal-
uzangbujiang (Tsangpo), Irrawady, Nujiang (Salween),
Lancangjiang (Mekong River), Jingshajiang (Upper
Yangtze), Yuanjiang (Red River), Nanpanjiang (Upper
Pearl River), and Brahmaputra basin. They are highly
adaptive to torrential environment. Its special distribu-
tion and phylogenetic interpretation would be helpful
in determining the development of water systems in this
area. It is assumed that speciation events within this
1055-7903/$ - see front matter 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.ympev.2004.12.015
*
Corresponding author. Fax: +86 27 68 78 0123.
E-mail address: clad@ihb.ac.cn (S. He).
Molecular Phylogenetics and Evolution 35 (2005) 344–362
MOLECULAR
PHYLOGENETICS
AND
EVOLUTION
www.elsevier.com/locate/ympev
group are linked to historical changes in the geography
of their main distribution habitat, which have been se-
verely affected by several uplift events along the Qin-
hai-Tibetan Plateau. These geological processes have
been considered to play a fundamental vicariant role
in species of many other vertebrates endemic to this re-
gion (Luo et al., 2004; Pang et al., 2003; Ru
¨ber et al.,
2004). Thus, we wanted to test whether the uplift of
the Qinghai-Tibetan Plateau also facilitated speciation
and adaptation process of the sisorids.
The currently accepted taxonomy of the sisorid catf-
ishes is outlined in Table 1. The revision by de Pinna
(1996), based on morphology, divided the family Sisori-
dae into two subfamilies: Sisorinae and Glyptosterninae.
The Sisoridae (sensu stricto) is composed of two major
lineages, one represented by (Bagarius (Sisor (Nangra,
Gagata))) and the other by (Glyptothorax (Pseudeche-
neis, ‘‘glyptosternoids’’)). There are 12 genera of sisorids
found in China, including genera Glyptothorax,Gagata,
Bagarius,andPseudecheneis as well as eight genera of
glyptosternoids, i.e., Glyptosternum,Exostoma,Pseud-
exostoma,Oreoglanis,Pareuchiloglanis,Euchiloglanis,
Parachiloglanis, and Glaridoglanis. The other genus
Myersglanis is only found in India. And most of the
eight genera glyptosternoids are highly specialized, with
strongly depressed heads and bodies, and greatly en-
larged pectoral and pelvic fins modified to form an adhe-
sive apparatus (e.g., Fig. 1).
Despite considerable studies, the phylogenetic and
taxonomic relationships within the sisorids, and the dee-
per relationships of the glyptosternoids, remain contro-
versial (various hypotheses are illustrated in Fig. 2). In
particular, considering disagreement surrounds the phy-
logenetic affinities of the genus Pseudecheneis. Early
studies demonstrated similarities and differences between
the genus Pseudecheneis and various ‘‘glyptosternoid
sisorids’’ but were not conclusive about their possible
phylogenetic affinities (Hora, 1952; Hora and Silas,
1952a,b). Hora (1952) implied that the glyptosternoids,
in addition to the genus Pseudecheneis, formed a mono-
phyletic group. Chu (1982) focused exclusively on the
monophyletic origin of Pseudecheneis, but without
addressing which group is most closely related. Tilak
(1976) proposed a close relationship between Pseudeche-
neis and the glyptosternoids, mainly on the basis of an
overall similarity in their pectoral spine morphology as
well as a study of the adhesive thoracic apparatus. In
support of TilakÕs view, de Pinna (1996) suggested that
Pseudecheneis is the sister group of the glyptosternoids
on the basis of a large set of synapomorphies. But his
work did not cover the intrarelationships of the glyptos-
ternoids. HeÕs work (1996) showed us a ‘‘full known’’ of
intrarelationship of the glyptosternoids based on 60 oste-
ological characters. Peng et al. (2004) first verified the
monophyly of glyptosternoids and proposed different
phylogenetic relationships based on the analysis of mito-
chondrial DNA cytochrome bgene sequences. However,
Peng et al. (2004) did not resolve the placement of genus
Table 1
Current classification of the sisorids, after de Pinna (1996)
Family Subfamily Tribe Genera Current valid species
a
Sisoridae Sisorinae Sisorini Sisor 1
Gagata 10
Nangra 5
Glyptosterninae Bagariini Bagarius 4
Glyptothoracini Glyptothorax 52
Glyptosternini Pseudecheneis 6
Glyptosternum 4
Exostoma 2
Pseudexostoma 1
Oreoglanis 9
Pareuchiloglanis 15
Euchiloglanis 2
Parachiloglanis 1
Glaridoglanis 1
Myersglanis 2
Ng (2004a, 2004b);Chu et al. (1999).
a
Partial data come from: http://clade.acnatsci.org/allcatfish/ACSI/taxa/Genera_by_Family/Genera_Sisoridae.html.
Fig. 1. Lateral and ventral views of Pseudexostoma yunnanensis
branchysoma.
X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362 345
Pseudecheneis. Strikingly, Guo et al. (2004) did not agree
with that of Peng et al. (2004) as for the monophyly of
glyptosternoids and the placement of Pseudecheneis
based on the analysis of mitochondrial 16S rRNA se-
quences. Such a perspective is therefore still far from
clear promoting us to pursue further studies of the
molecular phylogenetics of the sisorid catfishes.
It has long been recognized that paleo-drainages of
major continental East Asian Rivers, draining the
south-eastern Tibet plateau margin, differed markedly
from their current drainage patterns (Brookfield, 1998;
Clark et al., 2004; Gregory, 1925; Gregory and Gregory,
1923; Hallet and Molnar, 2001; Me
´tivier et al., 1999;
Seeber and Gornitz, 1983; Zeitler et al., 2001). In a re-
cent study, Clark et al. (2004) suggested that these rivers
were once tributaries to a single southward flowing sys-
tem, which drained into the South China Sea (Fig. 3A).
Subsequent reorganization into modern major river
drainages was primarily caused by river capture and
reversal events associated with the initiation of Miocene
uplifts in eastern Tibet (Clark et al., 2004). Although
large-magnitude tectonic shear, prompted by the In-
dian–Asian collision around the eastern Himalayan syn-
taxis (especially in the ‘‘Three River’’ area where the
Salween, Mekong, and Yangtze rivers run parallel, see
Fig. 3), river capture and reversal events cannot be ruled
out as an additional factor influencing these large-scale
changes in drainage patterns (Clark et al., 2004; Hallet
and Molnar, 2001). As reviewed by Ru
¨ber et al.
(2004), the evolution of drainage systems in Asia can
be summarized in four stages (Fig. 3,Clark et al.,
2004). (a) Upper Yangtze, Middle Yangtze, Upper Me-
kong, and Upper Salween rivers drained into the South
China Sea through the paleo Red River (Fig. 3A). (b)
Capture/reversal of the Middle Yangtze river redirected
drainage away from the Red River and into the East
China Sea through the Lower Yangtze river (Fig. 3B).
(c) Capture of the Upper Yangtze River into the Lower
Yangtze River, and of the Upper Mekong and Upper
Salween rivers into their modern drainage position.
The Tsangpo River was also captured to the south
through the Irrawaddy River (Fig. 3C). (d) Capture of
the Tsangpo river through the Brahmaputra river into
its modern drainage position (Fig. 3D).
Only a few studies have recognized the potential
importance of changes in drainage basin morphology
in understanding biogeographic patterns of the South
East Asian Ichthyofauna (e.g., Kottelat, 1989; Ru
¨ber
et al., 2004). Hora (1952) hypothesized that the glyptost-
enoids originated in Yunnan of China and radiated due
Fig. 2. Various recent hypotheses regarding the phylogenetic relationships of sisorid catfishes, with particular emphasis on the position of
Pseudecheneis. Trees have been redrawn and, in some cases, taxa that are not included in the current study have been removed for the sake of clarity.
346 X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362
to tectonic uplifting and movements in the region. Hora
and Silas (1952a) also hypothesized that the glyptoster-
noids achieved their current distribution by means of
dispersal, based on the geological hypothesis proposed
by Gregory and Gregory (1923) and Gregory (1925),
in which it is considered that, due to regional subsi-
dence, some rivers of this area changed their direction
by river-capture. They also considered that the orogenic
movements of the Himalayan mountain range formed a
highway, which supported the dispersal of the glyptister-
noids. However, recent historical biogeography study
(He, 1995), based on morphology, provided an alterna-
tive explanation for the distribution pattern of the glyp-
tosternoids, and hypothesized that the speciation of this
group has a direct relationship with three uplifting
events of the Qinhai-Tibetan Plateau. In addition, to
our knowledge, thus few molecular phylogenetic studies
have been conducted to test underlying vicariant/dis-
persed speciation hypotheses.
16S rRNA gene evolves at a slower rate than mito-
chondrial DNA cytochrome bgene, so the two genes
were selected to recover the maximum phylogenetic
information for the terminal nodes at the base of the
tree. Drawing on the combined molecular data, the
goals of this study were: (1) to elucidate existing gaps
in the phylogenetic relationships among the Chinese
sisorids; (2) to use molecular calibrations to investigate
if the divergence events within the sisorids are correlated
with the uplift events of the Qinhai-Tibetan Plateau; (3)
to resolve the biogeographic (dispersal/vicariance) con-
troversy surrounding the Qinhai-Tibetan Plateau distri-
bution of the glyptosternoids. To achieve the final aim,
we attempted to reconstruct the ancestral distributions
in the phylogeny of the glyptosternoids.
Fig. 3. Map of South-East Asia showing major changes in drainage basin morphology (Ru
¨ber et al., 2004; and references therein). Map (A) shows
the drainage pattern prior to the major captures, where the Upper Yangtze, Middle Yangtze, Upper Mekong, Upper Salween, and the Tsangpo rivers
drained together to the South China Sea through the paleo-Red River (paleo-Red River drainage shown in bold black). (B) Capture/reversal of the
Middle Yangtze into the Lower Yangtze (changes shown in bold black). (C) Capture of the Upper Yangtze River by the Middle Yangtze, and of the
Upper Mekong and Upper Salween rivers into their modern drainage positions. The Tsangpo River was captured by the Irrawaddy River (changes
shown in bold black). (D) Capture of the Tsangpo River through the Brahmaputra River into its modern course (changes shown in bold black). The
asterisk indicates the ‘‘Three River’’ area mentioned in the text. The indicated rivers are: B, Brahmaputra; I, Irrawaddy; M, Mekong; R, Red River;
S, Salween; T, Tsangpo; and Y, Yangtze.
X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362 347
2. Materials and methods
2.1. Sample collection
The specimens used in this study, including 32 indi-
viduals from 17 sisorid catfish species and five non-sisor-
ids, following the system of Chu et al. (1999), were
collected from a variety of locations in China (Table 2,
Fig. 4). Due to difficulties of sampling, Myserglanis
and Parachiloglanis were not included in this study.
Muscle tissue used was preserved in 95% ethanol, and
most specimens were deposited in the Fish Collection
of the Institute of Hydrobiology of the Chinese Acad-
emy of Sciences. As outgroups, Akysis brachybarbutus
and Liobagrus anguillicauda were included, which be-
long to Amblycipidae and Akysidae, respectively, puta-
tive close relatives to Sisoridae according to recent study
(de Pinna, 1996).
2.2. DNA extraction, amplification, and sequencing
Total genomic DNA was extracted from small
amounts of ethanol-preserved muscle using standard
proteinase K, phenol/chloroform extraction (Sambrook
et al., 1989). The complete cytochrome bgene was
amplified with primers L14724 and H15915, adapted
from Xiao et al. (2001). Then 580 bp of the mitochon-
drial 16S rRNA gene was amplied by using the universal
primers 16Sar and 16Sbr (Palumbi, 1996). The PCRs
contained approximately 100 ng of template DNA,
1lL of each primer, 5 lLof10·reaction buffer, 2 lL
dNTPs (each 2.5 mM), and 2.0 U Taq DNA polymerase
in total 50 lL volume. The PCR profile consisted of an
initial denaturation step (3 min at 94 C), followed by 35
cycles performed in the following order of denaturation
at 94 C for 1 min; annealing at 58 C for 45 s (52 C for
16S rRNA); and elongation at 72 C for 1 min; and a fi-
nal extension at 72 C for 8 min. PCR amplification
products were purified on a 0.8% low melting point aga-
rose gel stained with ethidium bromide, using the Bio-
Star glassmilk DNA purification kit following the
manufacturerÕs protocol. Nucleotide sequences of cyto-
chrome bgene and 16S rRNA gene were determined
using the purified PCR product. Sequencing reactions
were performed in both directions to allow verification
of character states. The sequences have been deposited
in GenBank (Accession Nos. are listed in Table 2).
2.3. Sequence alignment and analyses
No alignment was necessary for cytochrome bbe-
cause it is a protein-encoding gene. Initial alignments
of the 16S rRNA sequences were performed with Clustal
X(Thompson et al., 1997) with default gap costs.
Alignments were then refined by eye on the basis of
16S rRNA secondary structure (De Riijk et al., 1999).
Table 2
Details of specimens and sequences used in this study
Taxon and samples NCollecting locales GenBank Accession Nos.
a
16S rRNA Cytochrome b
Glyptothorax cavia 1 Yingjiang, Yunnan AY445906 AF477830
1
Glyptothorax fukiensis fukiensis 2 Wuyishan, Fujian AY574359 AF416884
1
Glyptothorax fukiensis hainanensis 2 — AF416887
1
Glyptothorax sinense 1 Hejiang, Sichuan AY574357 AY601764
Bagarius yarrelli 2 Tengchong, Yunnan AY445910 AF416897
1
Gagata cenia 1 Daojie, Yunnan AY445905 AF499599
1
Pseudecheneis sulcatus 2 Motuo, Tibet AY574355 AY601765
Glyptosternum maculatum 1 Lhasa, Lhasa, Tibet AY445908 AY601766
Euchiloglanis davidi 2 Baoxing, Sichuan AY445895 AF416883
1
Euchiloglanis kishinouyei 1 Yajiang, Sichuan AY445898 AY207480
1
Pareuchiloglanis sinensis 1 Mabian, Sichuan AY445901 AY191609
1
Pareuchiloglanis anteanalis 1 Yanjin, Yunnan AY445903 AY191610
1
Pareuchiloglanis kamengensis 1 1 Chayu, Tibet AY574360 AY601767
Pareuchiloglanis kamengensis 2 1 Chayu, Tibet AY574360 AY601768
Glaridoglanis andersonii 2 Chayu, Tibet AY574354 AY601769
Pseudexostoma yunnanensis 2 Tengchong, Yunnan AY445894 AF499602
1
Oreoglanis delacouri 2 — AF416878
1
Exostoma labiatum 3 Tengchong, Yunnan AY445907 AF499598
1
Outgroup
Liobagrus anguillicauda 2 Wuyishan, Fujian AY574353 AF416888
1
Akysis brachybarbutus 2 Jinghong, Yunnan AY574352 AF499603
1
Pseudobagrus tokiensis AB054127
2
AB054127
2
Pseudobagrus kyphus AB085621
3
AB085622
3
Mystus sp. AY458877
4
AY458893
4
a
Sources of additional sequences:
1
Peng et al. (2004);
2
Saitoh et al. (2003);
3
Watanabe et al. (2002);
4
Wilcox et al. (2004).
348 X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362
Most of cytochrome bgene sequences used in this study
were downloaded from GenBank (see Table 2). Despite
the fact that some sequences originated from different
specimens, there was little reason to assume inconsisten-
cies with regard to identification. Most of the samples
used in this study and those of Peng et al. (2004) had
been identified following the criteria in Chu et al. (1999).
To ensure accuracy, several representatives from each
species were included whenever possible. However, the
DNA of some samples used in this study was of poor
quality and did not amplify readily during PCR. For
this reason, some specimens included in the cytochrome
bsequences are not represented in the 16S rRNA se-
quences. Alignment gaps in 16S rRNA sequences were
treated as missing characters.
Phylogenetic congruence of cytochrome band 16S
rRNA data sets were tested by the partition homogene-
ity test of Farris et al. (1995) with PAUP* 4.0b10 (Swof-
ford, 2002). The partition homogeneity test supported
the combination of the cytochrome band 16S rRNA
data sets (P= 0.518). As mentioned above, some sam-
ples were of poor quality and did not amplify, resulting
in the inclusion of missing data. However, Wiens (1998)
suggested that unless the proportion of missing data is
large, addition of incomplete data sets is more likely
to improve phylogenetic accuracy than reduce them.
Consequently, the analyses were performed using the
combined data set, which included eighteen taxa of Siso-
ridae and five outgroups for two genes: cytochrome b
and 16S rRNA (Table 2).
Base compositional frequencies and nucleotide sub-
stitutions between pairwise distances were determined
using PAUP*. Also using PAUP*, random trees
(n= 1000) were generated to examine the phylogenetic
Fig. 4. Map showing the sampling localities of the sisorids used in this study.
X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362 349
signal (Hillis and Huelsenbeck, 1992). The possibility of
sequence saturation was investigated by plotting uncor-
rected sequence divergence versus pairwise numbers of
observed changes (for first, second, and third codon
positions, and for transitions and transversions sepa-
rately). The amount of sequence saturation is inferred
from the shape of the trend line, with a linear relation-
ship indicating that the sequence is unsaturated and an
asymptotic relationship indicating the presence of
saturation.
We performed a wide array of phylogenetic analyses
using different methods to gauge the robustness of our
resulting hypotheses. These methods were maximum
parsimony (MP), neighbor-joining with maximum like-
lihood distance (NJ), maximum likelihood (ML) as
implemented in PAUP*, and a Bayesian approach as
implemented in MrBayes ver. 3.0b (Huelsenbeck and
Ronquist, 2001). The TI bias was estimated with ML
and an associated transversion (TV) weighting sheme
(TV:TI = 3:1) was implemented for MP, using a stepm-
atrix. Character weights of 2:1 and 4:1 were applied to
first, and second-codon positions relative to third posi-
tions, respectively, for cytochrome b. The MP method
was performed using heuristic searches with 20 ran-
dom-addition-sequence replicates and tree-bisection-re-
connection (TBR) branch swapping. Likelihood ratio
tests (Goldman, 1993a,b; Huelsenbeck and Crandall,
1997), as implemented in MODELTEST 3.06 (Posada
and Crandall, 1998), were employed to choose models
for model-based methods (NJ, ML, and Bayesian anal-
yses). The GTR + G + I model (Gu et al., 1995; Yang,
1994) was selected by MODELTEST. Four indepen-
dent MCMC chains were simultaneously run for
1,000,000 replicates by sampling one tree per 100 repli-
cates with the Bayesian procedure. We discarded the
first 800 trees as part of a burn-in procedure, and used
the remaining 9200 sampling trees (whose log-likeli-
hoods converged to stable values) to construct a 50%
majority rule consensus tree. In addition to Bayesian
posterior probabilities, node supports were assessed
using ML, MP, and NJ bootstraps (Felsenstein, 1985)
with 100, 1000, and 1000 replicates, respectively. Alter-
native phylogenetic hypotheses were tested using Tem-
pleton test (Templeton, 1983) and Shimodaira–
Hasegawa test (SH; Shimodaira and Hasegawa, 1999)
using 1000 bootstrap replicates with RELL optimiza-
tion as implemented in PAUP*.
To estimate divergence times, we first performed tests
for substitution rate constancy (molecular clock test) of
cytochrme b, 16S rRNA, and combined data in different
lineages, respectively. The test compares the log-likeli-
hood of the most likely tree with and without a molecu-
lar clock enforced. Branch lengths were also calculated
under the assumption of clocklike evolution for the esti-
mation of lineage divergence times, by using the Kimura
(1980) two-parameter model of sequence evolution. To
date, the only fossil of sisorid catfishes is that of the
giant sisorid, Bagarius bagarius, from Pliocene deposit
in India and Sumatra (Hora, 1939), but no direct ances-
tor has yet been detected in Chinese fossil layers. Thus,
we chose to attempt clock calibrations using published
rates of evolution for mitochondrial genes to roughly
estimate divergence times between major clades. For
cytochrome b, we used a substitution rate of 0.91%/
Myr, which was derived for the same gene for schizotho-
racine fishes based upon well-dated geological events
(He et al., 2004). We adopted a substitution rate of
0.23%/Myr for 16S rRNA (Alves-Gomes, 1999), and
with 0.5–0.9%/Myr for combined data (Martin and Pa-
lumbi, 1993).
2.4. Primary BPA, DIVA, and WAAA analyses
Glyptosternoids historical biogeography was investi-
gated in three ways. First, we performed primary BPA
using the method of Brooks (Brooks et al., 2001) based
on the best phylogenetic tree and the distribution pattern
(Table 3). Pseudecheneis sulcatus was used as outgroup.
Owing to no general area cladogram (GAC) available,
dispersal–vicariance analysis was performed using DIVA
Table 3A
Known distribution of Pseudecheneis sulcatus and glyptosternoid fishes in this study
Taxon Brahmaputra Tsangpo Irrawaddy Salween Mekong Upper Yangtze Ganges
1Pseudecheneis sulcatus %%%%% %
2Glyptosternum maculatum qq
3Glaridoglanis andersonii
4Exostoma labiatum •••
5Pseudeoxtoma yunnanensis ¤¤
6Paruechiloglanis kamengensis nnnnn
7Oreoglanis delacouri ...
8Pareuchiloglanis sinensis n
9Pareuchiloglanis anteanalis h
10 Euchiloglanis davidi }
11 Euchiloglanis kishinouyei m
Data from Chu et al. (1999).
350 X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362
1.1 (Ronquist, 1996) to determine possible ancestral
areas for internal nodes, since this program can recon-
struct ancestral distributions without taking general bio-
geographic pattern into account. DIVA reconstructs
ancestral distributions by optimizing areas on a given
cladogram based on the vicariance model, while at the
same time allowing dispersals, duplication and extinc-
tions to occur (Ronquist, 1997). Duplications and vicar-
iance receive a cost of 0 in DIVA, whereas dispersals and
extinctions cost 1 per area added or deleted, respectively.
This is because ‘‘dispersals and extinctions are unpredict-
able events that can wipe out the traces of phylogeneti-
cally constrained processes like vicariance and
duplication’’ (Sanmartı
´n, 2003), and unless extinctions
and dispersals are assigned a cost spurious events may
be introduced in optimal reconstructions (Sanmartı
´n
and Ronquist, 2002). To determine more reliable esti-
mates of the distribution of the root node, we included
an additional outgroup Pseudecheneis sulcatus in the
analysis such that the node was no longer the root node.
Species were assigned to seven main distribution areas:
Brahmaputra, Tsangpo, Irrawaddy, Salween, Mekong,
Upper Yangtze, and Ganges (Table 3; see also Fig. 7)
and the combined data best tree was evaluated (DIVA
settings: maxareas = 7, bound = 250, hold = 32767,
weight = 1.000, age = 1.000). Weighted ancestral area
analysis (WAAA; Hausdorf, 1998), a cladistic method
for estimating ancestral areas using reversible parsi-
mony, was also performed on the best phylogenetic tree.
WAAA produces ‘‘probability indices’’ indicating the
relative probability that a particular area was part of
the ancestral area for a particular node on a tree (the
higher the number, the more likely the area was part of
the ancestral area).
3. Results
3.1. Characteristics of individual genes
3.1.1. Cytochrome b
For eight individuals, we sequenced the complete
cytochrome bgene and identified six haplotypes (see Ta-
ble 2). The other 17 complete cytochrome bsequences
(with the exception of Pseudobagrus kyphus, 1132 bp)
were obtained from GenBank. A total of 1138 positions
were analyzed, of which 564 characters were variable,
and 458 of these characters were phylogenetically infor-
mative (49.5 and 40.2%, respectively). Levels of se-
quence divergence (uncorrected distance) between the
outgroup and ingroup lineages ranged from 16.1% (be-
tween Akysis brachybarbatus and Glytosternum macula-
tum) to 25.9% (between Mystus sp. and Oreoglanis
delacouri). Sequence divergence within the ingroup taxa
was as high as 25.6% (Exostoma delacouri compared to
Gagata cenia). The smallest divergence between two spe-
cies was from Pareuchiloglanis sinensis to Pareuchilogl-
anis anteanalis (0). The saturation plots of uncorrected
sequence divergence against transitions and transver-
sions divided by codon position indicated saturation at
third position transitions (not shown). The g1 statistic
indicated that a significant phylogenetic signal was pres-
ent: g1=0.63; P= 0.001; means ± SD tree length =
4353.00 ± 101.51.
3.1.2. 16S rRNA
For 28 individuals, we sequenced the 16S rRNA frag-
ments (about 580 bp) and identified 17 haplotypes (see
Table 2). The other three 16S rRNA sequences were ob-
tained from GenBank. Twenty sequences of 567 bp
(excluding primers) of the 16S rRNA gene were ana-
lyzed. A total of 241 characters were variable, and 133
of these were phylogenetically informative (42.5 and
23.5%, respectively). Levels of sequence divergence
(uncorrected distance) between the outgroup and in-
group lineages ranged from 6.8% (between Liobagrus
anguillicauda and Glyptothorax cavia) to 33.0% (between
Pseudobagrus tokiensis and Pseudexostoma yunnanen-
sis). Sequence divergence within the ingroup was as high
as 13.9% (between Bagarius yarrelli and Euchiloglanis
kishinouyei). The smallest divergence between two taxa
was 0 (e.g., from Glyptothorax fukiensis fukiensis to Gly-
ptothorax sinense). When outgroup taxa were excluded,
scatterplots of uncorrected sequence divergence versus
transition and transversion suggested the transitions
and transversions were not saturated (not shown). The
g1 statistic indicated that significant phylogenetic signal
was present: g1=0.77; P= 0.001; mean ± SD tree
length = 838.50 ± 30.65.
3.2. Phylogenetic relationships
Fig. 5A showed the Bayesian tree constructed from a
set of 23 combined sequences, which confirmed the
monophyly of the sisorids and glyptosternoids
(PP = 97 and 95%, respectively). The other methods
yielded similar topologies (see Fig. 5B–D). Two major
clades within the sisorids were identified. Clade A con-
tained the tribe Glyptosternini, i.e., Pseudecheneis and
the glyptosternoids. Clade B contained the non-glyptos-
ternini sisorids in China. Clades A and B appeared to be
Table 3B
Basins in which species 1-11 occur coded for primary BPA in this study
Basins Species Binary code
Ganges River 1, 100000000000000000001
Brahmaputra basin 1, 2, 4, 6 110101000000000111111
Tsangpo River 1, 2, 3, 4, 6 111101000000000111111
Irrawaddy River 1, 3, 4, 5, 6, 7 101111100000001111111
Salween River 1, 5, 6, 7 100011100000001111111
Mekong River 1, 6, 7 100001100000001111111
Upper Yangtze 8, 9, 10, 11 000000011111111011111
X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362 351
Fig. 5. Reconstructed phylogeny of the Chinese sisorids using Bayesian, ML, and MP, NJ phylogenetic approach based on combined cytochrome b
and 16S rRNA sequences. Numbers represent node supports inferred from Bayesian posterior probability, ML bootstrap, MP bootstrap, and NJ
bootstrap analyses, respectively (only values above 50 are shown). (A) Bayesian tree using GTR + G + I model. (B) ML tree using GTR + G + I
model; a= 0.9456, I = 0.4081; ln L= 13112.03619. Bootstrap estimates are derived from 100 replicates. (C) MP tree, Tree Length = 5559,
CI = 0.5078, HI = 0.4922. Bootstrap estimates are derived from 1000 replicates. (D) NJ tree using GTR + G + I model. Bootstrap estimates are
derived from 1000 replicates.
352 X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362
sister-lineages (PP = 97%), despite the weak bootstrap
support in ML, MP, and NJ analyses. Phylogenetic res-
olution among the tribe Glyptosternini was generally
limited to recent relationships; interior nodes received
poor support (Fig. 5). The monophyly of clade A was
retained with moderate posterior probability
(PP = 59%) and weak bootstrap support (with 54, 60,
<50% in ML, NJ, and MP tree, respectively). Within
clade A, a sister-group relationship between Pseudeche-
neis and the glyptosternoids was supported by the
Bayesian, ML, and NJ topology (59, 54, and 60%,
respectively) but not MP. However, the monophyly of
Pareuchiloglanis was disputed here, with P. kamengensis
sister to Pseudexostoma (sister to Psudexostoma plus
Fig. 5. (continued)
X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362 353
Oreoglanis in NJ tree). When we forced all the samples
of Pareuchiloglanis (P. sinensis,P. anteanalis, and P.
kamengensis) to form a monophyletic group, the tree ob-
tained (ln L=13239.68551) differs significantly
(P< 0.001) from the ML tree (ln L=13100.93144)
based on a SH test (Shimodaira and Hasegawa, 1999).
Consequently, we reject the monophyly of Pareuchilogl-
anis. The monophyly of clade B was highly supported by
posterior probability (PP = 100%) and ML bootstrap
analysis (BP = 90%), but received mediocre bootstrap
support from the MP and NJ analyses (58 and 51%,
respectively). Within clade B, Gagata was the basal spe-
cies and a sister-group between Bagarius and Glyptotho-
rax was supported by the Bayesian and ML tree (67 and
76%, respectively), but not MP and NJ tree. When using
only closely related outgroup taxa, such as Akysis
brachybarbatus or Liobagrus anguillicauda, bootstrap
support for the monophyly of clade B did not increase
significantly (in the MP tree, support for this grouping
from 58 to 52%, whereas in the NJ tree it increased from
51 to 61%). This finding is in agreement with the results
of the saturation test described above. Within clade B,
Gagata was the basal species and a sister-group between
Bagarius and Glyptothorax was retained, but with weak
support from posterior probability (67%) and bootstrap
analysis (76% in ML, lower than 50% in NJ and MP
tree).
Templeton test and Shimodaira–Hasegawa test were
carried out to determine the best topology recovered un-
der the different methods (see Table 5). The two tests did
not reject the monophyly of sisorids. Bayesian tree was
the best tree based on Shimodaira–Hasegawa test. How-
ever, the hypothesis of Guo et al. (2004) was considered
as the best topology based on Templeton test. Given
that the SH test has two advantages over the more com-
monly used Kishino-Hasegawa (KH) and Templeton
tests (Buckley et al., 2001), we considered the Bayesian
tree as the best topology in this study.
Table 4
Results of the hierachical LRT test for constancy of rates of evolution for the data sets
Lineage Molecular data ln nonclock ln clock df P
Sisorids Cytochrome b7452.80274 9509.91751 15 0.00
*
Cytochrome b(excluding Oreoglanis) 6889.80496 6920.22976 14 0.00
*
16S rRNA 1835.28265 1851.56392 13 0.00
*
Combined 9449.07524 9518.78406 15 0.00
*
Combined (excluding Oreoglanis) 8833.57756 8871.71233 14 0.00
*
Glyptosternoids Cytochrome b4713.93648 4733.72699 8 0.00
*
Cytochrome b(excluding Oreoglanis) 4133.67947 4134.45309 7 0.99
16S rRNA (without Oreoglanis) 1326.80857 1331.6024 7 0.21
Combined data 6165.27371 6188.89477 8 0.00
*
Combined data (excluding Oreoglanis) 5534.05699 5536.58742 7 0.65
Glyptosternoids +Pseudecheneis Combined data (excluding Oreoglanis) 6107.98698 6120.53772 7 0.00
*
16S rRNA (without Oreoglanis) 1427.88269 1433.01452 7 0.17
Cytochrome b(excluding Oreoglanis) 3652.36010 4133.38204 7 0.00
*
Bagarius +Gagata +Glyptothorax Cytochrome b3695.04247 3696.65956 4 0.52
16S rRNA 1108.19229 1110.13442 3 0.27
Combined data 4858.77984 4861.77897 4 0.20
Associated probabilities are given and significantly (P< 0.05) different topologies are indicated by asterisks.
Table 5
Statistical comparison of alternative topologies including MP, NJ, ML, Bayesian trees, and various evolutionary hypotheses using Templeton (1983)
and Shimodaira and Hasegawa (1999) tests
Topology Templeton test Shimodaira–Hasegawa test
Maximum parsimony Maximum likelihood
Tree length NZ P ln Lln LDiff. P
Bayesian 2623 73 0.5852 0.5584 14496.39260 Best
ML 2631 103 1.2466 0.2125 14532.54008 36.14748 0.388
NJ 2645 119 2.3643 0.0181
*
14510.50011 14.10751 0.762
MP 2620 51 0.2722 0.7855 14511.70220 15.30960 0.744
He (1996)
a
2744 219 7.5703 <0.0001
*
14924.87937 428.48677 0.000
*
de Pinna (1996)
a
2633 103 1.4780 0.1394 14543.29542 46.90282 0.261
Peng et al. (2004)
a
2721 141 7.7283 <0.0001
*
14760.38364 263.99104 0.000
*
Guo et al. (2004)
a
2618 Best 14505.58844 9.19584 0.792
Associated probabilities are given and significantly (P< 0.05) worse topologies are indicated by asterisks.
a
Maximum likelihood tree recovered from constraint search.
354 X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362
3.3. Divergence time estimations
The results of molecular tests are illustrated in Table
4. A log-likelihood ratio test rejected the null hypothe-
sis of rate constancy when including all glyptosternoids
cytochrome bsequences (v2
½8¼39:58102, P< 0.001).
Similar test rejected clocklike evolution for combined
sequences (v2
½8¼47:24212, P< 0.001). However, simi-
lar test could not reject clocklike evolution for 16S
rRNA sequences (without Oreoglanis delacouri se-
quence; v2
½7¼9:58766, P= 0.21). More specifically,
rate constancy was statistically rejected for branch at
one node for cytochrome band one node for the glyp-
tosternoids combined data. For Bagarius + Gag-
ata + Glytothorax, clocklike evolution for all data
could not be rejected. For the glyptosternoids (exclud-
ing Oreoglanis) plus Pseudecheneis, clocklike evolution
for 16S rRNA could not be rejected (v2
½7¼10:26366,
P= 0.17). Molecular-clock calibrations were restricted
to clades for which clocklike evolution was supported.
Also molecular clock calibrations were used to estimate
the times of well-supported phylogenetic divergences
(see Table 6). Molecular-clock estimates for the diver-
gence of Pseudecheneis and glyptosternoids was 21.67
Mya (mean rate = 9.97% per site per million years,
SD = 1.20% for 16S rRNA estimates), falling within
the time frame of the second uplift of the Qinghai-Ti-
betan Plateau (25-17 MaBP; Shi et al., 1999). As can
be seen from the table, estimates for the divergence
of Bagarius and Glyptothorax was 14.63 ± 2.46 Mya
for 16S rRNA, probably falling within the boundary
of second uplift and second peneplain stage. The diver-
gence of Euchiloglanis and Pareuchiloglanis in Upper
Yangtze was calculated to be 2.96–5.32 Mya, falling
within the time frame of the second peneplain—the
third uplift boundary (3.4-1.7 MaBP; Shi et al.,
1999). Thus, the glyptosternoids most possibly origi-
nated in Oligocene–Miocene boundary, and radiated
from Miocene to Pleistocene.
3.4. Primary BPA, DIVA, and WAAA analyses
The results show the hypothesis of the relationships
among the basins where glyptosternini fishes occur
(Fig. 6). Using the Bayesian tree (Fig. 5A) and the distri-
bution pattern (Table 3A) for the tribe Glyptosternini
(glyptosternoid + Pseudecheneis), a matrix of region/
taxon for tribe Glyptosternini was created using the
method of primary BPA (Table 3B). We reconstructed
a taxon-area cladogram (Fig. 6) with PAUP*. In Fig.
6, the relationships and the vicariance sequence of the
principal basins in Qinhai-Tibetan Plateau were inter-
preted as follows. From west to east the River Ganges
and Brahmaputra became isolated in turn. In the eastern
Himalayas, the first isolated basin is Jinshajiang River
(Upper Yangtze); the second is Tsangpo; the third is
Table 6
Divergence time (MY = million years) for selected phylogenetic splits within the Chinese sisorids as based on molecular clock calibrations for cytochrome bgene (He et al., 2004), fish rRNA
(Alves-Gomes, 1999), and salmonid mtDNA (Martin and Palumbi, 1993)
Phylogenetic split Cyt b(%) 0.91%/MY 16S rRNA (%) 0.23%/MY mtDNA (%) 0.5–0.9%/MY
E. davidi–E. kishinouyei 2.60 ± 0.47 1.43 ± 0.26 0.19 ± 0.17 0.41 ± 0.36 1.81 ± 0.33 1.01 ± 0.18–1.81 ± 0.33
P. sinensis–P. anteanalis 00 00
Euchiloglanis—(P.sinensis,
P.anteanalis)
7.28 ± 0.84 4.00 ± 0.46 1.41 ± 0.50 3.07 ± 1.07 5.32 ± 0.49 2.96 ± 0.27–5.32 ± 0.49
Oreoglanis—(Euchiloglanis,
P. sinensis,P. anteanalis)
——— ——
(Pseudexostoma,P. kamengensis)–
(Oreoglanis,Euchiloglanis,
P. sinensis,P. anteanalis)
13.39 ± 0.98 7.36 ± 0.54 5.96 ± 0.99 12.95 ± 2.15 10.88 ± 0.77 6.04 ± 0.43–10.88 ± 0.77
Pseudexostoma–P. kamengensis 9.35 ± 0.96 5.14 ± 0.53 0.74 ± 0.35 2.11 ± 0.76 6.42 ± 0.67 3.57 ± 0.37–6.42 ± 0.67
Glyptosternum—(Glaridoglanis,
Oreoglanis, Exostoma,
Pseudexostoma,Pareuchiloglanis,
Euchiloglanis)
19.57 ± 1.19 10.75 ± 0.65 9.82 ± 1.18 21.34 ± 2.57 16.28 ± 0.92 9.04 ± 0.51–16.28 ± 0.92
Pseudecheneis–Glyptosternoids — — 9.97 ± 1.20 21.67 ± 2.61 —
Bagarius–Glyptothorax 21.83 ± 1.34 11.99 ± 0.74 6.73 ± 1.13 14.63 ± 2.46 17.86 ± 0.10 9.92 ± 0.06–17.86 ± 0.10
Gagata—(Bagarius, Glyptothorax) 22.09 ± 1.36 12.14 ± 0.74 7.61 ± 1.08 16.54 ± 2.35 18.15 ± 0.97 10.08 ± 0.54–18.15 ± 0.97
Clade A–Clade B — — — — — —
Nodes that failed rate constancy were excluded from calibrations.
X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362 355
Mekong; and the last separated basins are Salween and
Irrawaddy.
An exact search with DIVA resulted in 40 alternative
reconstructions of the distribution history, each requir-
ing 14 dispersals between drainages. The optimal distri-
butions at each ancestral node were given in Fig. 7A.
The reconstructions are various with some of the basal
events. The suggested solution implied a center of origin
for the glyptosternoids in the Irrawaddy–Tsangpo
drainages (Fig. 7B). Inferred ancestral areas (DIVA)
and probability indices (WAAA) are shown in Table
7. To distinguish between these scenarios, it was neces-
sary to consider geological evidence and the likelihood
of widespread ancestors.
4. Discussion
4.1. Intrarelationships of the Chinese sisorids
The Sisoridae has been consistently recognized as a
natural group (de Pinna, 1996; Diogo et al., 2002) based
on morphology. Despite the low support for this node in
ML, MP, and NJ analysis, these results are congruent
with morphology, which further lends support to the
hypothesis of monophyletic sisorids. Moreover, Bayes-
ian tree indicates a robust node support (PP = 97%).
In support of Guo et al. (2004), our result indicates that
Chinese sisorid catfishes are composed of two major lin-
eages, one represented by (Gagata (Bagarius,Glyptotho-
rax)) and the other by (glyptosternoids, Pseudecheneis).
However, de Pinna (1996) suggested that Glyptothorax
does not belong the major lineage represented by Baga-
rius (see Fig. 2A). This incongruence may be caused by
the absence of sampling genera Nangra and Sisor, since
taxon sampling has been shown to have a major impact
on phylogenetic inference.
One of the most important results with respect to the
phylogenetic relationships of sisorid catfishes is the dem-
onstration of the sister-group relationship between
Pseudecheneis and the glyptosternoids. Despite the low
support for this node in Bayesian, ML, and NJ analyses,
these results are congruent with morphology (de Pinna,
1996), which further lends support to the hypothesis
Fig. 6. Area cladogram derived from the cladogram of the glyptosternoids by using Brooks parsimony analysis.
Fig. 7. (A) Summary of the forty optimal reconstructions of the ancestral distributions of the glyptosternoids using dispersal-varicance analysis.At
each node, the optimal distribution prior to vicariance is given; alternative, equally optimal distributions are separated with comma. Each
reconstruction requires 14 dispersal events. Theses dispersals, and the associated vicariance events, differ among the reconstructions and are not
indicated on the cladogram. (B) Suggested biogeographical history of the glyptosternoids (see text). The reconstruction is one of the forty
alternatives, equally parsimonious biogeographical reconstructions (cf. 7A). Dispersal events are indicated on the branches. Note that the dispersal-
varicance analysis, the cost of dispersal is one per area added to an ancestral distribution. For instance, the cost of assumption that Pareuchiloglanis
kamengensis dispersed from Irrawady to the other major regions (+ABDE) is four dispersals. Implied between-area vicariance events are indicated by
hyphens in the ancestral distributions. The illustrated reconstruction differs from the other alternatives (cf. 7A) in postulating that the ancestral
species of the glyptosternoids was distributed in the Irrawaddy–Tsangpo drainages, and the ancestral species of Oreoglanis delacouri was most
widespread. In addition, the reconstruction implies a center of origin for the glyptesternoids in the Irrawaddy–Tsangpo drainages and several rapid
speciations in a relatively short time.
c
356 X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362
X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362 357
referring to Pseudecheneis as the sister group of mono-
phyletic glyptosternoids. In addition, our result agrees
well with TilakÕs hypothesis (1976) that the Pseudecheneis
is more closely related to Glyptothorax than to Bagarius.
Hora (1952) first defined the glyptosternoids as a nat-
ural group. From our molecular analysis (Bayesian,
ML, and NJ trees), we can see that this group is basi-
cally a monophyletic group, Glyptosternum is the ances-
tral genus. Though most branches are highly supported,
the branches where Glyptosternum,Glaridoglanis, and
Exostoma diverged are not certain. However, this did
not increase the possibility that Glyptosternum and Glar-
idoglanis as well as Exostoma evolved independently, be-
cause they live under strong stress of the torrential
environment, as recordings of the distribution are
mountain brooks or the rivers around the hills. Thus,
we hypothesize that both Exostoma and Glaridoglanis
probably derive directly from a Glyptosternum-like
ancestor and have a close relationship with Glyptoster-
num. In correspondence with our current results, He
(1996) and Peng et al. (2004) also suggested that the
glyptosternoids form a monophyletic group with Glypt-
ostenum as the most primitive genus, based on cladistic
analyses of 13 apormorphies and mitochondrial cyto-
chrome bsequences, respectively. The lack of phyloge-
netic resolution at deep levels within the
glyptosternoids may be attributed to several factors,
including saturation effects at the third positon of cyto-
chrome bgene, rate heterogeneity, and rapid cladogene-
sis within a relatively short time period.
Comparable to the molecular topology of Guo et al.
(2004), our results also provide support for the poly-
phyly of the genus Pareuchiloglanis. The hypothesis of
monophyly for the genus pareuchiloglanis was statisti-
cally rejected, when we compared the ML scores be-
tween the optimal ML tree and constrained
monophyletic tree using the Shimodaira-Hasegawa
parametric test as implemented in PAUP* (p< 0.001).
Supports for some of the basal nodes within the glyptos-
ternoids are not high and relationships among the differ-
ent clades are likely to be subject to change by using
different molecular data sets. However, in no case have
we found a topology in which Pareuchiloglanis form a
monophyletic group. Thus, in congruence with HeÕs
(1996) morphology result, we reject the monophyly of
Pareuchiloglanis based on the combined evidence. How-
ever, important taxa from the genus Pareuchiloglanis
were not intensively sampled for this study. Notably,
the branches where Oreoglanis delacouri and (Euchilogl-
anis +P. sinensis +P. anteanalis) diverged are not cer-
tain. Specially, Bayesian, ML, MP, and NJ analyses
produced different placement of Oreoglanis delacouri.
Given the non-monophyletic nature of Oreoglanis (Ng
and Rainboth, 2001), we should test it with molecular
data. Thus, it is imperative to include additional species
from genera Pareuchiloglanis and Oreoglanis to better
investigate the phylogenetic relationships among the
glyptosternoids.
4.2. Historical biogeographic scenario of Chinese sisorids
It is suggested that the sisorid catfishes derived from
the same ancestor as the Doumeinae, and moved to Asia
via the drift of the Indian subcontinent (He and Meu-
nier, 1998). A comparison of mtDNA divergences with-
in the sisorids suggests that the ancestral sisorid lineage
split into two main clades (‘‘Gagata +Bagarius +Gly-
ptothorax’’ and ‘‘Pseudecheneis + glyptosternoids’’) in
a relatively short period. Based on the molecular clock
calibration, we can at least tentatively date the origin
of Chinese sisorids earlier than the Oligocene–Miocene
boundary (19–24 Myr). Additionally, our results agree
with HeÕs hypothesis (1995) that the divergence of this
group has a direct relationship with three uplifting
events of the Qinhai-Tibetan Plateau.
These rough divergence timings should be interpreted
with caution, for they are based on calibrations derived
from other fishes, a strategy necessitated by a lack of
Chinese sisorids fossil data. Nevertheless, the use of a
wide range of calibrated rates should allow us to com-
Table 7
Results of dispersal–vicariance and weighted ancestral area analyses
Node
a
Suggested distribution
(DIVA)
Probability indices (WAAA)
ABCDEF
1 C 1.00 1.00 111.00 0.00
2 F 0.00 0.00 0.00 0.00 0.00 1
3 F 0.00 0.00 0.00 0.00 0.00 1
4 F 0.00 0.00 0.00 0.00 0.00 1
5 CDEF 0.00 0.00 1.00 1.00 1.00 1.00
6 C 0.33 0.33 3.00 3.00 1.00 0.33
7 C 1.60 1.60 5.50 0.63 0.40 0.18
8 C 0.50 3.50 8.33 0.33 0.25 0.12
9 BC 1.61 4.51 1.07 0.22 0.18 0.09
a
Node numbers and distribution legend refer to Fig. 7B.
358 X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362
pare molecular divergence times with the available data
on geological events in this area, and provide tests of
vicariance/dispersal hypotheses. Vicariance theorists as-
sume that common distributional patterns result from
shared vicariance events. Under a vicariance model, taxa
with parallel distributions would be expected to exhibit
similar amounts of genetic divergence. The hypothesis
of a vicariant event between the formerly connected
Tsangpo and Upper Irrawaddy is also supported by
data from Badidae species (Ru
¨ber et al., 2004). The
uncorrected cytochrome bsequence divergence between
Glyptosternum vs. the remaining glyptosternoids lin-
eages represented by Glaridoglanis is 17.35 ± 0.85%
(range 16.50–18.20%). Similar level of cytochrome b
divergence was found between the B. badis and B.assam-
ensis species group vs. the B. corycaeus species group
(18.24–19.56%) that show comparable allopatric distri-
bution (Ru
¨ber et al., 2004). Although preliminary, these
data indicate that the levels of cytochrome bsequence
divergence between species pairs occurring in the Tsang-
po–Brahmaputra drainages and the Irrawaddy drainage
are in the same range. Thus, our data support a hypoth-
esized paleo connection of the Tsangpo river with the
Irrawaddy drainage that was most likely interrupted
during Miocene orogenic events through tectonic uplifts
in eastern Tibet.
The reconstructed taxon-area cladogram by BPA was
partly not congruent with the geological evidence of
large-scale changes in drainage patterns in the Miocene
affecting the Irrawaddy- and Tsangpo–Brahmaputra
drainages in the eastern Himalayas. Thus, the general
area relationships may not be expected to conform to
a hierarchical pattern. Dispersal–vicariance analysis
and weighted ancestral area analysis provide much in-
sight into cladogenesis within glyptosternoids. A con-
nective ancestral distribution comprising the Tsangpo
river and Irrawaddy drainage is inferred for the glyptos-
ternoids ancestor (node 9, Fig. 7B; Table 7). In the glyp-
tosternoids, we can suggest an ancestral divergence into
Irrawaddy- and Tsangpo-clade, following several rapid
duplications (sympatric speciation, nodes 6–8 in Fig.
7B) within a relatively short time period. This suggests
that vicariance has played a fundamental role in the
glyptosternoids cladogenesis, but that vicariance alone
is not a sufficient explanation. The dispersal–vicariance
analysis highlights this finding, suggesting a minimum
of 12 dispersal events in the glyptosternoids (see Fig.
7B). For example, molecular-clock calibration between
Pseudexostoma yunnanensis and Pareuchiloglanis kam-
engensis (9.35 ± 0.96 Mya) postdates the cladogenesis
of Irrawaddy and Tsangpo drainages, and thus favors
dispersal rather than vicariance for the wide distribu-
tions of Pareuchiloglanis kamengensis (Fig. 7B). From
the suggested biogeographical history (Fig. 7B) and
molecular-clock calibration, we can infer the divergence
of the Euchiloglanis and Pareuchiloglanis sinensis and P.
anteanalis might have been caused by the third uplift
(3.4-1.7 MaBP) of the Plateau. Geological studies indi-
cated that the uplift events of the Qinhai-Tibetan Pla-
teau occurred more strongly and frequently from 3.4
Myr ago (Yu et al., 2000; and references therein). The
rich geological and ecological diversity of the Qinhai-Ti-
betan Plateau, together with habitat isolation due to
changing climatic conditions during the third uplift of
the Plateau, may have promoted rapid speciation in
small, isolated populations, thus allowing the fixation
of derived morphological characters. The main period
of salt forming was after 3.5 MaBP (An et al., 2001;
Li et al., 2001). This period was named Qinghai-Tibetan
Movement A phase. Violent climate shifts in the Qinhai-
Tibetan Plateau occurred about 1.7 MaBP, and the per-
iod was called Qinghai-Tibetan Movement C Phase (Li
et al., 1996). During that time, rivers around the Plateau
eroded headward and cut down intensively, such as the
Upper Yangtze and all those rivers trans-Himalayan be-
gan to erode and form terraces. It justified the dispersal
and duplication (sympatric speciation) events illustrated
in Fig. 7B.
Chu (1979) demonstrated that the origin of the glyp-
tosternoid fishes is most probably in the late Pliocene.
However, we argue that this time of origin estimates is
inconsistent with the historical episodes of geological
change. According to recent environmental data, biol-
ogy and chronology (Fang et al., 2002; Ge et al.,
2002; Sun, 1996), the Qinhai-Tibetan Plateau was sub-
ject to three cycles of uplift and two large-scale pene-
plain, during 45-38 MaBP, 25-17 MaBP, and 3.4-1.7
MaBP, respectively, with each uplift less than 2000 m
(Shi et al., 1999). Variations in the environment of
the Plateau accompanied the changes in elevation.
For sisorid catfishes, elevation of habitat can reflect
differences of environment. The environment suitable
for one genus is commensurate with the conditions that
are necessary when the genus colonized in evolutionary
history. In a sense, elevations adapted to each genus of
the sisorid catfishes may also represent the uplift height
of the Plateau attained when the genus appeared.
Glyptosternum fishes mainly inhabit altitudes of 2900–
3850 m. Given that the Plateau had raised up to more
than 2000 m, less than 4000 m above sea level during
25-17 MaBP (Shi et al., 1999), the origin of glyptoster-
noids could also have happened at that time, i.e., dur-
ing the late Oligocene and Miocene. Interestingly, this
speculation is practically in accordance with the molec-
ular- clock calibration that the origin of the glyptoster-
noids occurred at 19.03-24.08 MaBP, falling within the
time frame of the second uplift of the Qinghai-Tibetan
Plateau.
Our molecular evidence tentatively highlights the
importance of paleo river connections between the
Tsangpo and the Irrawaddy drainages. Calibration of
a molecular clock allowed us to date the proposed
X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362 359
vicariant scenario at the Oligocene–Miocene boundary
(19–24 Mya), which seems to agree with geological evi-
dence for the separation of these drainages caused by
tectonic uplifts in Eastern Tibet. These geological
events appear to have played a primary role in the
diversification of Chinese sisorids. Based on dis-
persal–vicariance analysis, we propose an evolutionary
scenario for the sisorids that combines aspects of both
vicariance and dispersal theory. Unfortunately, recon-
structions of biogeographic events require various
assumptions regarding the weighting of particular
events (e.g., dispersal, extinction, and vicariance) (Ron-
quist, 1997). Nearly any phylogenetic pattern of rela-
tionships can be explained by invoking various
combinations of vicariance, dispersal, and extinction.
Historical biogeography can be particularly difficult
when studying taxa with poor or nonexistent fossil re-
cords, because existing methods that are designed to
work in the absence of a general area cladogram
(e.g., DIVA and WAAA) may produce inferred ances-
tral distributions that are implausible. Undoubtedly,
additional tests of the phylogenies and biogeographic
scenarios proposed here are necessary. These tests
should include addition of new sequence data and taxa
for phylogenetic analysis, as well as exploration of dis-
persal/vicariance cost ratios or application of maxi-
mum-likelihood techniques, analogous to recent work
done on character state reconstruction (e.g., Omland,
1997; Schluter et al., 1997). Similar biogeographic anal-
yses for other taxa endemic to Qinhai-Tibetan Plateau,
in combination with the results presented here, will be
critical for constructing a general area cladogram of
basins in the Qinhai-Tibetan Plateau.
Acknowledgments
We extend our sincerest gratitude to Mrs. Sheng-
hong Zhou, Zhen-quan Ye, and Dr. Zuo-gang Peng
for assistance in collecting specimens or providing tis-
sues in their care, and Dr. C. Smith (University of
Leicester) for helpful suggestions. This work was sup-
ported by the Chinese Academy of Sciences (KSCX2-
SW-101B), the Innovation Program (220101) of the
Institute of Hydrobiology of Chinese Academy of Sci-
ences to S.P. He. Insightful comments from the anony-
mous reviewers improved the clarity and focus of the
manuscript.
References
Alves-Gomes, J.A., 1999. Systematic biology of gymnotiform and
mormyriform electric fishes: phylogenetic relationships, molecular
clocks and rates of evolution in mitochondrial rRNA genes. J. Exp.
Biol. 202, 1167–1183.
An, Z., Kutzbach, J.E., Prell, W.L., Porter, S.C., 2001. Evolution of
Asian monsoons and phased uplift of the Himalayas–Tibetan
Plateau since Late Miocene times. Nature 411, 62–66.
Brookfield, M.E., 1998. The evolution of the great river systems of
southern Asia during the Cenozoic India–Asia collision: rivers
draining southwards. Geomorphology 22, 285–312.
Brooks, D.R., Veller, M., Van McLennan, D.A., 2001. How to do
BPA, really. J. Biogeogr. 28, 345–358.
Buckley, T.R., Simon, C., Shimodaira, H., Chambers, G.K., 2001.
Evaluating hypotheses on the origin and evolution of the New
Zealand alpine cicadas (Maoricicada) using multiple-comparison
tests of tree topology. Mol. Biol. Evol. 8, 223–234.
Chu, X., 1979. Systematics and evolutionary pedigree of the
glyptosternoid fishes (Family Sisoridae). Acta Zootaxa. Sinica
4, 72–82.
Chu, X., 1982. Phylogeny of the genus Pseudecheneis (Siluriformes:
Sisoridae), with descriptions of two new species. Acta Zootaxon-
omica Sinica (in Chinese) 7, 428–437.
Chu, X., Zheng, B., Dai, D., 1999. Fauna Sinica, Class Teleostei,
Siluriformes (in Chinese). Scientific Press, Beijing.
Clark, M.K., Schoenbohm, L.M., Royden, L.H., Whipple, K.X.,
Burchfiel, B.C., Zhang, X., Tang, W., Wang, E., Chen, L., 2004.
Surface uplift, tectonics, and erosion of eastern Tibet from large-
scale drainage patterns. Tectonics 23, TC106, doi:10.1029/
2002TC001402.
de Pinna, M.C.C., 1996. A phylogenetic analysis of the Asian catfish
families Sisoridae, Akysidae, and Amblycipitidae, with a hypoth-
esis on the relationships of the neotropical Aspredinidae (Teleostei:
Ostariophysi). Fieldiana Zool. (N.S.) 84, 1–83.
De Riijk, P., Robbrecht, E., de Hoog, S., Caers, A., de Peer, V.Y., de
Wachter, R., 1999. Database on the structure of large subunit
ribosomal RNA. Nucleic Acids Res. 27, 174–178.
Diogo, R., Michel Chardon, M., Vandewalle, P., 2002. Osteology and
myology of the cephalic region and pectoral girdle of Glyptothorax
fukiensis (Rendahl, 1925), comparison with other sisorids, and
comments on the synapomorphies of the Sisoridae (Teleostei:
Siluriformes). Belg. J. Zool. 132, 93–101.
Fang, X., Li, J., Li, B., An, Z., Jiang, P., Chen, X., 2002. Loess in
Kunlun Mountains and its implications on desert development and
Tibetan Plateau uplift in West China. Sci. China (Series D) 45,
289–299.
Farris, J.S., Ka
¨llersjo
¨, M., Kluge, A.G., Bult, C., 1995. Testing
significance of incongruence. Cladistics 10, 315–319.
Felsenstein, J.P., 1985. Confidence limits on phylogenies: an approach
using the bootstrap. Evolution 39, 783–791.
Ge, X., Liu, Y., Ren, S., 2002. Uplift dynamics of Qinghai–Tibet
Plateau and Altun fault. Geol. China (in Chinese) 29, 346–
350.
Goldman, N., 1993a. Simple diagnostic statistical tests of models of
DNA substitution. J. Mol. Evol. 37, 650–661.
Goldman, N., 1993b. Statistical tests of models of DNA substitution.
J. Mol. Evol. 36, 182198.
Gregory, J.W., 1925. The evolution of the river system of South-
Eastern Asia. Scottish Geograph. Mag. XLI, 129–141.
Gregory, J.W., Gregory, C.J., 1923. The Alps of Chinese Tibet and
their geographical relations. Geograph. J. LXI, 153–179.
Gu, X., Fu, Y.X., Li, W.H., 1995. Maximum likelihood estimation of
the heterogeneity of substitution rate among nucleotide sites. Mol.
Biol. Evol. 12, 546–557.
Guo, X., Zhang, Y., He, S., Chen, Y., 2004. Mitochondrial 16S rRNA
sequence variations and phylogeny of the Chinese sisorid catfishes.
Chinese Sci. Bull. 49, 1586–1595.
Hallet, B., Molnar, P., 2001. Distorted drainage basins as markers of
crustal strain east of the Himalaya. J. Geophys. Res. 106, 13697–
13709.
Hausdorf, B., 1998. Weighted ancestral area analysis and a solution to
the redundant distribution problem. Syst. Biol. 47, 445–456.
360 X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362
He, D., Chen, Y., Chen, Y., Chen, Z., 2004. Molecular phylogeny of
the specialized schizothoracine fishes (Teleostei: Cyprinidae), with
their implications for the uplift of the Qinghai-Tibetan Plateau.
Chinese Sci. Bull. 49, 39–48.
He, S., 1995. The analysis of historical biogeography for the
glyptosternoid fishes (Teleoste: Siluriformes, Sisoridae). Biogeo-
graphica 71, 145–160.
He, S., 1996. The phylogeny of the glyptosternoid fishes (Teleostei:
Siluriformes, Sisoridae). Cybium 20, 115–159.
He, S., Meunier, F., 1998. The biogeography of Amphiliidae and
Sisoridae (Teleositei: Siluriformes). Acta Zootax. Sinica. (in Chi-
nese) 23 (Suppl.), 1–9.
Hillis, D.M., Huelsenbeck, J.P., 1992. Signal, noise, and reliability in
molecular phylogenetic analyses. J. Heredity 83, 189–195.
Hora, S.L., 1939. The game fishes of Indian, Part VI. The Goonch,
Bagarius bagarius (Hamilton). J. Bombay. Nat. Hist. Soc., 583–
593.
Hora, S.L., 1952. Parallel evolution of Pseudecheneis Blyth and similar
fishes of South-East Asia. J. Asiatic Soc. 18, 123–128.
Hora, S.L., Silas, E.G., 1952a. Evolution and distribution of glyptos-
ternoid fishes of the family Sisoridae (Order: Siluroidea). Proc. Nat.
Inst. Sci. Indian 18, 309–322.
Hora, S.L., Silas, E.G., 1952b. Notes on fishes in the Indian Museum.
XLVII. Revision of glyptosternoid fishes of the family Sisoridae,
with description of new genera and species. Rec. Indian Mus. 49, 5–
30.
Huelsenbeck, J.P., Crandall, K.A., 1997. Phylogeny estimation and
hypothesis testing using maximum likelihood. Annu. Rev. Evol.
Syst. 28, 437–466.
Huelsenbeck, J.P., Ronquist, F., 2001. MrBayes: bayesian inference of
phylogeny. Bioinformatics 17, 754–755.
Kimura, M., 1980. A simple method for estimating evolutionary rate
of base substitutions through comparative studies of nucleotide
sequences. J. Mol. Evol. 16, 111–120.
Kottelat, M., 1989. Zoogeography of the fishes from Indochines
inlandwaters with an annotated check-list. Bull. Zoo
¨l. Mus. 12, 1–
56.
Li, J., Fang, X., Ma, H., et al., 1996. Geomorphologic and environ-
mental evolution in upper reaches of Yellow River during the late
Cenozoic. Sci. China (Series D) 39, 380–390.
Li, J., Fang, X., Pan, B., 2001. Late Cenozoic intensive uplift of
Qinghai-Xizang Plateau and its impact on environments in
surrounding area. Q. Sci. 5, 381–391.
Luo, J., Yang, D., Suzuki, H., Wang, Y., Chen, W., Campbell, K.L.,
Zhang, Y., 2004. Molecular phylogeny and biogeography of
Oriental voles: Genus Eothenomys (Muridae, Mammalian). Mol.
Phylogenet. Evol. 33, 349–362.
Martin, A.P., Palumbi, S.R., 1993. Body size, metabolic rate,
generation time, and the molecular clock. Proc. Natl. Acad. Sci.
USA 90, 4087–4091.
Me
´tivier, F., Gaudemer, Y., Tapponnier, P., Klein, M., 1999. Mass
accumulation rates in Asia during the Cenozoic. Geophys. J. Int.
137, 280–318.
Ng, H.H., 2004a. Oreoglanis macronemus, a new species of
glyptosternine catfish (Teleostei: Siluriformes: Sisoridae)
from northern Laos. The Raffles Bulletin of Zoology 52, 209–
213.
Ng, H.H., 2004b. Two new glyptosternine catfishes (Teleostei: Siso-
ridae) from Vietnam and China. Zootaxa 428, 1–12.
Ng, H.H., Rainboth, W.J., 2001. A review of the sisorid catfish genus
Oreoglanis (Siluriformes: Sisoridae) with descriptions of four new
species. Occasional Papers of the Museum of Zoology, University
of Michigan, 732, 1–34.
Omland, K.E., 1997. Examining two standard assumptions of ances-
tral reconstructions: Repeated loss of dichromatism in dabbling
ducks (Anatini). Evolution 51, 1636–1646.
Palumbi, S.R., 1996. Nucleic acids II: the polymerase chain reaction.
In: Hillis, D.M., Moritz, C., Mable, B.K. (Eds.), Molecular
Systematics, second ed. Sinauer Associates, Sunderland, MA.
Pang, J., Wang, Y., Zhong, Y., Hoelzel, A.R., Papenfuss, T.J., Zeng, X.,
Ananjeva, N.B., Zhang, Y., 2003. A phylogeny of Chinese species in
the genus Phrynocephalus (Agamidae) inferred from mitochondrial
DNA sequences. Mol. Phylogenet. Evol. 27, 398–409.
Peng, Z., He, S., Zhang, Y., 2004. Phylogenetic relationships of
glyptosternoid fishes (Siluriformes: Sisoridae) inferred from mito-
chondrial cytochrome bgene sequences. Mol. Phylogenet. Evol. 31,
979–987.
Posada, D., Crandall, K.A., 1998. MODELTEST: testing the model of
DNA substitution. Bioinformatics 14, 817–818.
Regan, C.T., 1911. The classification of the Teleostean fishes of the
Order Ostariophysi-2. Siluridea. Ann. Mag. N. Hist. vol. 8 (Ser. 8),
553–557.
Ronquist, F., 1996. DIVA version 1.1. Computer program and manual
available by anonymous FTP from Uppsala University (ftp.uu.se
or ftp.systbot.uu.se).
Ronquist, F., 1997. Dispersal–vicariance analysis: a new approach to the
quantification of historical biogeography. Syst. Biol. 46, 195–203.
Ru
¨ber, L., Britz, R., Kullandr, S.O., Zardoya, R., 2004. Evolutionary
and biogeographic patterns of the Badidae (Teleostei: Perciformes)
inferred from mitochondrial and nuclear DNA sequence data. Mol.
Phylogenet. Evol. 32, 1010–1022.
Saitoh, K., Miya, M., Inoue, J.G., Ishiguro, N.B., Nishida, M., 2003.
Mitochondrial genomics of ostariophysan fishes: perspectives on
phylogeny and biogeography. J. Mol. Evol. 56, 464–472.
Sambrook, J., Fritsch, E.F., Maniatis, T., 1989. Molecular Cloning: A
Laboratory Manual, second ed. Cold Spring Harbor Laboratory
Press, New York.
Sanmartı
´n, I., 2003. Dispersal versus vicariance in the Mediterranean:
historical biogeography of the palearctic Pachydeminae (Coleop-
tera, Scarabaeoidea). J. Biogeogr. 30, 1883–1997.
Sanmartı
´n, I., Ronquist, F., 2002. New solutions to old problems:
widespread taxa, redundant distributions, and missing areas in
event-based biogeography. Animal Biodiv. Conserv. 25, 75–93.
Schluter, D., Price, T., Mooers, A.Ø., Ludwig, D., 1997. Likelihood of
ancestor states in adaptive radiation. Evolution 51, 1699–1711.
Seeber, L., Gornitz, V., 1983. River profiles along the Himalayan arc
as indicators of active tectonics. Tectonophysics 92, 335–367.
Shi, Y.F., Li, J., Li, B., et al., 1999. Uplift of the Qinghai-Tibetan
Plateau and East Asia environmental change during late Cenozoic.
Acta Geographical Sinica (in Chinese). 54, 10–20.
Shimodaira, H., Hasegawa, M., 1999. Multiple comparisons of
loglikelihoods with applications to phylogenetic inference. Mol.
Biol. Evol. 16, 1114–1116.
Sun, H.L., 1996. Science monographic series: formation and evolution
of Qinghai-Xizang Plateau. In: Sun, H.L. (Ed.), Scientific Mono-
graph Series. Shanghai Science and Technology Press, Shanghai,
pp. 101–110.
Swofford, D.L., 2002. PAUP*. Phylogenetic Analysis Using Parsi-
mony (*and other methods). Version 4. Sinauer, Sunderland, MA.
Templeton, A.R., 1983. Phylogenetic inference from restriction endo-
nuclease cleavage site maps with particular reference to the
evolution of humans and the apes. Evolution 37, 221–244.
Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins,
D.G., 1997. The Clustal X windows interface: flexible strategies for
multiple sequences alignment aided by quality analysis tools.
Nucleic Acids Res. 25, 4876–4882.
Tilak, R., 1976. The adhesive thoracic apparatus in the evolution of
glyptothoracoid fishes (Siluriformes: Sisoridae). Zool. Anz. 196,
255–261.
Watanabe, K., Zhang, C., Zhao, Y., 2002. Redescription of the East
Asian bagrid catfish, Pseudobagrus kyphus Mai, 1978, and a new
record from China. Ichthyol. Res. 49, 384–388.
X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362 361
Wiens, J.J., 1998. Combining data sets with different phylogenetic
histories. Syst. Biol. 47, 568–581.
Wilcox, T.P., Garcia de LeO
´n, F.J., Hendrickson, D.A., Hillis, D.M.,
2004. Convergence among cave catfishes: long-branch attraction and
a Bayesian relative rates test. Mol. Phylogenet. Evol. 31, 1101–1113.
Xiao, W., Zhang, Y., Liu, H., 2001. Molecular systematics of
Xenocyprinae (Teleostei: Cyprinidae): Taxonomy, biogeography,
and coevolution of a special group restricted in East Asia. Mol.
Phylogenet. Evol. 18, 163–173.
Yang, Z., 1994. Estimating the pattern of nucleotide substitution. J.
Mol. Evol. 39, 105–111.
Yu, N., Zheng, C., Zhang, Y.P., Li, W.H., 2000. Molecular systematics
of pikas (Genus Ochotona) inferred from mitochondrial DNA
sequences. Mol. Phylogenet. Evol. 16, 85–95.
Zeitler, P.K., Meltzer, A.S., Koons, P.O., Craw, D., Hallet, B.,
Chamberlain, C.P., Kidd, W.S.F., Park, S.K., Seeber, L., Bishop,
M., Shroder, J., 2001. Erosion, Himalayan geodynamics, and the
geomorphology of metamorphosis. GSA Today 11, 4–9.
362 X. Guo et al. / Molecular Phylogenetics and Evolution 35 (2005) 344–362
