From the Old World to the New World: A Molecular Chronicle of the Phylogeny and Biogeography of Hystricognath Rodents

Abstract

Hystricognath rodents include Old World Phiomorpha and New World Caviomorpha. These two groups have an enigmatic biogeographical history. Using a nuclear marker, the exon 28 of the von Willebrand Factor gene (vWF), we reconstructed the phylogenetic relationships among 23 Hystricognathi species. These taxa encompass the complete familial diversity of the Hystricognathi. Our results indicate a basal trifurcation of hystricognaths leading to Hystricidae, Phiomorpha s.s. (Bathyergidae, Thryonomyidae, and Petromuridae), and Caviomorpha. The monophyly of caviomorphs is robustly supported, confirming a single colonization event of South America by hystricognaths. Caviomorpha are divided into four lineages: Cavioidea, Erethizontoidea, Chinchilloidea, and Octodontoidea. Furthermore, we suggest that (1) Chinchillidae and Dinomyidae are sister clades, (2) Abrocomidae is a true Octodontoidea, and (3) Capromyidae, Echimyidae, and Myocastoridae cluster together. Surprisingly, Erethizontidae does not appear to be the most diverged caviomorph lineage. The molecular results are discussed in the light of previous paleontological and morphological observations. Local molecular clocks are used to estimate divergence dates among hystricognath lineages. An Asian origin is suggested for Caviomorpha, and a colonization route through Australia and Antarctica is indicated as an alternative to the hypothesis of a transatlantic migration of Caviomorpha from Africa to South America.

Full text

From the Old World to the New World: A Molecular Chronicle
of the Phylogeny and Biogeography of Hystricognath Rodents
Dorothe´e Huchon1and Emmanuel J. P. Douzery2
Laboratoire de Pale´ontologie, Pale´obiologie et Phyloge´nie-CC064, Institut des Sciences de l’Evolution UMR 5554-CNRS,
Universite´ Montpellier II, Place E. Bataillon, 34 095 Montpellier Cedex 05, France
Received October 10, 2000; revised February 14, 2001
Hystricognath rodents include Old World Phiomor-
pha and New World Caviomorpha. These two groups
have an enigmatic biogeographical history. Using a
nuclear marker, the exon 28 of the von Willebrand
Factor gene (vWF), we reconstructed the phylogenetic
relationships among 23 Hystricognathi species. These
taxa encompass the complete familial diversity of the
Hystricognathi. Our results indicate a basal tri-
furcation of hystricognaths leading to Hystricidae,
Phiomorpha s.s. (Bathyergidae, Thryonomyidae, and
Petromuridae), and Caviomorpha. The monophyly of
caviomorphs is robustly supported, confirming a sin-
gle colonization event of South America by hystricog-
naths. Caviomorpha are divided into four lineages:
Cavioidea, Erethizontoidea, Chinchilloidea, and Oct-
odontoidea. Furthermore, we suggest that (1) Chin-
chillidae and Dinomyidae are sister clades, (2)
Abrocomidae is a true Octodontoidea, and (3) Capro-
myidae, Echimyidae, and Myocastoridae cluster to-
gether. Surprisingly, Erethizontidae does not appear
to be the most diverged caviomorph lineage. The mo-
lecular results are discussed in the light of previous
paleontological and morphological observations. Lo-
cal molecular clocks are used to estimate divergence
dates among hystricognath lineages. An Asian origin
is suggested for Caviomorpha, and a colonization
route through Australia and Antarctica is indicated as
an alternative to the hypothesis of a transatlantic mi-
gration of Caviomorpha from Africa to South America.
© 2001 Academic Press
Key Words: phylogeny; nuclear gene (vWF); Roden-
tia; Hystricognathi; Caviomorpha; Dinomyidae; Chin-
chilloidea; Phiomorpha; biogeography; South Amer-
ica; Antarctica.
INTRODUCTION
South America was an isolated continent during the
Tertiary. This isolation from the Late Eocene to the
Mio–Pliocene produced a variety of endemic mamma-
lian groups belonging to marsupials, edentates, ungu-
lates, primates, and rodents. The Cenozoic mammalian
evolution in South America may be summarized into
three-phases (Flynn and Wyss, 1998): establishment of
archaic groups during the Paleocene (various marsupi-
als, xenarthrans, notoungulates, and litopterns), ar-
rival of two new immigrants (i.e., primates [Platyr-
rhini] and rodents [Caviomorpha]) at the Eocene–
Oligocene transition, and establishment of modern
faunas with Northern invaders crossing the Panama-
nian land bridge during the Great American Inter-
change at the Mio–Pliocene. The isolation of South
America raises intriguing questions about the biogeo-
graphical origins and phylogenetic relationships of
platyrrhines and caviomorphs: (i) Are they monophy-
letic lineages, arising from a single invasion event, or
polyphyletic groups, possibly stemming from multiple
invasions? (ii) Are they, as it is generally supposed,
closely related to African forms (Catarrhini primates
and Phiomorpha rodents)? (iii) What were their migra-
tion routes from the Old World to the New World?
In this study, we focus on Rodentia, an order repre-
senting 43% of the 4629 recognized living mammalian
species (Wilson and Reeder, 1993). Among rodents, the
South American caviomorphs belong to the infraorder
Hystricognathi. Hystricognaths (230 species in 68 gen-
era and 17 families) are well defined by morphological
(Luckett and Hartenberger, 1993) and molecular
(Catzeflis et al., 1995; Nedbal et al., 1996; Huchon et
al., 1999) characters. The hystricognath sister group is
represented by a family of North African rodents, the
Ctenodactylidae—or gundis—(Bryant and McKenna,
1995; Huchon et al., 2000). Hystricognathi is divided
into two well-defined biogeographical taxa: the Old
World Phiomorpha (sensu Lavocat, 1973) and the New
World Caviomorpha (sensu Wood, 1955). Caviomorphs
suddenly appeared and diversified in South America in
1Present address: Tokyo Institute of Technology, Faculty of Bio-
science and Biotechnology, Molecular Evolution Laboratory, 4259
Nagatsuta-cho, Midori-ku, 226-8501 Yokohama, Japan.
2To whom correspondence should be addressed. Fax: 33 4 67 14 36
10. E-mail: douzery@isem.univ-montp2.fr.
Molecular Phylogenetics and Evolution
Vol. 20, No. 2, August, pp. 238–251, 2001
doi:10.1006/mpev.2001.0961, available online at http://www.idealibrary.com on
1055-7903/01 $35.00
Copyright © 2001 by Academic Press
All rights of reproduction in any form reserved.
238

the Early Oligocene (31 Myr ago; Wyss et al., 1993;
Vucetich et al., 1999). Their route of colonization is
fully enigmatic as hystricognath fossils have been iden-
tified only in South America, Africa, and Eurasia. A
North American origin of caviomorphs was proposed
(e.g., Wood, 1985) but this hypothesis might be based
on erroneous interpretation of fossil characters (e.g.,
Meng, 1990; Martin, 1994). The prevailing hypothesis
suggests an African origin of caviomorphs and a trans-
atlantic migration (e.g., Lavocat, 1969). Moreover, the
number of migration events is debated, and some phy-
logenetic studies based on morphology suggest that
caviomorphs are paraphyletic and have arisen from
two independent colonizations (Bugge, 1985; Woods
and Hermanson, 1985; Bryant and McKenna, 1995;
McKenna and Bell, 1997) or even more (Landry, 1999).
Most studies rather suggest a more parsimonious sce-
nario, with a single origin of caviomorphs (Wood and
Patterson, 1959; Nedbal et al., 1994).
Old World hystricognaths (Phiomorpha sensu lato)
include four extant families (Wilson and Reeder, 1993;
McKenna and Bell, 1997): Thryonomyidae (cane rats)
and Petromuridae (dassie rats), which are considered
to cluster together into the superfamily Thryonomy-
oidea (Lavocat, 1973; Nedbal et al., 1994); Bathyergi-
dae (African mole rats); and Hystricidae (Old World
porcupines). The fossil record indicates that Thryono-
myidae arose in the Late Eocene–Early Oligocene of
Africa. Petromuridae is a recent family appearing in
the Pleistocene (Winkler, 1994; McKenna and Bell,
1997). The origin of Bathyergidae is debated, with ei-
ther an African or an Asian Miocene origin (Winkler,
1994). Hystricidae is the most enigmatic. It presents
many ancestral characters but their predecessors are
unknown, and the earliest fossils are recorded only
since the Early Miocene of Europe (McKenna and Bell,
1997). However, paleontologists hypothesized that hys-
tricids have an Asian or Indian origin (e.g., Hussain et
al., 1978; Wood, 1985; Winkler, 1994). Caviomorpha
are more diversified, with 13 extant New World fami-
lies (Wilson and Reeder, 1993) grouped into four super-
families: Erethizontoidea (New World porcupines),
Cavioidea (guinea pigs), Octodontoidea (spiny rats),
and Chinchilloidea (chinchillas).
Because the morphological evolution of rodents is
characterized by high levels of homoplasy (Wood and
Patterson, 1959; Jaeger, 1988), an alternative way to
understand the origin and diversification of South
American hystricognaths is to explore their phylogeny
with molecular characters. Studies based on the mito-
chondrial 12S rRNA gene failed to give robust phylo-
genetic results on the relationships among the main
caviomorph and phiomorph lineages (Nedbal et al.,
1994; Catzeflis et al., 1995). Studies based on exon 28 of
the nuclear von Willebrand factor (vWF) gene ap-
peared promising but included only a few hystricog-
nath species (Huchon et al., 1999, 2000). To accurately
estimate the number of South America colonizations by
rodents, it is important to sample all the currently
available hystricognath biodiversity, without excep-
tion. To reach this goal, we increased the vWF data
bank to obtain a complete set of orthologues sequences.
It is the first time, to our knowledge, that at least one
representative of each of the 13 caviomorph and of each
of the 4 phiomorph families has been sampled in a
study at the nucleotide level.
This study addresses three questions about rodent
evolution. (1) How many waves of hystricognath ro-
dents colonized South America? A single colonization
event implies that caviomorphs are monophyletic. (2)
What is the closest sister clade of caviomorphs? Its
identification could help to understand the caviomorph
routes of migration. (3) What is the timing of these
colonization events as deduced from molecular data? If
some nucleotide substitutions or amino acid replace-
ments behave clock-like in the vWF molecule, a tem-
poral framework will be drawn to better understand
the evolution of hystricognaths.
MATERIAL AND METHODS
Species Sampling
The dataset includes at least one representative for
each hystricognath family (Table 1). Because there is a
general agreement about the number of rodent families
and their content (e.g., Hartenberger, 1985), it seems
reasonable to consider any species sampled a fair rep-
resentative of its family. Two to four species were con-
sidered for the Echimyidae, the Bathyergidae, and the
key Old World and New World porcupine families.
Ctenodactylidae (gundis) were chosen as a close out-
group because it has been shown that they are the
sister group of hystricognath rodents on the basis of
paleontological (Bryant and McKenna, 1995), morpho-
logical (Landry, 1999), and molecular (Huchon et al.,
2000) characters. Because Robinson et al. (1998) sug-
gested that for distance analysis the most reliable out-
groups are those closely related to the ingroup and
slowly evolving, Aplodontia (mountain beaver; Ap-
lodontidae) and Spalax (blind mole rat; Spalacidae)
were selected as more distant outgroups because of
their slow-evolving vWF exon 28 (Huchon et al., 2000).
DNA Sequencing of vWF Exon 28
Tissue samples were derived mostly from the Collec-
tion of Mammalian Tissues of Montpellier (France)
(Catzeflis, 1991). Taxonomy, origin, and references of
the tissues are indicated in Table 1. DNA extractions
and vWF exon 28 amplifications with V1/W1 primers
(V1direct ⫽5⬘-TGTCAACCTCACCTGTGAAGCCTG-3⬘
and W1reverse ⫽5⬘-TGCAGGACCAGGTCAGGAGCCT-
CTC-3⬘) were conducted according to Huchon et al.
(1999). For Atherurus macrourus, V1 was replaced by
239SOUTH AMERICA COLONIZATION BY CAVIOMORPH RODENTS

primer V5direct (5⬘-GGAGGCCTGGHRGTGCCCCC-3⬘)
(Huchon et al., 2000). For most species, smaller over-
lapping PCR products were obtained after a reamplifi-
cation step with V1 (or V5)/W2 and V2/W1 pairs of
primers (V2direct ⫽5⬘-CCCTCAGAGCTGCGGCGCA-
T-3⬘and W2reverse ⫽5⬘-ACGTCCATGCGCTGGATCA-
CCT-3⬘) (Huchon et al., 1999). For Echimys chrysurus,
Abrocoma bennettii, Myocastor coypus, and Capromys
pilorides, one part of the vWF exon 28 was cloned in
the pCR 2.1 plasmid vector with the Original TA clon-
ing kit (Invitrogen, Carlsbad, CA) and bacterial trans-
formation in Escherichia coli strain INV
␣
F⬘. PCR prod-
ucts and recombinant plasmids were purified and
directly sequenced on both strands with [
␣
33P]ddNTP
and the Thermo Sequenase radiolabeled terminator
cycle sequencing kit (Amersham, Cleveland, OH). The
exon 28 vWF DNA sequences have been deposited in
the EMBL/GenBank/DDBJ databases (Table 1).
TABLE 1
Taxonomic Frame for the Hystricognath Rodents Studied, following McKenna and Bell (1997) for Caviomor-
pha Superfamily Content and Wilson and Reeder (1993) for Family Names and Numbers of Genera/Species
(between Parentheses)
Genera/species Latin name Common name Accession Origin (donator)/references
Phiomorpha (10/26)
Hystricidae (3/11) Atherurus macrourus Brush-tailed porcupine AJ251131aVietnam (J. L. Patton)/T-1751
Trichys fasciculata Long-tailed porcupine AJ224675 Huchon et al. (1999)
Petromuridae (1/1) Petromus typicus Dassie rat AJ251144aZoological Society of Philadelphia,
USA (R. Hoyt)
Thryonomyidae (1/2) Thryonomys swinderianus Cane rat AJ224674 Huchon et al. (1999)
Bathyergidae (5/12) Cryptomys hottentotus Common mole-rat AJ251132aPietermaritzburg, South Africa
(E. Nevo)/T-266
Heliophobius argenteocinereus Silvery mole-rat AJ251133aKenya (C. G. Faulkes)/T-1846
Heterocephalus glaber Naked mole-rat AJ251134aKenya (C. G. Faulkes)/T-1848
Bathyergus suillus Dune mole-rat AJ238384 Huchon et al. (2000)
Caviomorpha (59/208)
Erethizontidae (4/12) Erethizon dorsatum North American porcupine AJ251135a(J. A. W. Kirsh)/T-1789
Coendou melanurus Prehensile tailed porcupine AJ224664 Huchon et al. (1999)
Cavioidea
Agoutidae (1/2) Agouti paca Paca AJ251136aPetit Saut, French Guyana
(J.-C. Vie´)/T-1555
Dasyproctidae (2/13) Dasyprocta leporina Agouti U31607 Porter et al. (1996)
Dinomyidae (1/1) Dinomys branickii Pacarana AJ251145aCleveland Metroparks Zoo, USA
(T. L. Bettinger)
Caviidae (5/14) Cavia porcellus Guinea pig AJ224663 Huchon et al. (1999)
Hydrochaeridae (1/1) Hydrochaeris hydrochaeris Capybara AJ251137aPetit Saut, French Guyana
(J.-C. Vie´)/T-1618
Octodontoidea
Octodontidae (6/9) Octodon lunatus Degu AJ238386 Huchon et al. (2000)
Ctenomyidae (1/38) Ctenomys maulinus Tuco-tuco AJ251138aTalca, Chili (L. Contreras)/T-1005
Echimyidae (21/80) Proechimys oris Spiny rat (or casiragua) AJ251139aBreeding colony, Brazil
(F. Petter)/T-0311
Echimys chrysurus White-faced tree rat AJ251141aPetit Saut, French Guyana
(F. Catzeflis)/A-024
Myocastoridae (1/1) Myocastor coypus Nutria AJ251140aGard, France (B. de Sousa)/T-1811
Capromyidae (12/25) Capromys pilorides Cuban hutia AJ251142aZoo Viena, Austria (H. Burger
and A. Kuebber-Heiss)/T-1845
Chinchilloidea
Chinchillidae (3/7) Chinchilla lanigera Chinchilla AJ238385 Huchon et al. (2000)
Abrocomidae (1/3) Abrocoma bennettii Chinchilla rat AJ251143aParque National Fray Jorge, Chili
(L. Contreras)/T-1004
Outgroups
Ctenodactylidae Ctenodactylus vali Gundi AJ238387 Huchon et al. (2000)
Massoutiera mzabi Gundi AJ238388 Huchon et al. (2000)
Aplodontidae Aplodontia rufa Mountain beaver AJ224662 Huchon et al. (1999)
Muridae Spalax polonicus Blind mole-rat U31621 Porter et al. (1996)
Note. Information provided: Latin name, common name, EMBL data bank accession numbers for vWF exon 28 sequences, origin of the
animal (name of the donator)/reference number in the collection of mammalian tissues of the “Institut des Sciences de l’Evolution of
Montpellier” (Catzeflis, 1991).
aThis paper.
240 HUCHON AND DOUZERY

Sequence Alignment
Sequences were aligned by hand with the ED editor
(MUST package; Philippe, 1993) and compared with
published orthologues. The alignment was unequivo-
cal, with only one deletion of 1 codon for Proechimys ori
and of 10 codons for Petromus typicus. In the subse-
quent analyses, all sites were kept, and gaps were
coded as missing data. Among the 1263 aligned posi-
tions, 664 and 455 were, respectively, variable and
parsimony informative.
Phylogenetic Analyses
Data were analyzed by neighbor-joining (NJ), maxi-
mum-parsimony (MP), and maximum-likelihood (ML)
methods with PAUP* (Swofford, 1998), versions 4.0b2
and 4.0b4a. Complementary ML analyses, with the
quartet puzzling method (MLQ), and the Kishino and
Hasegawa (1989) tests were conducted with TREE-
PUZZLE 4.0.1 (Strimmer and von Haeseler, 1996). For
ML analyses, different models of sequence evolution
were compared with the likelihood ratio test (LRT)
following the approach of Sullivan and Swofford
(1997). After these comparisons, the most general
model of sequence evolution available under the pro-
gram considered was chosen (see Results for details):
GTR under PAUP 4.0b2 and TN93 under TREE-PUZ-
ZLE 4.0.1 for nucleotides, and JTT for amino acids.
Rate variation among sites was described by a gamma
distribution with eight categories [⌫8] (Yang, 1996).
GTR and gamma parameters were optimized during a
ML heuristic search with NNI branch swapping.
Neighbor-joining reconstructions were performed on
GTR distances with gamma (⌫) rates or on constant
site removal LogDet (LD) distances. Description of the
substitution rate heterogeneities among sites by a
gamma distribution improves phylogenetic reconstruc-
tion and dating estimation (Yang, 1996). Use of LogDet
distances allows management of a data matrix for
which the assumption of stationary base frequencies is
violated (Lockhart et al., 1994).
For standard MP analyses (i.e., equal weights for all
nucleotide changes), heuristic searches were done with
the TBR branch swapping option and 100 random ad-
dition of sequences. For weighted MP analyses (MPw),
each of the six nucleotide changes (e.g., A 7G) at each
of the three codon positions was weighted according to
the product CI ⫻S (e.g., for A 7G changes at the first
codon position, CI represents the consistency index
excluding uninformative characters of the most parsi-
monious cladogram reconstructed only from first codon
position A and G states, i.e., coding C and T as missing
data in the matrix), and S represents the slope of the
saturation plot between observed changes against in-
ferred substitutions (Hassanin et al., 1998a,b; Hassa-
nin and Douzery, 1999). The cladograms used for cal-
culating CI and S may be different for each of the 18
pairs of nucleotide changes. This point allows an a
priori measurement of the homoplasy and saturation of
each nucleotide change, instead of an a posteriori eval-
uation of the levels of homoplasy and saturation on a
single MP tree (e.g., derived from a standard MP anal-
ysis).
The robustness of the nodes of the trees was assessed
by (i) bootstrap percentages (BP; with PAUP 4.0b2)
after 1000 replicates of resampling for NJ, 1000 for MP
(one random addition of sequences; TBR branch swap-
ping), and 100 for ML (NJ starting tree; TBR branch
swapping); and (ii) reliability percentages (RP; with
TREE-PUZZLE 4.0.1) estimated after 1000 ML quar-
tet puzzling steps.
MOLPHY 2.3b3 (Adachi and Hasegawa, 1996) was
used to write all the bifurcating topologies connecting
the various major caviomorph and phiomorph clades.
Likelihood scores were then compared by the one-
tailed normal approximation test of the difference (
␦
)of
two log-likelihoods (Kishino and Hasegawa, 1989).
Molecular Evolutionary Rates and Molecular Dating
The two-cluster and branch-length tests of the LIN-
TRE package (Takezaki et al., 1995) were used to iden-
tify species evolving significantly slower and faster
relative to the others. Because of the substitution rate
differences observed between vWF sequences, we
adopted a compromise between constraining a unique
substitution rate for all branches of the ML tree (i.e.,
setting a global clock) and having independent rates
for each branch (i.e., no clock), by using the local clocks
approach proposed by Yoder and Yang (2000). Local
clocks were enforced with PAML, version 3.0b (Yang,
2000). Likelihood ratio tests (Felsenstein, 1988) be-
tween log-likelihoods of clock-constrained and non-
clock-constrained trees allowed testing for the hypoth-
esis of local molecular clocks in the vWF data.
Few data are available about the origin of the cur-
rent hystricognath families, and a single calibration
point was available for our clock-like tree: the cavi-
omorph radiation at 31 Myr. The choice of this date is
justified by (i) the occurrence of the first caviomorph
fossil in the Tinguirirican (South American land mam-
mal age, SALMA; 31–37 Myr; Wyss et al., 1993),
whereas all South American rodent families are iden-
tified in the Deseadan (SALMA; 24.5–29 Myr; e.g.,
Walton, 1997); and (ii) its compatibility with other
mammalian divergence dates (Huchon et al., 2000).
RESULTS
Phylogenetic Reconstructions
The vWF of Heliophobius was the single sequence
among the 27 sequences having a base composition
deviating from the frequency distribution assumed in
the ML model (1% significance of a
␹
2test in TREE-
PUZZLE 4.0.1). If the third codon positions were re-
241SOUTH AMERICA COLONIZATION BY CAVIOMORPH RODENTS

moved, no base composition deviation was observed. To
evaluate the impact of third codon positions and the
presence of Heliophobius, analyses were conducted
with and without these characters and this taxon.
Different ML models have been compared to describe
the data. First, the GTR model fitted the complete vWF
better than TN93 when the highest-likelihood phylo-
gram was considered (LRT statistics ⫽13.22, P⫽
0.001). Second, GTR ⫹⌫
8was better than GTR without
rate heterogeneity (LRT ⫽997.20, P⬍0.001). Third,
GTR ⫹⌫
8⫹I was marginally not better than GTR ⫹
⌫8without invariable sites (LRT ⫽3.38, P⫽0.07).
Fourth, when third codon positions were excluded, the
same results were observed. To run ML analyses with
and without third codon positions, we chose the GTR ⫹
⌫8model as it resulted in a significantly better fit of the
data relative to models with fewer parameters.
ML phylograms were rooted with Aplodontia and
Spalax (Fig. 1). All ML reconstructions indicated that
(i) Ctenodactylidae was the sister group of all hystri-
cognaths which clustered in an unambiguously mono-
phyletic group; (ii) Caviomorpha, here represented by
at least one member of each living family, was mono-
phyletic; (iii) Phiomorpha s.l. was composed of three
distinct clades; Hystricidae, Thryonomyidae ⫹Petro-
muridae (Thryonomyoidea), and Bathyergidae; the lat-
FIG. 1. Highest-likelihood phylograms of 23 hystricognath and 2 ctenodactylid vWF exon 28 sequences, rooted by Aplodontia and Spalax,
reconstructed from all codon positions (left) and from first and second codon positions (right). The GTR model of sequence evolution with
eight-categories gamma rates (parameter
␣
) has been used for the two maximum-likelihood (ML) reconstructions, with the following values:
(i) rate matrix ⫽(1.27 [A 7C substitutions], 6.77 [AG], 0.90 [AT], 1.25 [CG], 9.28 [CT], 1.00 [GT]), and
␣
⫽0.49 for all positions (lnL ⫽
⫺9599.86); (ii) rate matrix ⫽(2.42, 8.02, 1.09, 2.76, 5.97, 1.00), and
␣
⫽0.50 after removal of third codon positions (lnL ⫽⫺4528.15). ML
bootstrap percentages after 100 replicates (NJ starting trees and TBR branch swapping) are given at the nodes. Each star indicates one codon
deletion. Branches leading to caviomorph species are thickened. Note that the scale is the same for both phylograms.
242 HUCHON AND DOUZERY

ter two were moderately associated in a clade here
called Phiomorpha sensu stricto; (iv) within Bathyergi-
dae the relationships conformed to Faulkes et al. (1997)
observations based on cytochrome band 12S rRNA;
Bathyergus and Cryptomys clustered together, with
Heliophobius being outside, and Heterocephalus being
in a very distant sister group position; (v) Caviomorpha
contained four major clades; New World porcupines
(Erethizontidae), Cavioidea, Octodontoidea, and an un-
expected association between Chinchilla and Dinomys;
(vi) within Cavioidea, Agouti and Dasyprocta were sis-
ter group to a Cavia plus Hydrochaeris clade; and (vii)
within Octodontoidea, Octodon, Abrocoma, and Cteno-
mys were sister group to a robustly defined clade in-
cluding Capromyidae, Echimyidae, and Myocastori-
dae: the Echimyidae sensu lato.
Comparison of the phylogenies reconstructed with and
without third codon positions reveals a general agree-
ment, with exceptions at two levels (Fig. 1). At the topo-
logical level, the removal of third positions induced sev-
eral shifts: (i) Hystricidae moved from a sister
relationship with caviomorphs to a basalmost position
among hystricognaths; (ii) Erethizontidae moved from a
sister relationship with Cavioidea to one with Octodon-
toidea ⫹Chinchilla ⫹Dinomys; and (iii) Myocastor
moved from a clustering with Proechimys to one with
Echimys. At the node robustness level, the removal of
third codon positions (i) increased the support for Bathy-
ergidae (BP ⫽67 vs 44) and Bathyergidae ⫹Thryono-
myoidea (BP ⫽81 vs 66) and (ii) decreased the support
for Cavioidea (BP ⫽67 vs 100) and Octodontoidea ⫹
Chinchilla ⫹Dinomys (BP ⫽54 vs 77).
All (other) phylogenetic reconstructions—based on
nucleotides with and without Heliophobius sequences
(data not shown), third codon positions, or amino ac-
ids—evidenced the same robust nodes (Table 2). The
best-supported clades (BP ⬎85) were Thryonomy-
oidea, Caviomorpha, Octodontoidea, and Echimyi-
dae ⫹Capromyidae ⫹Myocastoridae, and then Chin-
chilla ⫹Dinomys (99 ⬎BP ⬎79). Few discrepancies
occurred for weakly supported interrelationships, such
as those between the main hystricognath lineages and
the caviomorph superfamilies (Table 2).
Tests of Alternative Hypotheses
The reference topology used for all statistical com-
parisons was rooted by the four sciurognath sequences
(two ctenodactylids, one aplodontid, and one spalacid),
and weakly supported nodes were represented by mul-
tifurcations. Three data matrices were used: nucleo-
tides with or without third positions and amino acids.
Within hystricognaths, phylogenetic analyses iden-
tified seven major clades of family or superfamily rank:
three belonged to Phiomorpha s.l. (Hystricidae, Bathy-
ergidae, and Thryonomyoidea) and four to Caviomor-
pha (Cavioidea, Erethizontidae, Chinchillidae ⫹Dino-
myidae, Octodontoidea). The 10,395 bifurcating
topologies connecting these seven clades were evalu-
ated by ML with TREE-PUZZLE 4.0.1; the TN93 and
gamma rates parameters were set to the values esti-
mated for the best quartet puzzling tree. At the nucle-
otide level with all codon positions, the best alternative
tree showing the paraphyly of Caviomorpha had a log-
likelihood 2.09 standard error (SE) worse (probability
TABLE 2
Robustness Estimators for Representative Nodes of the Hystricognath vWF Tree, Deduced from Three
Data Matrices: All Nucleotide Positions, First and Second Codon Positions, Amino Acids
Nodes
All codon positions Codon positions one and two Amino acids
NJ⌫NJLD MP MPWML NJ⌫NJLD MP MPWML NJ MP MLQ
Petromus ⫹Thryonomys 100 100 100 100 100 100 100 100 100 100 100 99 89
Bathyergidae 36 54 13 28 44 79 85 35 64 67 66 * 44
Phiomorpha s.s.492455 47 66 65 65 74 53 81 51 55 84
Caviomorpha 100 100 99 100 100 97 98 98 93 99 95 97 98
Caviomorpha ⫹Phiomorpha s.s.14*511733 30 25 37 11 39 26 33 80
Caviomorpha ⫹Hystricidae 80 92 36 68 56 53 67 37 27 48 67 47 *
Cavioidea 97 94 95 95 100 64 57 77 16 67 66 81 90
Chinchilla ⫹Dinomys 91 98 90 99 99 82 89 86 85 88 95 79 84
Octodontoidea 100 100 100 100 100 98 98 100 97 98 95 98 94
Echimyidae ⫹Capromys ⫹Myocastor 100 100 100 100 100 92 88 95 95 97 87 93 91
Cavioidea ⫹Erethizontidae 91 86 80 66 87 17 11 29 28 16 * * 9
Chinchilla ⫹Dinomys ⫹Octodontoidea * * 50 62 77 30 29 44 15 54 19 54 10
Chinchilla ⫹Dinomys ⫹Erethizontoidea * * * * * 36 44 11 8 *672473
Note. Bootstrap percentages are obtained from the majority-rule consensus trees. Minority percentages for alternative branchings are
underlined. A star (*) indicates that the node is not supported by the corresponding bootstrap analysis. The different robustness estimators
were obtained after bootstrap resampling with the following phylogenetic reconstructions (from left to right in the table): neighbor-joining
(NJ⌫) on gamma distances (
␣
⫽0.49 or 0.50, with or without third codon positions), NJ on LogDet distances (NJLD), standard maximum-
parsimony (MP), MP weighted (MPW) by the CI ⫻S product (see text for details), maximum-likelihood (ML) with heuristic searches, NJ on
mean character changes (NJ), and ML with quartet puzzling (MLQ).
243SOUTH AMERICA COLONIZATION BY CAVIOMORPH RODENTS

of the one-tailed Kishino–Hasegawa test: PKH ⬍0.02)
than the log-likelihood of the best tree (Fig. 1). Simi-
larly, for codon positions one and two and at the amino
acid level, the disruption of the caviomorph monophyly
involved a dramatic drop in log-likelihood, which was,
respectively, at least 2.29 SE and 2.24 SE worse (PKH ⬍
0.02) than that of the best tree. Altogether, the vWF
data show that the hypothesis of a paraphyly of cavio-
morphs is strongly rejected.
After log-likelihood ranking of the 10,395 topologies,
we independently evaluated ML parameters with
PAML 3.0b for all codon positions on three selected
trees: the best tree (lnL ⫽⫺9605.61, Ti/Tv ⫽3.341,
␣
⫽0.492), the first tree disrupting Caviomorpha
monophyly (lnL ⫽⫺9626.46 [PKH ⫽0.01], Ti/Tv ⫽
3.339,
␣
⫽0.489), and the worst tree (lnL ⫽⫺9668.48
[PKH ⬍0.0001], Ti/Tv ⫽3.366,
␣
⫽0.477). It appeared
a posteriori that variations in parameter estimates
were very slight. Identical results were obtained for
codon positions one and two and amino acids. It is
therefore likely that the confidence probabilities of the
Kishino–Hasegawa comparisons were not affected by
the use of the same ML parameters for different topol-
ogies.
The identification of the phylogenetic relationships
among the three phiomorph clades and the cavio-
morphs was less conclusive. None of the 15 topologies
connecting these four lineages appear to be signifi-
cantly worse than the best tree at the 5% threshold.
The two best competing topologies showed either Hys-
tricidae or Phiomorpha s.s. (Thryonomyidae ⫹Bathy-
ergidae) as the basalmost clade among hystricognaths
(Fig. 1). We note that the Kishino–Hasegawa tests
rank similarly the competing topologies (except for the
best trees), irrespective of the inclusion or exclusion of
third codon positions. In contrast, alternative topolo-
gies are not similarly ranked when amino acids are
used. For this reason no reliable conclusion can be
drawn concerning the order of the first hystricognath
splits.
About the caviomorph relationships, various hypoth-
eses suggested by morphology and paleontology were
also tested. (i) The grouping of Abrocoma and Chin-
chilla (e.g., McKenna and Bell, 1997) was significantly
worse than the reference topology clustering Chin-
chilla and Dinomys (PKH ⬍0.002), whatever the data
matrix considered and whatever the phylogenetic po-
sition of the Abrocoma ⫹Chinchilla clade among the
other caviomorph lineages (i.e., Dinomyidae, Cav-
ioidea, Erethizontoidea, and Octodontoidea). (ii) The
association of Chinchillidae, Dinomyidae, Agoutidae,
and Dasyproctidae in the same clade (e.g., Wood and
Patterson, 1959) is also significantly rejected (PKH ⬍
0.02). (iii) The clustering of Dinomyidae with Erethi-
zontidae (Grand and Eisenberg, 1982) was signifi-
cantly worse than the best tree when all codon posi-
tions were considered (0.002 ⬍PKH ⬍0.03), but
marginally worse with codon positions one and two
(0.04 ⬍PKH ⬍0.09) and with amino acids (0.02 ⬍PKH ⬍
0.09). (iv) The affinity between Dinomyidae and Cav-
ioidea (McKenna and Bell, 1997) yielded a topology
significantly worse (PKH ⬍0.05) than the best one as-
sociating Dinomyidae ⫹Chinchillidae, when all codon
positions were analyzed. When codon positions one and
two or amino acids were considered, all trees disrupt-
ing the Chinchilla ⫹Dinomys clade were not signifi-
cantly different from the best tree (PKH ⬍0.16). (v)
Among Octodontoidea, the 3 possible topologies asso-
ciating Proechimys, Echimys, and Myocastor were not
significantly different (PKH ⬎0.25). This was also the
case for the 15 topologies associating Octodontidae,
Abrocomidae, Ctenomyidae, and Echimyidae s.l.
(PKH ⬎0.06). It was noteworthy that with the nucleo-
tide data sets, the best topologies describing the rela-
tionships among the four main Octodontoidea taxa al-
ways clustered either Ctenomys or Abrocoma with the
Echimyidae s.l. In other words, the vWF analyses sug-
gested a basal emergence of the Octodontidae among
the Octodontoidea.
Finally, the likelihood of the 15 possible trees clus-
tering the four caviomorph lineages were compared.
None of them was significantly better than the others,
but all topologies having a
␦
lnL⬍1 SE (whatever the
characters considered) clustered Octodontoidea with
Chinchillidae ⫹Dinomyidae.
Seeking a Molecular Clock
The highest-likelihood phylograms reconstructed from
codon positions one and two (Fig. 1) and amino acids (best
topology identified by the Kishino–Hasegawa tests) were
taken as a reference, as both displayed the same topol-
ogy. The two slowest-rate species (Aplodontia, Spalax)
and the fastest species (Proechimys; Fig. 1) were first
discarded to maximize the probabilities of obtaining a
clock-like tree. Then, the two-cluster and the branch-
length tests indicated that several branches and taxa
displayed significantly contrasted amino acid vWF
substitution rates: slower for Chinchilla ⫹Dinomys
(PLINTRE ⬍0.01), Trichys ⫹Atherurus (PLINTRE ⫽0.02),
and Heterocephalus (PLINTRE ⫽0.03), and higher for
Echimys (PLINTRE ⬍0.01). Finally, local clocks corre-
sponding to the heterogeneous evolving branches were
enforced with PAML 3.0b.
The hypothesis of five local clocks was accepted for
amino acids, with rates of r1⫽0.41 (Trichys ⫹Athe-
rurus), r2⫽0.55 (Heterocephalus), r3⫽0.34 (Chin-
chilla ⫹Dinomys), r4⫽3.41 (Echimys), and r0⫽1.00
(default rate for the remaining species): lnL ⫽
⫺4090.59 with clock vs ⫺4070.65 without clock (LRT
statistics ⫽39.88, df ⫽27, P⫽0.05). For nucleotides
(codon positions one and two), local clocks were en-
forced for the same taxa, but the clock hypothesis was
244 HUCHON AND DOUZERY

accepted only after the introduction of an additional
rate (r5) for Coendou ⫹Erethizon, i.e., the two slowest-
evolving caviomorphs after Chinchilla ⫹Dinomys (cf.
Fig. 1, right). The following rates were calculated: r0⫽
1.00, r1⫽0.64, r2⫽0.54, r3⫽0.45, r4⫽2.01, and r5⫽
0.56 (lnL ⫽⫺5544.07 with clock vs ⫺5523.75 without
clock; LRT statistics ⫽40.64, df ⫽28, P⫽0.06).
After the local clocks were calibrated by the cavi-
omorph radiation at 31 Myr, close divergence date es-
timates were computed from either amino acids or
codon positions one and two. Except for erethizontoids,
chinchilloids, and octodontoids, amino acids gave older
date estimates, and the deeper the node the greater the
difference of estimation (Fig. 2). According to the vWF
clocks, the first splits in the Hystricognathi tree lead-
ing to the hystricid, phiomorph s.s. lineages, and cavi-
omorphs occurred during Paleocene to Middle Eocene
(63–43 Myr). Bathyergidae split from Thryonomyoidea
in the Middle Eocene (48–41 Myr). From 43 Myr
(younger estimate) to 31 Myr, Caviomorpha did not
produce lineages that are still living. Actually, their
first diversification occurred after colonization of South
America, during the Late Oligocene. Octodontoid radi-
ated at the Early/Middle Miocene transition. Closer
genera—such as Coendou and Erethizon or Myocastor
and Echimys—separated during Late Miocene or Plio–
Pleistocene.
DISCUSSION
A Single Colonization Event of South America
by Caviomorphs
The sampling of all living caviomorph and phio-
morph families warranted an exhaustive consideration
of Caviomorpha monophyly. The vWF phylogeny
strongly supports the monophyly of Caviomorpha, and
all alternatives are significantly less likely. This indi-
cates that all living caviomorphs have a single origin,
therefore reflecting a single colonization event of South
America by hystricognath rodents. This molecular re-
sult contradicts the view of a reciprocal paraphyly of
caviomorphs and phiomorphs, and of a double invasion
of hystricognaths into South America, suggested for
example by myology (Woods and Hermanson, 1985),
arterial patterns (Bugge, 1985), cranial characters
(Bryant and McKenna, 1995), and parasitology (Quen-
tin, 1973; Hugot, 1982). One should note that these
studies, supporting caviomorph paraphyly, are not mu-
tually congruent. Studies on muscles, arteries, skulls,
and teeth all suggest that Erethizontidae might be one
of the earliest (or the earliest) hystricognath lineages
(Woods and Hermanson, 1985; Bryant and McKenna,
1995). However, parasitology gives a different conclu-
sion because Dinomyidae, Erethizontidae, Hystricidae,
and the sciurognath Pedetidae share closely related
endoparasites (Hugot, 1982).
Subsequent Adaptative Radiations of Caviomorphs
in the New World
Caviomorph invaders likely replaced South Ameri-
can endemic species of the Paleocene and Eocene in
their ecological niches (Flynn and Wyss, 1998). Their
arrival has been thought to have caused the extinction
of small mammals, such as rodent-like marsupials. The
success of caviomorphs in the South American environ-
ment led to their diversification into four extant groups
that are identified since the Early Oligocene: erethi-
FIG. 2. Local clocks in the vWF maximum-likelihood tree show-
ing the time frame for the radiation of the main hystricognath
families. Divergence ages in million years deduced from amino acids
and codon positions one and two are, respectively, given above and
below branches or at the left and right of slashes. The white circle
indicates the calibration point: diversification of caviomorph lineages
at 31 Myr. The ML topology is deduced from analysis of amino acids
and nucleotides, and branch lengths are computed after the amino
acids matrix (scale: number of substitutions per site). The thickness
of branches is proportional to the amino acid local clock rates: r1⫽
0.41, r2⫽0.55, r3⫽0.34, r4⫽3.41 (see the text for details). The
vertical dashed lines indicate the time frame for the remaining
species, all evolving with the default local clock (r0⫽1.00). The
absolute ages (Ma) and the name of the main Tertiary divisions are
given (epochs: PAL, Paleocene; EOC, Eocene; OLI, Oligocene; MIO,
Miocene; PP, Plio–Pleistocene).
245SOUTH AMERICA COLONIZATION BY CAVIOMORPH RODENTS

zontoids, cavioids, chinchilloids, and octodontoids (Fig.
1). Disagreements between morphopaleontological sys-
tematics (e.g., McKenna and Bell, 1997; Table 1) and
molecular data (Fig. 1) involve the content of these four
groups.
Chinchilloidea (Dinomyidae ⫹Chinchillidae). The
systematic position of Dinomyidae has been intensely
debated. Our vWF results suggest the clustering of
Dinomyidae (pacaranas) with Chinchillidae (chinchil-
las) to form the Chinchilloidea. This rejects the possi-
bility of a sister group relationship between Dinomyi-
dae and Erethizontidae and indicates that the fusion of
the second and third cervical vertebrae shared by Di-
nomys, Erethizon, and Coendou (Ray, 1958; cited in
Woods and Hermanson, 1985, p. 533) likely represents
a convergence, as for their many other morphological
similarities (reviewed in Grand and Eisenberg, 1982).
Likewise, the fact that Dinomyidae and Erethizontidae
share the same pinworm parasites (Nematoda: Wellco-
mia; Quentin, 1973; Hugot, 1982) can be explained by
a horizontal transfer. For example, Patterson and
Wood (1982, p. 474) indicate that the pinworm of Er-
ethizon also infects the domestic cat.
Dinomyidae has also been considered to be related to
Cavioidea or even included within them (McKenna and
Bell, 1997). Wood and Patterson (1959) identified sim-
ilarities in the dental patterns of Dinomyidae, Agouti-
dae, and Dasyproctidae together with Chinchillidae.
Later, Patterson and Wood (1982) estimated that the
three former families should in fact be the sister clade
of Caviidae ⫹Hydrochaeridae. In the present study, all
alternative phylogenetic positions of Dinomyidae were
significantly less likely than those of the best tree when
the three codon positions were analyzed, but not when
the two first codon positions alone or the amino acids
were considered. Consequently, even if the vWF
strongly supports the unexpected relationship of Chin-
chillidae with Dinomyidae, it will need to be confirmed
by additional molecular analyses based on genetically
independent genes.
The first Dinomyidae fossils were discovered in the
“Mayoan” Middle Miocene (10–12 Myr; Vucetich et al.,
1999). This paleontological date is younger than our
molecular estimation of a Chinchilla–Dinomys split
between 17 and 21 Myr (Fig. 2), therefore suggesting
that there is a gap in the Dinomyidae fossil record.
The second new phylogenetic result suggested by
vWF sequence comparisons is the strong rejection of a
relation between Chinchillidae and Abrocomidae (e.g.,
McKenna and Bell, 1997). We rather suggest the inclu-
sion of Abrocomidae among Octodontoidea as previ-
ously suggested by dental characters (Martin, 1994).
Octodontoidea. Octodontoidea is the most diversi-
fied caviomorph clade, with a first radiation having
produced the three extant families Octodontidae, Abro-
comidae, and Ctenomyidae and a fourth lineage, which
subsequently diversified into Capromyidae, Myocasto-
ridae, and Echimyidae (Fig. 1). The existence of the
latter subclade is well defined by vWF (Table 2) and
supported by dental and myological studies (Woods
and Hermanson, 1985). Myocastoridae has been con-
sidered either an independent family or part of the
Capromyidae or an Echimyidae subfamily (e.g., Patter-
son and Wood, 1982; Wilson and Reeder, 1993; Mc-
Kenna and Bell, 1997). In conjuction with immunolog-
ical data (Sarich, 1985), the vWF results suggest the
merging of Myocastoridae within Echimyidae (Fig. 1).
Similarly, Capromyidae has been thought to be an
Echimyidae subfamily (e.g., Patterson and Wood,
1982). The vWF trees here suggest a distinct position
for Capromys, possibly at the family level (Fig. 1).
Based on the local clock analysis, the split between
Capromyidae and Echimyidae ⫹Myocastoridae is es-
timated to be 7–10 Myr old (Fig. 2). These dates are
younger than the paleontological record, which gives
an Early Miocene age for the first Myocastoridae
(21–19 Myr) and Capromyidae (19–16.3 Myr) fossils
(McKenna and Bell, 1997). This indicates that the mo-
lecular clock might not click regularly along this lin-
eage (cf. the long Echimys branch) and/or that the
taxonomic position of fossils may need to be reexam-
ined (at least for the first Myocastoridae). Anyway, the
close molecular relationship among these species and
the fact that echimyid fossils are known since Late
Oligocene (29–24 Myr) suggest that Echimyidae, Myo-
castoridae, and Capromyidae are part of the same tax-
onomic group, which should be further sampled in fu-
ture phylogenetic studies.
Ctenomyidae was considered either an independent
family or an Octodontidae subfamily (Pascual et al.,
1965; cited in Lessa and Cook, 1998). In agreement
with Nedbal et al. (1994) and Lessa and Cook (1998),
our results support the ranking of ctenomyids at the
familial level (Fig. 1). Furthermore, Abrocomidae
(chinchilla rats) might be more related to Octodontidae
than to Ctenomyidae (analysis of all codon positions).
Abrocoma is a chinchilla-like animal, living in the Cor-
dillera from south Peru and Bolivia to north Argentina
and Chile. Only a few morphological studies have ad-
dressed the issue of Abrocomidae origins (e.g., Martin,
1994), and we note that this family was not investi-
gated in Luckett and Hartenberger (1985). We here
include for the first time one Abrocoma sequence in a
molecular analysis. We suggest that Abrocoma is one of
the main lineages produced by the Octodontoidea ra-
diation. Consequently, it appears important that fu-
ture evolutionary studies involving octodontids also
include Abrocomidae representatives.
The association of Abrocomidae with Octodontidae is
surprising because the latter family shares various
synapomorphies with Ctenomyidae, such as highly de-
rived kidney-shaped molars. One should note that the
grouping of Octodontidae with Ctenomyidae is not a
246 HUCHON AND DOUZERY

significantly less likely alternative than groupings in
the best tree and that all taxa at the base of the
Octodontoidea radiation are slowly evolving, most par-
ticularly the Octodontidae. We cannot exclude the pos-
sibility that the topology within Octodontoidea is the
result of a long-branch attraction phenomenon, clus-
tering Ctenomyidae with the fast-evolving Echimyidae
s.l. Additional sampling within these families might
help to resolve this issue.
Cavioidea. The molecular trees robustly cluster
Caviidae, Hydrochaeridae, Dasyproctidae, and Agouti-
dae to form the Cavioidea clade (Fig. 1, all codon posi-
tions). This is in agreement with the morphology-based
systematics, except for the Dinomyidae (see above).
Among the former four families, Caviidae and Hydro-
chaeridae appear to be a sister group, as suggested by
morphological studies (Wood and Patterson, 1959).
Dasyproctidae has been regarded an Agoutidae sub-
family (see comments in Wilson and Reeder, 1993;
McKenna and Bell, 1997). The vWF analysis suggests
the ranking of Dasyproctidae at the family level as this
taxon does not cluster with Agoutidae. However, alter-
native hypotheses are not significantly less likely, in-
dicating that longer nuclear sequences are required to
reach higher confidence levels.
Erethizontoidea. It should be noted that South
American porcupines are not the most divergent hys-
tricognaths as suggested by morphological studies
(Bryant and McKenna, 1995; Lavocat and Parent,
1985). Erethizontoids are part of the caviomorph radi-
ation, but additional molecular studies should be con-
ducted to evaluate whether they diverged slightly be-
fore the other caviomorph superfamilies as suggested
by Bugge (1985).
Finally, the pattern of radiation of the four major
caviomorph clades—Chinchilloidea, Octodontoidea,
Cavioidea, and Erethizontoidea—is difficult to estab-
lish because of the lack of robust resolution and con-
flicting branching order between topologies recon-
structed with and without third codon positions (Fig.
1). An association between Chinchilloidea and Oct-
odontoidea is, however, suggested at the DNA level,
with greater support when third codon positions are
included (Fig. 1).
Timing of the Caviomorpha Radiations
The Caviomorpha subtree presents three ambiguous
branching orders, representing either a lack of resolu-
tion of the vWF and/or different radiation events. The
three branching points are those of the Caviomorpha
superfamilies, the Octodontoidea families, and the
Echimyidae genera (including Myocastor). The two
former events might be related to two ecological events.
(i) The arrival of caviomorphs in South America corre-
sponds to a cooling period (36–25 Myr) in which indig-
enous species developed hypsodont teeth as an adap-
tative response to environmental modifications (Kay et
al., 1999). The first caviomorph fossils are high
crowned (i.e., hypsodont; Wyss et al., 1993) and many
genera from the Deseadan display a high degree of
hypsodonty (Vucetich et al., 1999). Climatic changes
may have been responsible for the success of the cavi-
omorphs and their fast diversification. (ii) Important
climatic changes contemporary to the Quechua phase
of the Andean orogeny have been described during
Middle Miocene (e.g., Vucetich et al., 1999). Such cli-
matic perturbations may have allowed for the occur-
rence of species adapted to new ecological conditions, in
association with environmental barriers preventing
north–south intermigration that have been described
at the same time (16–11 Myr) (Walton, 1997). Our
datings suggest 13–18 Myr for the origin of modern
octodontoids (Fig. 2), in agreement with Vucetich et al.
(1999), who correlated the diversification of Octodon-
toidea with the Middle Miocene climatic changes.
What Does vWF Tell Us about Phiomorpha
Phylogenetics and Evolution?
The vWF data confirm that the two African families
Petromuridae and Thryonomyidae are closely related
and justify their grouping into the Thryonomyoidea
superfamily, as previously indicated by fossil and mi-
tochondrial data (Lavocat, 1973; Nedbal et al., 1994).
The sister group of Dassie rats and cane rats is shown
to be the Bathyergidae, as previously suggested by
Lavocat (1973) and Nedbal et al. (1994). One should
note that the association of Bathyergidae and Thryono-
myoidea suggests an African but not an Asian origin of
Bathyergidae (Winkler, 1994).
Highest-likelihood phylograms suggest that Phio-
morpha s.l. contains two major clades (Hystricidae,
Bathyergidae ⫹Thryonomyoidea) whose relationships
with caviomorphs are sensitive to the molecular char-
acters considered (Fig. 1). Consequently, it would be
better to restrict the use of the term “Phiomorpha” to
the Thryonomyoidea ⫹Bathyergidae clade, excluding
the Hystricidae. The latter family is a puzzling taxon
with an enigmatic origin. Hystricidae fossils are known
only since the Miocene, when they appear simulta-
neously in Asia, Europe, and Africa (McKenna and
Bell, 1997). vWF sequences indicated that hystricids
are part of the basal hystricognath radiation, suggest-
ing that the origin of the group is much older than the
Miocene and is possibly of Eocene age.
All the molecular dates for the Phiomorpha s.s. splits
(Fig. 2) appear to be older than the relevant dates in
the fossil record. The vWF sequences propose (1) an
Early to Middle Eocene age for the Thryonomyoidea/
Bathyergidae divergence vs a Late Eocene–Early Oli-
gocene paleontological record for the first hystricog-
nath, (2) an Eocene age for the first Bathyergidae split
(but this might reflect the deep clustering of Hetero-
cephalus with the remaining Bathyergidae) with first
247
SOUTH AMERICA COLONIZATION BY CAVIOMORPH RODENTS

Bathyergidae fossils known only since the Early Mio-
cene, and (3) a Petromuridae–Thryonomyidae split
during the Oligocene vs a Pleistocene age for the first
fossil Petromuridae. Gaps in the fossil record can be
invoked to explain these discrepancies. However, the
fast-evolving Bathyergidae and Thryonomyoidea se-
quences (cf. Huchon et al., 2000) might have increased
the depth of the split between the two groups and
subsequently affected the divergence date estimations.
The Potential Colonization Routes of Caviomorphs
to South America
Based on paleontological data, an Asian origin for
hystricognaths is suggested (Flynn et al., 1986; Bryant
and McKenna, 1995). In sharp contrast, however, the
migration patterns to explain the current distribution
of most phiomorphs in Africa and caviomorphs in
South America remain moot. The current consensus
(Lavocat, 1969; Martin, 1994) proposed that cavi-
omorphs originated from an African phiomorph stock
and directly migrated to South America by rafting over
the Atlantic ocean (Fig. 3, route 1). Despite the fact
that Africa and South America were separated by the
Atlantic oceanic barrier (i.e., at least 1000 km) when
colonization took place, the probability of a successful
colonization event might have been increased by the
existence of marine currents, paleowinds, “stepping
stone” islands, and rafts carried by tropical rivers, com-
bined with dramatic climatic and oceanographic
changes at the Eocene/Oligocene transition (Wyss et
al., 1993; Flynn and Wyss, 1998; Houle, 1999).
Our molecular data indicate an almost contemporary
origin for hystricognaths (46–63 Myr) and cavio-
morphs (43–54 Myr). Given the probable Asian origin
of hystricognaths and the short time span between the
origin of hystricids and that of caviomorphs (cf. Fig. 2),
our data suggest an Asian rather than an African ori-
gin for the Caviomorpha. This hypothesis has already
been suggested by paleontology. Hussain et al. (1978)
proposed an Asian origin for hystricognaths, with sub-
sequent colonization(s) in South America through
North America for caviomorphs (Fig. 3, Route 2) and
migration in Africa for phiomorphs. The drawback of
this scenario is the lack of obvious hystricognath fossils
in North America (Meng, 1990; Martin, 1994). Given
this, we cannot exclude the possibility that Caviomor-
pha might also have originated in Asia (cf. Hussain et
al., 1978) but migrated through Africa (Fig. 3, Route 3)
to South America (cf. Lavocat, 1969). This would in-
volve the paraphyly of African hystricognath fossil
taxa. Interesting to note, another alternative, namely a
FIG. 3. Four hypotheses for the migration routes of the hystricognath rodents during the Eocene (55–34 Myr). The vWF phylogenetic tree
for amino acids and codon positions one and two was superimposed on a Middle Eocene tectonic plate distribution map (after Smith et al.,
1994) according to the localization of the fossils of each family at the end of the Eocene. Phiomorpha s.l. rodents are italicized. The solid arrow
lines indicate the currently recognized routes of migration. The dashed lines suggest four hypothetical routes of migration of the caviomorphs.
Route 1: Caviomorpha originated from Africa and directly crossed the Atlantic ocean to reach South America (Lavocat, 1969). Route 2:
Caviomorpha originated from Asia (e.g., Hussain et al., 1978) and migrated through North America (cf. Wood, 1985). Route 3: Caviomorpha
originated from Asia (cf. Hussain et al., 1978) and migrated to South America through Africa (cf. Lavocat, 1969). Route 4: Caviomorpha
originated from Asia and migrated to South America through Australia and Antarctica, after two ocean crossings (cf. Houle, 1999). One
should note that the possibility of migration from Africa to Antarctica and then to South America has never been proposed. These migration
routes are drawn to explain the current distribution of living phiomorphs and caviomorphs in Africa and South America (bold outlines). Note
that some living hystricids and erethizontids are also distributed, respectively, in Africa and North America.
248 HUCHON AND DOUZERY

southern migration route for eutherian mammals from
Asia to South America, through Australia and Antarc-
tica, has never been extensively investigated (Fig. 3,
Route 4). Until the Late Eocene (41–34 Myr), ecological
conditions in both Australia and Antarctica were fa-
vorable for placental fauna, and Antarctica was con-
nected to South America (Houle, 1999; Kay et al.,
1999). To date, no rodent fossils have been found in
Australia or Antarctica, but the mammalian fossil
record is scarce for the Early and Middle Eocene for
these continents. All Eocene Australian fossils come
from the Murgon faunal zone, which has a minimum
age estimate of 54.6 Myr (Godthelp et al., 1992), and all
Tertiary mammalian fossils found in Antarctica date to
the Late Eocene deposits at Seymour Island (Kay et al.,
1999). Our data indicate a minimum time lag of 12 Myr
(43 to 31 Myr; Fig. 2) between the divergence of cavi-
omorphs relative to phiomorphs and their subsequent
South American diversification. We therefore suggest
that Caviomorpha might have originated in Asia and
followed an Australian–Antarctic migration route to
reach South America (Fig. 3, Route 4) during the 12
Myr in the Middle and Late Eocene (Fig. 2).
CONCLUSIONS
Nedbal et al. (1994) concluded from their mitochon-
drial 12S rRNA analysis that the rapid radiation of the
caviomorphs, the number of taxa analyzed, and the
long branches in the ingroup might explain the lack of
resolution for some caviomorph relationships. In their
caviomorph subtrees including 10 families, a single
node (Octodontoidea) was supported by more than 60%
of bootstrap. Our improvements in the reconstruction
of Caviomorpha relationships indicate that, despite all
these constraints, the use of a nuclear DNA marker,
here the vWF, provides molecular signal pertaining to
the phylogeny of this group.
Our main results involve (1) the division of Hystri-
cognathi into three clades, Hystricidae, Phiomorpha
s.s., and Caviomorpha; (2) the validation at the molec-
ular level of a four-clade division within Caviomorpha;
(3) the grouping of Dinomyidae with Chinchillidae; (4)
the recognition of the monophyly of Echimyidae ⫹Ca-
promyidae ⫹Myocastoridae; and (5) the suggestion
that Caviomorpha might not have an African but
rather has an Asian origin. These results will need to
be validated by additional independent molecular data
and by paleontological and morphological observa-
tions. The study of the phylogeny of Primates, with
special focus on Platyrrhini and Catarrhini, appears
complementary to better understand the biogeograph-
ical relationships between South America and the
other land masses during the Tertiary.
ACKNOWLEDGMENTS
This work would not have been possible without the essential
contribution of Franc¸ois Catzeflis (curator of the collection of Mont-
pellier) and of all tissue collectors: Heinrich Burger and Anna Kueb-
ber-Heiss (Zoo of Viena, Austria), Luis Contreras, Chris G. Faulkes,
John A. W. Kirsh, Eviatar Nevo, James L. Patton, Francis Petter,
Benoit de Sousa, and Jean-Christophe Vie´. D.H. thanks Tammie L.
Bettinger, Christopher J. Bonar and the Cleveland Metroparks Zoo,
and Reg Hoyt and the Zoological Society of Philadelphia for their
sample gifts. We thank Franc¸ois Catzeflis for laboratory support,
Ste´phane Ducrocq, Jean-Louis Hartenberger, Jean-Jacques Jaeger,
Laurent Marivaux, Bettine Jansen van Vuuren, and two anonymous
reviewers for useful comments and paleontological discussions, and
Ziheng Yang for advice on BASEML and CODEML programs of the
PAML package. This work has been supported by ACC-SV7 (Re´seau
National de Biosyste´matique), ACC-SV3 (Re´seau coordonne´ par D.
Mouchiroud), and European Community TMR Network “Mamma-
lian phylogeny” FMRX-CT98-0221. D.H. acknowledges the financial
support of a M.E.N.E.S.R. grant (No. 97132). This is contribution No.
2001-026 of the Institut des Sciences de l’Evolution de Montpellier
(UMR 5554-CNRS).
REFERENCES
Adachi, J., and Hasegawa, M. (1996). MOLPHY 2.3: Programs for
molecular phylogenetics based on maximum likelihood. Comput.
Sci. Monogr. 28: 1–150.
Bryant, J. D., and McKenna, M. C. (1995). Cranial anatomy and
phylogenetic position of Tsaganomys altaicus (Mammalia: Roden-
tia) from the Hsanda Gol Formation (Oligocene), Mongolia. Am.
Mus. Novit. 3156: 1–42.
Bugge, J. (1985). Systematic value of the carotid arterial pattern in
rodents. In “Evolutionary Relationships among Rodents: A Multi-
disciplinary Analysis” (W. P. Luckett and J.-L. Hartenberger,
Eds.), pp. 355–379. Plenum, New York.
Catzeflis, F. M. (1991). Animal tissue collections for molecular ge-
netics and systematics. Trends Ecol. Evol. 6: 168.
Catzeflis, F. M. Ha¨nni, C., Sourrouille, P., and Douzery, E. (1995).
Molecular systematics of hystricognath rodents: The contribution
of sciurognath mitochondrial 12S rRNA sequences. Mol. Phylo-
genet. Evol. 4: 357–360.
Faulkes, C. G., Bennett, N. C., Bruford, M. W., O’Brien, H. P.,
Aguilar, G. H., and Jarvis, J. U. M. (1997). Ecological constraints
drive social evolution in the African mole-rats. Proc. R. Soc. Lond.
B264: 1619–1627.
Felsenstein, J. (1988). Phylogenies from molecular sequences: Infer-
ence and reliability. Annu. Rev. Genet. 22: 521–565.
Flynn, L. J., Jacobs, L. L., and Cheema, I. U. (1986). Baluchimyinae,
a new ctenodactyloid rodent subfamily from the Miocene of Ba-
luchistan. Am. Mus. Novit. 2841: 1–58.
Flynn, J. J., and Wyss, A. R. (1998). Recent advances in South
American mammalian paleontology. Trends Ecol. Evol. 13: 449–
454.
Godthelp, H., Archer, M., Cifelli, R., Hand, S. J., and Gilkeson, C. F.
(1992). Earliest known Australian Tertiary mammal fauna. Na-
ture 356: 514–516.
Grand, T. I., and Eisenberg, J. F. (1982). On the affinities of the
Dinomyidae. Sa¨uget. Mitteil. 30: 151–157.
Hartenberger, J.-L. (1985). The order Rodentia: Major questions on
their evolutionary origin, relationships and suprafamilial system-
atics. In “Evolutionary Relationships among Rodents: A Multidis-
ciplinary Analysis” (W. P. Luckett and J.-L. Hartenberger, Eds.),
pp. 1–33. Plenum, New York.
Hartenberger, J.-L. (1998). Description de la radiation des Rodentia
249SOUTH AMERICA COLONIZATION BY CAVIOMORPH RODENTS

(Mammalia) du Pale´oce`ne supe´rieur au Mioce`ne: Incidences phy-
loge´ne´tiques. C. R. Acad. Sci. Paris Sci. Terre Plane`tes 326: 439–
444.
Hassanin, A., Lecointre, G., and Tillier, S. (1998a). The “evolutionary
signal” of homoplasy in protein-coding gene sequences and its
consequences for a priori weighting in phylogeny. C. R. Acad. Sci.
Paris Life Sci. 321: 611–620.
Hassanin, A., Pasquet, E., and Vigne, J.-D. (1998b). Molecular sys-
tematics of the subfamily Caprinae (Artiodactyla, Bovidae) as de-
termined from cytochrome bsequences. J. Mammal. Evol. 5: 217–
236.
Hassanin, A., and Douzery, E. J. P. (1999). The tribal radiation of the
family Bovidae (Artiodactyla) and the evolution of the mitochon-
drial cytochrome bgene. Mol. Phylogenet. Evol. 13: 227–243.
Houle, A. (1999). The origin of platyrrhines: An evaluation of the
antarctic scenario and the floating island model. Am. J. Phys.
Anthropol. 109: 541–559.
Huchon, D., Catzeflis, F. M., and Douzery, E. J. P. (1999). Molecular
evolution of the nuclear von Willebrand Factor gene in mammals
and the phylogeny of rodents. Mol. Biol. Evol. 16: 577–589.
Huchon, D., Catzeflis, F., and Douzery, E. J. P. (2000). Variance of
molecular datings, evolution of rodents and the phylogenetic affin-
ities between Ctenodactylidae and Hystricognathi. Proc. R. Soc.
Lond. B 267: 393–402.
Hugot, J.-P. (1982). Sur le genre Wellcomia (Oxyuridae, Nematoda),
parasite de Rongeurs archaı¨ques. Bull. Mus. Natl. Hist. Nat. Paris.
4e´me Se´r. Sect. A 4: 25–48.
Hussain, S. T., de Bruijn, H., and Leinders, J. M. (1978). Middle
Eocene rodents from the Kala Chitta Range (Pujab, Pakistan) (III).
Proc. Kon. Ned. Akad. Wetensch. Ser. B 81: 101–112.
Jaeger, J.-J. (1988). Rodent phylogeny: New data and old problems.
In “The Phylogeny and Classification of the Tetrapods” (M. J.
Benton, Ed.), pp. 177–199. Clarendon, Oxford.
Kay, R. F., Madden, R. H., Vucetich, M. G., Carlini, A. A., Mazzoni,
M. M., Guillermo, H. R., Heizler, M., and Sandeman, H. (1999).
Revised geochronology of the Casamayoran South American Land
Mammal Age: Climatic and biotic implications. Proc. Natl. Acad.
Sci. USA 96: 13235–13240.
Kishino, H., and Hasegawa, M. (1989). Evaluation of the maximum
likelihood estimate of the evolutionary tree topologies from DNA
sequence data, and the branching order in Hominoidea. J. Mol.
Evol. 29: 170–179.
Kumar, S., and Hedges, S. B. (1998). A molecular timescale for
vertebrate evolution. Nature 392: 917–920.
Landry, S. O. J. (1999). A proposal for a new classification and
nomenclature for the glires (Lagomorpha and Rodentia). Mitt.
Mus. Nat. Kd. Berl. Zool. Reihe 75: 283–316.
Lavocat, R. (1969). La syste´matique des rongeurs hystricomorphes et
la de´rive des continents. C. R. Acad. Sci. Paris Se´r. D 269: 1496–
1497.
Lavocat, R. (1973). Les rongeurs du Mioce`ne d’Afrique Orientale. I.
Mioce`ne infe´rieur. Me´m. Trav. Inst. Montpellier Ecole Pratique
Hautes Etudes 1: 1–284.
Lavocat, R., and Parent, J.-P. (1985). Phylogenetic analysis of middle
ear features in fossil and living rodents. In “Evolutionary Rela-
tionships among Rodents: A Multidisciplinary Analysis” (W. P.
Luckett and J.-L. Hartenberger, Eds.), pp. 333–354. Plenum, New
York.
Lessa, E. P., and Cook, J. A. (1998). The molecular phylogenetics of
tucos-tucos (genus Ctenomys, Rodentia: Octodontidae) suggests an
early burst of speciation. Mol. Phylogenet. Evol. 9: 88–99.
Lockhart, P. J., Steele, M. A., Hendy, M. D., and Penny, D. (1994).
Recovering evolutionary distances under a more realistic model of
sequence evolution. Mol. Biol. Evol. 11: 605–612.
Luckett, W. P., and Hartenberger, J.-L. (Eds.). (1985). “Evolutionary
Relationships among Rodents: A Multidisciplinary Analysis,” Ple-
num, New York.
Luckett, W. P., and Hartenberger, J.-L. (1993). Monophyly or
polyphyly of the order Rodentia: Possible conflict between morpho-
logical and molecular interpretations. J. Mammal. Evol. 1: 127–
147.
Martin, T. (1994). African origin of caviomorph rodents is indicated
by incisor enamel microstructure. Paleobiology 20: 5–13.
McKenna, M. C., and Bell, S. K. (1997). “Classification of Mammals
above the Species Level,” Columbia Univ. Press, New York.
Meng, J. (1990). The auditory region of Reithroparamys delicatissi-
mus (Mammalia, Rodentia) and its systematic implications. Am.
Mus. Novit. 2972: 1–35.
Nedbal, M. A., Allard, M. W., and Honeycutt, R. L. (1994). Molecular
systematics of hystricognath rodents: Evidence from the mitochon-
drial 12S rRNA gene. Mol. Phylogenet. Evol. 3: 206–220.
Nedbal, M. A., Honeycutt, R. L., and Schlitter, D. A. (1996). Higher-
level systematics of rodents (Mammalia, Rodentia): Evidence from
the mitochondrial 12S rRNA gene. J. Mammal. Evol. 3: 201–237.
Patterson, B., and Wood, A. E. (1982). Rodents from the Deseadan
Oligocene of Bolivia and the relationships of the Caviomorpha.
Bull. Mus. Comp. Zool. 149: 371–543.
Philippe, H. (1993). MUST: A computer package of management
utilities for sequences and trees. Nucleic Acids Res. 21: 5264–
5272.
Quentin, J.-C. (1973). Affinite´s entre les Oxyures parasites de ron-
geurs Hystricide´s, Erethizontide´s et Dinomyide´s: Inte´reˆt pale´obio-
ge´ographique. C. R. Acad. Sci. Paris Se´r. D 276: 2015–2017.
Robinson, M., Gouy, M., Gautier, C., and Mouchiroud, D. (1998).
Sensitivity of the relative-rate test to taxonomic sampling. Mol.
Biol. Evol. 15: 1091–1098.
Sarich, V. M. (1985). Rodent macromolecular systematics. In “Evo-
lutionary Relationships among Rodents: A Multidisciplinary Anal-
ysis” (W. P. Luckett and J.-L. Hartenberger, Eds.), pp. 423–452.
Plenum, New York.
Smith, A. G., Smith, D. G., and Funnell, B. M. (1994). “Atlas of
Mesozoic and Cenozoic Coastlines,” Cambridge Univ. Press, Cam-
bridge, UK.
Strimmer, K., and von Haeseler, A. (1996). Quartet puzzling: A
quarter maximum-likelihood method for reconstructing tree topol-
ogies. Mol. Biol. Evol. 13: 964–969.
Sullivan, J., and Swofford, D. L. (1997). Are guinea pigs rodents? The
importance of adequate models in molecular phylogenetics. J.
Mammal. Evol. 4: 77–86.
Swofford, D. L. (1998). PAUP*. Phylogenetic Analysis Using Parsi-
mony (* and Other Methods). Version 4, Sinauer, Sunderland, MA.
Takezaki, N., Rzhetsky, A., and Nei, M. (1995). Phylogenetic test of
the molecular clock and linearized trees. Mol. Biol. Evol. 12: 823–
833.
Vucetich, M. G., Verzi, D. H., and Hartenberger, J.-L. (1999). Review
and analysis of the radiation of the South American Hystricog-
nathi (Mammalia, Rodentia). C. R. Acad. Sci. Paris Earth Planet.
Sci. 329: 763–769.
Walton, A. H. (1997). Rodents. In “Vertebrate Paleontology in the
Neotropics: The Miocene Fauna of La Venta, Columbia” (R. F. Kay,
R. H. Madden, R. L. Ciffelli, and J. J. Flynn, Eds.), pp. 392–409.
Smithsonian Institution Press, Washington, DC.
Wilson, D. E., and Reeder, D. M. (1993). “Mammal Species of the
World: A Taxonomic and Geographic Reference,” Smithsonian In-
stitution Press, Washington, DC.
Winkler, A. J. (1994). The middle/upper Miocene dispersal of major
rodent groups between southern Asia and Africa. In “Rodent and
Lagomorph Families of Asian Origins and Diversification” (Y. To-
250 HUCHON AND DOUZERY

mida, C.-K. Li, and T. Setoguchi, Eds.), pp. 173–184. National
Science Museum Monogr., Tokyo.
Wood, A. E. (1955). A revised classification of the rodents. J. Mam-
mal. 36: 165–187.
Wood, A. E., and Patterson, B. (1959). The rodents of the Deseadan
Oligocene of Patagonia and the beginnings of South American
rodent evolution. Bull. Mus. Comp. Zool. 120: 279–428.
Wood, A. E. (1985). The relationships, origin and dispersal of the
hystricognathous rodents. In “Evolutionary Relationships among
Rodents: A Multidisciplinary Analysis” (W. P. Luckett and J.-L.
Hartenberger, Eds.), pp. 475–513. Plenum, New York.
Woods, C. A., and Hermanson, J. W. (1985). Myology of hystricog-
nath rodents: An analysis of form, function, and phylogeny. In
“Evolutionary Relationships among Rodents: A Multidisciplinary
Analysis” (W. P. Luckett and J.-L. Hartenberger, Eds.), pp. 515–
548. Plenum, New York.
Wyss, A. R., Flynn, J. J., Norell, M. A., Swisher, C. C., III, Charrier,
R., Novacek, M. J., and McKenna, M. C. (1993). South America’s
earliest rodent and recognition of a new interval of mammalian
evolution. Nature 365: 434–437.
Yang, Z. (1996). Among-site rate variation and its impact on phylo-
genetic analyses. Trends Ecol. Evol. 11: 367–372.
Yang, Z. (2000). Phylogenetic Analysis by Maximum Likelihood
(PAML), Version 3.0b. University College London.
Yoder, A. D., and Yang, Z. (2000). Estimation of primate speciation
dates using local molecular clocks. Mol. Biol. Evol. 17: 1081–1190.
251SOUTH AMERICA COLONIZATION BY CAVIOMORPH RODENTS