Content uploaded by Anne Ropiquet
Author content
All content in this area was uploaded by Anne Ropiquet on Sep 03, 2018
Content may be subject to copyright.
Molecular Phylogenetics and Evolution 40 (2006) 101–117
www.elsevier.com/locate/ympev
1055-7903/$ - see front matter © 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.ympev.2006.02.017
Mitochondrial and nuclear phylogenies of Cervidae (Mammalia,
Ruminantia): Systematics, morphology, and biogeography
Clément Gilbert a,b,c, Anne Ropiquet a,b, Alexandre Hassanin a,b,¤
a UMR 5202—Origine, Structure et Evolution de la Biodiversité, Département Systématique et Evolution, Muséum National d’Histoire Naturelle,
Case postale N° 51, 55, rue BuVon, 75005 Paris, France
b Service de Systématique Moléculaire, Département Systématique et Evolution, Muséum National d’Histoire Naturelle, 43, rue Cuvier, 75005 Paris, France
c Evolutionary Genomics Group, Department of Botany and Zoology, University of Stellenbosch, Private Bag X1, Matieland 7602, South Africa
Received 14 November 2005; revised 2 February 2006; accepted 19 February 2006
Available online 3 April 2006
Abstract
The family Cervidae includes 40 species of deer distributed throughout the northern hemisphere, as well as in South America and South-
east Asia. Here, we examine the phylogeny of this family by analyzing two mitochondrial protein-coding genes and two nuclear introns for 25
species of deer representing most of the taxonomic diversity of the family. Our results provide strong support for intergeneric relationships.
To reconcile taxonomy and phylogeny, we propose a new classiWcation where the family Cervidae is divided in two subfamilies and Wve tribes.
The subfamily Cervinae is composed of two tribes: the tribe Cervini groups the genera Cervus, Axis, Dama, and Rucervus, with the Père
David’s deer (Elaphurus davidianus) included in the genus Cervus, and the swamp deer (Cervus duvauceli) placed in the genus Rucervus; the
tribe Muntiacini contains Muntiacus and Elaphodus. The subfamily Capreolinae consists of the tribes Capreolini (Capreolus and Hydropotes),
Alceini (Alces), and Odocoileini (Rangifer + American genera). Deer endemic to the New World fall in two biogeographic lineages: the Wrst
one groups Odocoileus and Mazama americana and is distributed in North, Central, and South America, whereas the second one is composed
of South American species only and includes Mazama gouazoubira. This implies that the genus Mazama is not a valid taxon. Molecular dat-
ing suggests that the family originated and radiated in central Asia during the Late Miocene, and that Odocoileini dispersed to North Amer-
ica during the Miocene/Pliocene boundary, and underwent an adaptive radiation in South America after their Pliocene dispersal across the
Isthmus of Panama. Our phylogenetic inferences show that the evolution of secondary sexual characters (antlers, tusk-like upper canines, and
body size) has been strongly inXuenced by changes in habitat and behaviour.
© 2006 Elsevier Inc. All rights reserved.
Keywords: Cervidae; Ruminantia; Phylogeny; Taxonomy; Evolution; Biogeography; Sexual dimorphism; Mitochondrial DNA; Nuclear DNA
1. Introduction
With 40 species of deer, the family Cervidae constitutes
the second most speciose family of artiodactyls after the
Bovidae (Grubb, 1993). Widely divergent in size, habitat,
and behaviour, Cervidae are united by a series of synapo-
morphies, including the possession of antlers in males
(Janis and Scott, 1987). In the classiWcation of Grubb
(1993), the 16 genera of Cervidae are arranged in four sub-
families: Cervinae, Muntiacinae, Hydropotinae, and Odo-
coileinae. The subfamily Cervinae includes four genera: (1)
Dama in Eurasia with two species of fallow deer, (2) Axis
with four Asian species (e.g., hog deer and chital), (3) Ela-
phurus davidianus (Père David’s deer) in China, and (4) the
broadly distributed genus Cervus, with nine species in Asia
and Cervus elaphus, which is widespread throughout the
whole northern hemisphere. The subfamily Muntiacinae
contains two Asian genera: Muntiacus, with nine species of
muntjacs, and the monotypic Elaphodus cephalophus
(tufted deer). The subfamily Hydropotinae is only repre-
sented by the antlerless Hydropotes inermis (Chinese water
deer). Finally, the subfamily Odocoileinae is the most
*Corresponding author. Fax: +33 1 40 79 30 63.
E-mail address: Hassanin@mnhn.fr (A. Hassanin).
102 C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117
heterogeneous group, with nine genera currently deWned:
(1) Alces (moose) and (2) Rangifer (reindeer), which are
both largely dispersed in North America and Eurasia, (3)
the Eurasian genus Capreolus with two species of roe deer,
(4) Odocoileus (mule deer and white-tailed deer) and (5)
Mazama (six species of brocket deer), which are both pres-
ent in North and South America, and the four South Amer-
ican genera: (6) Blastocerus (marsh deer), (7) Hippocamelus
(two species of huemuls), (8) Ozotoceros (pampas deer), and
(9) Pudu (two species of pudus).
Other morphological classiWcations and phylogenies
agree with the monophyly of the 16 genera of Cervidae, but
many inconsistencies remain concerning the intergeneric
relationships. In the Wrst classiWcation of Cervidae, Brooke
(1878) recognized the four subfamilies retained by Grubb
(1993), but two higher taxa were also proposed on the basis
of diVerences in the metacarpals: (1) the Plesiometacarpa-
lia, which unites Cervinae and Muntiacinae, as both possess
only the proximal part of the lateral metacarpals; and (2)
the Telemetacarpalia, which includes Odocoileinae and
Hydropotinae, as both possess only the distal part of the
lateral metacarpals. In contrast, Groves and Grubb (1987)
did not consider Telemetacarpalia to be a valid taxon, as
they suggested that Hydropotes was the sister group to all
other Cervidae because it does not possess antlers.
The subfamily Odocoileinae is the most problematic
taxon, and intergeneric relationships are particularly con-
troversial within this group. Brooke (1878) grouped Rang-
ifer with the six genera endemic to America (Blastocerus,
Hippocamelus, Mazama, Odocoileus, Ozotoceros, and
Pudu), because all of them possess a vomerine septum that
completely separates the choana. Simpson (1945) did not
follow this hypothesis and split the Odocoileinae into
the Wve tribes Odocoileini (Blastocerus, Hippocamelus,
Mazama, Odocoileus, Ozotoceros, and Pudu), Alcini (Alces),
Capreolini (Capreolus), Hydropotini (Hydropotes), and
Rangiferini (Rangifer). Following Simpson (1945), McK-
enna and Bell (1997) placed Alces and Capreolus in their
own tribe Alceini (diVerent spelling from “Alcini”; Simp-
son, 1945) and Capreolini, but they included Rangifer in the
tribe Odocoileini and separated Hydropotes in the subfam-
ily Hydropotinae. In an analysis of morphological charac-
ters, Webb (2000) excluded Alces from the Odocoileinae
and recognized only two tribes in this subfamily: Rangife-
rini, which contains Rangifer, Hippocamelus, and Pudu; and
Odocoileini, which includes Odocoileus, Blastocerus,
Mazama, and Ozotoceros. Additional subfamilies have been
proposed on the basis of antler characteristics: Pocock
(1923) proposed the inclusion of Rangifer in its own sub-
family Rangiferinae, because the reindeer is the only species
with antlered females; and Kraglievich (1932) included
Mazama and Pudu in the subfamily Mazaminae, as they
share simple, one tined antlers.
The discrepancies between morphological phylogenies
of Cervidae can be explained by the fact that all these stud-
ies have used diVerent matrices of characters (Groves and
Grubb, 1987; Meijaard and Groves, 2004). It is also very
diYcult to evaluate the reliability of these hypotheses,
because none of them has used a clearly deWned methodo-
logical approach. Moreover, the usefulness of morphologi-
cal characters in deciphering the phylogeny of Cervidae,
and to a larger extent of ruminants, has been repeatedly
questioned because of their high level of homoplasy (Gen-
try, 1994; Groves and Grubb, 1987; Hassanin and Douzery,
2003; Janis and Scott, 1987; Scott and Janis, 1987).
Several molecular studies have also been published on
Cervidae. The cytochrome b (Cyb) analysis of Randi et al.
(1998) supported the monophyly of Plesiometacarpalia but
not that of Telemetacarpalia. The multigene analysis of
Hassanin and Douzery (2003) favored the major Plesio-/
Telemetacarpalia dichotomy, but only six genera of Cervi-
dae were included in this study. Within the subfamily
Muntiacinae, the analyses of several mitochondrial markers
(D-loop, Cyb, 12S and 16S rRNA, ND4 and ND4L) have
provided a strong support for the monophyly of Muntiacus
and interspeciWc relationships (Amato et al., 2000; Wang
and Lang, 2000). The monophyly of the subfamily Cervinae
has been supported by analyzing a large species sample
with mitochondrial sequences of the control region (D-
loop) and Cyb gene (Bonnet, 2001; Liu et al., 2003; Ludt
et al., 2004; Pitra et al., 2004; Randi et al., 1998; Randi et al.,
2001). Within Cervinae, all Cyb studies have shown
the polyphyly of Cervus and Axis, with C. eldi grouped
with Elaphurus, A. axis allied with C. duvauceli and
C. schomburgki, and A. porcinus linked to C. timorensis
(Liu et al., 2003; Ludt et al., 2004; Pitra et al., 2004). The
polyphyly of Cervus was also reported in two analyses of
the D-loop (Bonnet, 2001; Randi et al., 2001), but Axis was
found monophyletic in Bonnet (2001).
The monophyly of Odocoileinae sensu Grubb (1993) has
been questioned based on molecular studies. Indeed, Capre-
olus should be excluded from this group, as it was found
grouped with Hydropotes in Cyb studies (Pitra et al., 2004;
Randi et al., 1998), and in the multigene analysis of Hassa-
nin and Douzery (2003). This molecular hypothesis sug-
gests that antlers were completely lost in the lineage leading
to the extant Hydropotes (Randi et al., 1998). In addition,
Cyb sequences have suggested that the American genera
Mazama and Odocoileus are closely related, and that they
share aYnities with Rangifer (Randi et al., 1998). The posi-
tion of the moose (Alces) is unresolved: it is supported as
either the sister group to the clade uniting American deer
and Rangifer, or allied with Capreolus and Hydropotes
(Randi et al., 1998).
Molecular investigations have greatly helped in delimit-
ing clades within Cervidae, and in understanding the evolu-
tion of morphological characters. Nevertheless, because of
insuYcient gene and taxon sampling, in particular for deer
of America, many nodes remain unresolved, impeding the
interpretation of many aspects of the biogeographic history
of the family. In particular, questions regarding the origin
of American deer and their colonization of South America
have never been addressed. Here, we propose to Wll these
gaps by combining the strength of a large taxonomic sam-
C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117 103
ple including 25 species of Cervidae (representing 15 gen-
era) with the phylogenetic signal from both mitochondrial
and nuclear markers. To provide a consistent signal span-
ning the whole evolutionary history of the family, two rap-
idly evolving mitochondrial genes, i.e., cytochrome b (Cyb)
and the subunit II of the cytochrome oxidase (CO2), are
combined with two slower nuclear markers, i.e., intron 2 of
the -lactalbumin (LAlb) and intron 1 of the protein
kinase C iota gene (PRKCI). This enables us to address the
following four phylogenetic questions: (1) Are the supra-
generic taxa previously described in the literature (Plesio-/
Telemetacarpalia; Cervinae; Muntiacinae; Odocoileinae)
monophyletic? (2) Within Cervinae, are the genera Cervus
and Axis monophyletic? (3) What is the phylogenetic posi-
tion of the South American genera? (4) Among American
deer, are the genera Mazama and Odocoileus monophy-
letic? In addition, divergence times are estimated with the
relaxed Bayesian molecular clock approach by using sev-
eral calibration points. The results are compared with avail-
able data on fossils, paleoclimatic changes, and the
evolution of landscapes to provide a comprehensive sce-
nario for the biogeography and morphological evolution of
Cervidae during the Neogene.
2. Materials and methods
2.1. Taxonomic sample
The ingroup includes 25 species of Cervidae representing
15 genera. As the monophyly of Muntiacus was conWrmed
with mitochondrial sequences (Amato et al., 2000; Wang
and Lang, 2000), the monophyly of the subfamily Muntiac-
inae is here tested by including one species of Muntiacus
and the monospeciWc genus Elaphodus. The subfamily Cer-
vinae was densely sampled (12 species) to clarify the status
of the genera Axis and Cervus. All genera of Odocoileinae
except Ozotoceros were incorporated in the analyses to bet-
ter understand the evolution of American deer. The out-
group contains four species belonging to four diVerent taxa
of the suborder Ruminantia: Antilocapra americana
(Antilocapridae), Moschus moschiferus (Moschidae), Gaz-
ella granti (Bovidae, Antilopinae), and Tragelaphus imber-
bis (Bovidae, Bovinae) (Table 1).
2.2. DNA extraction, ampliWcation, and sequencing
Most of the tissues used in this study come from the col-
lections of the National Museum of Natural History of
Paris (Muséum National d’Histoire Naturelle, MNHN)
(Table 1). Cells and blood samples were digested in CTAB
(hexade Cyl Trimethyl Ammonium Bromide) using the pro-
tocol detailed in Winnepenninckx et al. (1993); DNA was
puriWed in chloroform isoamyl alcohol, and was then pre-
cipitated with isopropanol. For bone samples, DNA was
extracted following Hassanin et al. (1998). First, bones were
ground in liquid nitrogen and digested in a lysis solution
(Tris–HCl 10 mM; EDTA 0.5 M, pH 8.5; SDS 0.5%; pro-
teinase K 200 g/ml). Second, the supernatant was dialysed,
and DNA was puriWed several times by using phenol and
chloroform. Finally, DNA was precipitated with isopropa-
nol.
Four genes were sequenced: two mitochondrial protein-
coding genes, i.e., Cyb (1140bp) and CO2 (582 bp), and two
nuclear fragments, i.e., intron 2 of -lactalbumin (LAlb),
which is 462 bp long in Cervus elaphus (Accession No.
AY122017), and intron 1 (and two small exonic regions) of
the gene encoding the protein kinase C iota (PRKCI),
which is 514 bp long in C. elaphus (Accession No.
AY846793). Most sequences were obtained using several
overlapping PCR ampliWcations. The exact matching
between the overlapping portions of two diVerent PCR
fragments has been checked as an evidence of authenticity
of sequences. Most primers come from previous studies:
Hassanin et al. (1998) and Hassanin and Douzery (1999)
for Cyb; Hassanin and Douzery (2003) for LAlb; Hassanin
and Ropiquet (2004) for CO2, and Ropiquet and Hassanin
(2005a) for PRKCI. AmpliWcations were done in 50 l using
the following PCR standard conditions: buVer 10£ with
MgCl2 (1.5 mM), 5 l; dNTPs, 5 l (6.6mM); Taq Appligen,
0.3 l (2.5 U); and primers, 2.5 l at 10 M. The standard
PCR program used was: 94 °C for 4 min; 94 °C for 1 min,
50–60 °C for 1 min, 72 °C for 1min (30 cycles). AmpliWca-
tion products were puriWed using the Montage PCR Cen-
trifugal Filter Devices (Millipore). All sequences were
obtained by double-strand DNA cycle sequencing with a
CEQ2000 Dye terminator cycle Sequencing Quick Start kit
in a CEQ2000 Beckman (v4.3.9) sequencer. The resulting
output was edited using Sequencher 4.1 (Gene Codes, Ann
Arbor, Michigan).
2.3. Phylogenetic analyses
Alignments were done by eye on Bioedit v5.0.6 (Hall,
2001). The mitochondrial protein-coding genes, i.e., Cyb
and CO2, did not pose any problem of primary homology
because no gaps were included in the alignment. In con-
trast, four nuclear regions which were found ambiguous for
the position of gaps were excluded from analyses: 3 nucleo-
tides (nt) and 9 nt, respectively, at position 70 and 218 of the
LAlb sequence of C. elaphus (Accession No. AY122017),
and 14 nt and 17 nt at position 148 and 244 of the PRKCI
sequence of C. elaphus (Accession No. AY846793). We also
excluded Wve autapomorphic single nucleotides (pos. 113,
119, 179, and 246 of LAlb and pos. 304 of PRKCI). The
alignments are available in the supplementary material S1
on the MPE’s web page. Bayesian and maximum likelihood
(ML) analyses were performed on each of the four genes
separately and on a matrix combining these four genes. The
models of sequence evolution were selected by using
MrMODELTEST (v2.2) (Nylander, 2004). These models
are GTR+ I+ for CO2 and Cyb, HKY+ for LAlb and
PRKCI, and GTR +I + for the concatenated dataset.
Bayesian analyses were performed using MrBayes
(v3.0b4) (Huelsenbeck and Ronquist, 2001). To ensure a
104 C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117
Table 1
Origin of the sequences
aThis study; Cyto: cytogenetic collection of the MNHN; Spot: tissue collection of the MNHN; MJP: Ménagerie du jardin des plantes of the MNHN; Zoothèque: osteological collection of the
MNHN; ISEM: Institut des Sciences de l’Evolution de Montpellier; Unp: Unpublished; (1) Honeycutt et al. (1995); (2) Miyamoto et al. (1994); (3) Ludt et al. (2004); (4) Randi et al. (1998); (5) Hassanin
and Douzery (2003); (6) Hassanin and Douzery (1999); (7) Hassanin et al. (1998); (8) Ropiquet and Hassanin (2005a); (9) Matthee and Davis (2001).
Taxa Common name Collection reference CO2 Cyb PRKCI LAlb
Cervinae Axis axis Chital Cyto 02.090 DQ379310aDQ379302aDQ379329aDQ379348a
porcinus Hog deer Cyto 02.054 DQ379311aDQ379301aDQ379367aDQ379349a
Cervus albirostris Thorold’s deer Cyto 02.052 DQ379312aAF423202 (3) DQ379330aDQ379350a
duvauceli Swamp deer Cyto 02.057 DQ379313aDQ379303aDQ379331aDQ379351a
elaphus Red deer Spot 01.176 DQ365689aAY244490 (3) AY846793 (8) AY122017 (5)
eldi Eld’s deer Cyto 02.060 DQ379314aAY157735 (Unp) Negative PCR DQ379353a
nippon Sika MJP 12887 C2 DQ379315aAY035876 (3) DQ379332aDQ379352a
timorensis Timor’s deer Cyto 02.65 DQ379316aAF423200 (3) DQ379333aDQ379354a
unicolor Sambar W. Robichaud T0009A DQ379317aAF423201 (3) DQ379334aDQ379355a
Dama dama Fallow deer Cyto 02-104 DQ379318aAJ000022 (4) DQ379335aDQ379356a
mesopotamica Persian fallow deer Cyto 02.077 DQ379319aDQ379304aDQ379336aDQ379357a
Elaphurus davidianus Père David’s deer Cyto 02.070 DQ379320aAF423194 (3) DQ379337aDQ379358a
Muntiacinae Elaphodus cephalophus Tufted deer Zoothèque 1896–689 DQ379321aDQ379305aDQ379339aDQ379359a
Muntiacus reevesi Reeves’ muntjac — NC_004069 (Unp) NC_004069 (Unp) AF165677 (9) AY122018 (5)
Alcinae Alces alces Moose Vincennes’s Zoo DQ379322aAJ000026 (4) DQ379338aDQ379360a
Capreolinae Capreolus capreolus Roe deer Cyto 2000.256 DQ365690aAJ000024 (4) DQ365692aAY122021 (5)
Hydropotes inermis Chinese water deer ISEM T4307 DQ379323aAJ000028 (4) DQ379340aAY122020 (5)
Odocoileinae Blastocerus dichotomus Marsh deer Zoothèque Trophy 175 DQ379324aDQ379306aDQ379341aDQ379361a
Hippocamelus antisensis Huemul Zoothèque 1957–1302 DQ379325aDQ379307aDQ379342aDQ379362a
Mazama americana Red brocket Cyto 94.072 DQ379326aAJ000027 (4) DQ379343aDQ379363a
gouazoubira Gray brocket ISEM T1627 DQ379368aDQ379308aDQ379344aDQ379364a
Odocoileus hemionus Mule deer ISEM T176 DQ379369aAF091630 (4) DQ379345aAY122022 (5)
virginianus White-tailed deer Cyto 02.133 U18816 (1) DQ379370aDQ379346aDQ379365a
Pudu puda Pudu Spot 330 DQ379327aDQ379309aDQ379347aDQ379366a
Rangifer tarandus Reindeer ISEM T45353 DQ379328aAJ000029 (4) AF165693 (9) AY122019 (5)
Moschidae Moschus moschiferus Musk deer Spot 1258 DQ365691aAY121995 (5) DQ365693aAY122033 (5)
Bovidae Tragelaphus imberbis Lesser kudu — U18815 (1) AF036279 (6) AF165733 (9) AY122025 (5)
Gazella granti Grant’s gazelle — U18824 (1) AF034723 (7) AF165749 (9) AY122029 (5)
Antilocapridae Antilocapra americana Pronghorn — U62571 (2) AF091629 (6) AF165669 (9) AY122014 (5)
C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117 105
good exploration of the posterior probabilities (PP) surface
and to reach the stationary distribution of Markov chains,
Wve chains were run for 2,000,000 generations and sampled
every 100 generations. Ten thousand sampled trees were
discarded as “burn in.” Each analysis was repeated twice to
check that the chains always converged towards the same
likelihood score. Unambiguous indels (insertions or dele-
tions) were coded as binary characters. Two diVerent char-
acter partitions were therefore used for each nuclear
dataset: (1) DNA sequences were analyzed with the model
selected by MrMODELTEST, and (2) indels were analyzed
using the parsimony options. Five character partitions were
used for the combined analysis: the four diVerent genes
were analyzed by applying four diVerent GTR + I+ mod-
els, and indels were analysed with the parsimony options.
For the Bayesian bootstrap analysis, 100 pseudoreplicates
of the combined matrix were Wrst created using SEQBOOT
3.5c (Felsenstein, 2004). Bootstrap Bayesian values (BPB)
were obtained by constructing the consensus of the 100
Bayesian trees with CONSENSE 3.5c (Felsenstein, 2004).
The bootstrap ML analyses (BPML) were performed using
PHYML 2.1b1 (Guindon and Gascuel, 2003) with 1000
replicates. We used the same models as for the Bayesian
analyses.
2.4. Molecular dating
Divergence times were calculated according to the
relaxed Bayesian molecular clock approach for multigene
datasets described in Thorne et al. (1998) and Thorne and
Kishino (2002) and implemented in MULTIDIVTIME.
Divergence times were estimated for each four genes sep-
arately and for the combined matrix. The mean of the dis-
tribution of the root’s age (rttm) was set at rttmD25
Million Years Ago (MYA) with a standard deviation (rtt-
msd) set at rttmsd D10 MYA. To approximate the mean of
the prior distribution of the rate of evolution at the root of
the tree (rtrate), we followed the procedure recommended
in the documentation of the software. We therefore used
the following values: rtrate D0.07 for the Cyb, 0.06 for
CO2, 0.013 for LAlb, 0.008 for PRKCI, and 0.05 for the
combined analysis. The Markov chains were sampled
10,000 times every 100 generations and the “burn in”
period was set at 100,000 generations.
Three calibration points were used for the analyses: the
Wrst refers to the oldest fossil of Cervidae (20 §2 MYA;
Ginsburg, 1988), the second refers to the oldest fossil attrib-
uted to the subfamily Muntiacinae or tribe Muntiacini
(8 §1 MYA; Dong et al., 2004) and the third refers to the
oldest fossil of the clade Rangifer + American genera or
tribe Odocoileini (5 §1 MYA; Vislobokova, 1980) (see
paragraphs 4.1 and 4.2). The use of such constraints
involves the assumption that the age of the oldest fossil
attributed to a node is a good approximation of the mini-
mum age of this node. The divergence times were estimated
using the three possible pairs of calibration points: (1) Cer-
vidae and Muntiacinae, (2) Cervidae and (Rangifer +
American genera), and (3) Muntiacinae and (Rangifer +
American genera).
2.5. Reconstruction of the ancestral morphotype of Cervidae
The tree resulting from the combined analysis (Fig. 4)
was used for studying whether sexual characters (type of
antlers, upper canines, and body size) have evolved in
relation to changes in behaviour and habitat . The four
discrete states used for antlers were those previously
deWned by Pocock (1933) as follows: (1) one tine, (2) two
tines, (3) three tines, and (4) four or more tines. The pres-
ence/absence and shape of upper canines were observed
on a series of skulls available in the MNHN collections of
Paris (Catalog numbers of the specimens are available in
the supplementary material S2 on the MPE’s web page).
Two character states were considered for the type of habi-
tat: (1) open habitat, which includes grasslands, marsh-
lands, and open forests; and (2) closed habitat, which
includes dense forests and marshes with reeds. We coded
two categories of body size, i.e., minimal shoulder size
higher or smaller than 650 mm. The presence/absence of
sexual dimorphism in body size was coded by taking into
account the body mass for each sex (Geist and Bayer,
1988; Loison et al., 1999; Merino et al., 2005; Mooring
et al., 2004). Habitat types and shoulder heights were
taken from Nowak (1999). The matrix of character states
is available in Fig. 4. The evolution of these characters
was optimized using the two parsimony options ACC
TRAN and DELTRAN of the software PAUP 4.0b10
(SwoVord, 2003). Only one ancestral character state was
found ambiguous because of contradictions between
ACCTRAN and DELTRAN inferences.
3. Results
3.1. Phylogenetic analyses
The four genes have Wrst been analyzed separately
(Fig. 1) and then concatenated to constitute a matrix
including 2718 nucleotides and 6 unambiguous indels (four
in PRKCI and two in LAlb; see in Fig. 2).
3.1.1. Monophyly of the family Cervidae
Each of the four genes strongly supports the monophyly
of the Cervidae whatever the type of analysis (0.99< PP < 1,
and 70 < BPML < 100). The family is characterized by 11
exclusive synapomorphies (the positions of all synapomor-
phies are deWned on the sequences of C. elaphus, see acces-
sion numbers in Table 1): the most striking is a 16
nucleotides deletion (CATAAAAGGCAACAGG) at posi-
tion (pos.) 355 of the LAlb; seven are transversions
(LAlb: T!G and A!C in pos. 164 and 330; PRKCI:
A!T and G!C in pos. 167 and 286; CO2: A!T, pos. 6;
Cyb: C!A and G!T in pos. 114 and 712), and three are
transitions (LAlb: C!T, pos. 86; PRKCI: A!G in pos. 7
and 212).
106 C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117
3.1.2. The Plesio-/Telemetacarpalia dichotomy
The analysis combining the four markers reveals a main
dichotomy within Cervidae separating the Plesiometacarpa-
lia, which includes the subfamilies Cervinae and Muntiacinae
(PP D1; BPB/ML D100), and the Telemetacarpalia, which
groups all other Cervidae (members of the subfamilies Odo-
coileinae and Hydropotinae) (PPD1; BPBD83; BPML D70).
The monophyly of Plesiometacarpalia was recovered in the
independent analyses of Cyb (PP D0.92; BPML D75), LAlb
(PP D0.99; BPML D52), and PRKCI (PP D0.58; BPML D32).
Fig. 1. Phylogenetic analyses of the four markers: CO2, Cyb, LAlb and PRKCI. The trees were obtained with the Bayesian approach using the evolution-
ary model selected under MrModeltest 2.2, i.e., GTR+I+ for mitochondrial markers, and HKY+ for nuclear markers. The indels found in the nuclear
genes were coded as binary characters, and analyzed in a diVerent partition from the nucleotides characters (see Section 2). The values on each node corre-
spond to the Bayesian posterior probabilities (PP> 0.5) and bootstrap proportions calculated with the maximum likelihood method (BPML).
Alces alces
Capreolus capreolus
Hydropotes inermis
1/96
Mazama americana
Mazama gouazoubira
Pudu puda
Rangifer tarandus
Odocoileus hemionus
Odocoileus virginianus
0.68/50
Blastocerus dichotomus
Hippocamelus antisensis
0.65/53
0.76/61
1/80
Muntiacus reevesi
Cervus albirostris
Cervus duvauceli
Elaphodus cephalophus
Elaphurus davidianus
Cervus eldi
Cervus timorensis
Cervus unicolor
Cervus elaphus
Cervus nippon
1/97
Dama dama
Dama mesopotamica
1/100
Axis axis
Axis porcinus
1/93
0.67
0.99/52
1/93
Tragelaphus imberbis
0.81/84
1/87
0.99/80
0.98/69
1/85
1/85
0.97/61
0.52
0.58/32
Antilocapra americana
Moschus moschiferus
Gazella granti
1/84
0.84/43
0.88/81
1/90
0.64/47
αLAlb PRKCI
0.82/39
0.98/91
0.99/90
0.8/39
1/100
1/86
1/100
0.98/84
1/100
0.97/74
1/91
1/99
0.92/75
0.50/50
0.8/60
0.85/76
1/91
1/96
0.9/90
1/99
0.8/60
0.75/69
1/100
0.67/57
1/100
Antilocapra americana
Moschus moschiferus
Axis axis
Cervus duvauceli
Axis porcinus
Dama dama
Dama mesopotamica
0.98/87
Elaphurus davidianus
Cervus eldi
Cervus timorensis
Cervus unicolor
0.84/71
Cervus elaphus
Cervus albirostris
Cervus nippon
1/97
0.57/55
0.86/96
0.64/74
0.56/53
Elaphodus cephalophus
Muntiacus reevesi
Rangifer tarandus
Alces alces
Mazama americana
Odocoileus hemionus
Odocoileus virginianus
1/83
1/99
Blastocerus dichotomus
Hippocamelus antisensis
Mazama gouazoubira
Pudu puda
0.89/71
1/88
1/71
Capreolus capreolus
Hydropotes inermis
0.63/50
0.77
/26
0.52
0.99/70
Gazella granti
Tragelaphus imberbis
0.77/36
0.86/47
CO2 Cyb
PP/BP ML
C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117 107
The monophyly of Telemetacarpalia is also recovered by the
analysis of PRKCI (PPD0.64; BPML D47). Moreover, all
members of this clade are diagnosed by a deletion of 12
nucleotides (AATACCCCGTA) at pos. 238 of LAlb, and
by a transition A!G in pos. 251 of PRKCI.
3.1.3. Within Plesiometacarpalia
According to the combined analysis, the Plesiometacarpa-
lia are divided in two strongly supported clades which corre-
spond to the subfamilies Cervinae (PPD1; BPB/ML D100)
and Muntiacinae (PPD1; BPBD77; BPML D72). The Cyb
data alone strongly supports the two subfamilies (PPD1;
BPML D99 and 96). They are also recovered with PRKCI,
and although this marker provides a weak quantitative
signal, it contains a molecular signature for each subfamily,
i.e, an insertion of two nucleotides (AT) in pos. 81 for Cervi-
nae, and a transition A!G in pos. 433 for Muntiacinae.
Within Cervinae, the monophyly of the genus Dama is
found in the independent analyses of the four genes
(0.75 < PP < 1; 69 < BPML < 100), and in the combined
analysis (PP D1; BPB/ML D100). In addition, Dama is
diagnosed by a transversion T!G in pos. 375 of PRKCI.
The genus Axis was found monophyletic with the com-
bined analysis (PPD1; BPB/ML D100), as well as with
independent analyses of Cyb, LAlb, and PRKCI (PP D1;
85 < BPML < 100). Moreover, both species of Axis share
six exclusive synapomorphies, including four transitions
(Cyb: T!C, pos. 884; LAlb: A!G and G!A in pos.
125 and 362; PRKCI: T!C, pos. 461, and two G!T
transversions Cyb: pos. 462; PRKCI: pos. 142). By con-
trast, the genus Cervus was found polyphyletic in the com-
bined analysis, and in the independent analyses of
mitochondrial markers: C. duvauceli is allied with Axis
(PP D0.96; BPB/ML D90/91); C. eldi is linked to Elaphurus
(PP D0.87; BPB/ML D91/97); and the latter clade plus all
other species of Cervus (C. timorensis, C. unicolor, C. albi-
rostris, C. elaphus, and C. nippon) form the sister group to
Dama (PP D0.96; BPB/ML D93/88). Nuclear genes do not
Fig. 2. Bayesian tree resulting from the analysis combining the four markers CO2, Cyb, LAlb and PRKCI. The selected model is GTR +I + for both
Bayesian and maximum likelihood (ML) analyses. The values shown for each node correspond to posterior probabilities (PP), Bayesian and ML boot-
strap proportions (BPB and BPML, respectively) in the order indicated in the rectangle. The diVerent symbols indicate the four diagnostic indels which can
be found in the two nuclear genes: D deletion of ATT in PRKCI, D deletion of AATACCCTGTA in LAlb, Dinsertion of AT in PRKCI and
Ddeletion of ACATAAAAGGCAACAG in LAlb.
PLESIOMETACARPALIA
Antilocapra americana
Dama dama
Dama mesopotamica
1/100/100
Cervus timorensis
Cervus unicolor
1/100/100
Cervus albirostris
Cervus elaphus
Cervus nippon
1/52/56
1/93/96
1/100/100
Elaphurus davidianus
Cervus eldi
0.99
0.96/93/88
Cervus duvauceli
Axis axis
Axis porcinus
1/100/100
0.96/90/91
1/100/100
Elaphodus cephalophus
Muntiacus reevesi
1/77/72
1/97/99
Capreolus capreolus
Hydropotes inermis
1/100/100
Alces alces
Rangifer tarandus
Mazama americana
Odocoileus hemionus
Odocoileus virginianus
1/100/100
1/100/100
Blastocerus dichotomus
Mazama gouazoubira
Hippocamelus antisensis
Pudu puda
0.51/52/46
1/99/98
1/98/97
1/100/99
0.73
1/83/70
1/100/100
Moschus moschiferus
Gazella granti
Tragelaphus imberbis
0.99/50/50
1/80/63
0.87/91/97
0.75/52/51
C
E
R
V
BOVIDAE
I
D
A
E
/86/89
PP/BP
B
/BP
ML
TELEMETACARPALIA
108 C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117
produce a strong signal for the position of the various
species of Cervus: only two nodes are supported by PP
values superior to 0.50, and they are incongruent with the
mitochondrial analyses: C. nippon is allied with C. elaphus
in the LAlb tree (PPD1; BPML D97; diagnostic transi-
tion T!C in pos. 220), whereas it is grouped with C. albi-
rostris in the mtDNA trees (CO2/Cyb: PP D1/0.5;
BPML D 97/50; diagnostic transition A!G in pos. 111 of
CO2); C. timorensis is allied with Elaphurus in the PRKCI
tree (PP D0.97; BPML D61; diagnostic transversion A!T
in pos. 362), whereas it is allied with C. unicolor in the
mtDNA trees (CO2/Cyb: PP D0.84/1, and BPML D71/100;
diagnostic transition A!G in pos. 36 of Cyb).
3.1.4. Within Telemetacarpalia
The subfamily Odocoileinae sensu Grubb (1993) is
found to be paraphyletic because Hydropotes (Hydropoti-
nae) is robustly allied with Capreolus. This clade is found in
the combined analysis (PPD1; BPB/ML D100), and in all the
separate analyses of the four genes (0.63< PP < 1;
50 < BPML < 99). Moreover, this grouping is supported by
three transitions (LAlb: A!G and C!T in pos. 130 and
pos. 43; PRKCI: A!G, pos. 109) and two transversions
(Cyb: A!T, pos. 318; LAlb: C!A, pos. 357). The Ameri-
can genera, i.e., Blastocerus, Hippocamelus, Mazama,
Odocoileus, and Pudu, form a monophyletic assemblage in
the combined analysis (PP D1; BPB/ML D98/97), and in
the mitochondrial analyses (CO2/Cyb: PP D1/0.98;
BPML D71/91). Their grouping with Rangifer receives
strong support in the Cyb, LAlb, and PRKCI analyses
(PP D1; 80 < BPML < 86) and in the combined analysis
(PP D1; BPB/ML D100/99). The position of Alces is ambigu-
ous: it is allied with Capreolus and Hydropotes in the Cyb
tree (PP D0.8; BPML D60), and in the bootstrap analyses of
the combined matrix (BPB/ML D61/74), whereas it is
grouped with Rangifer and American genera in the PRKCI
tree (PP D0.84; BPML D43), and in the Bayesian analysis of
the combined analysis (PPD0.73).
The monophyly of Odocoileus is strongly supported by
the mitochondrial genes (CO2/Cyb: PP D1; BPML D83/
100), and in the combined analysis (PP D1; BPB/ML D100),
and it is also weakly favored by the nuclear LAlb
(PP D0.68; BPML D50). In contrast, the genus Mazama is
found polyphyletic in the combined analysis and in the
independent analyses of mtDNA markers: Mazama ameri-
cana is grouped with the genus Odocoileus (combined anal-
ysis/CO2/Cyb: PP D1, BP D99–100), resulting in a clade
present in both North and South America, whereas
Mazama gouazoubira clusters with the genera Blastocerus,
Hippocamelus, and Pudu (combined analysis/CO2/Cyb:
PP D0.99–1, BPD88–99), forming a strictly South Ameri-
can clade.
3.2. Divergence times estimates
Three pairs of calibration points were used for estimat-
ing divergence times (Fig. 3): in the Wrst one, the node Cer-
vidae at 20 §2 MYA was combined with the node
Muntiacinae ( DMuntiacini) at 8 §1 MYA; in the second
one, the node Cervidae at 20§2 MYA was combined with
the node (Rangifer + American genera) ( DOdocoileini) at
5§1 MYA; in the third one, the node Muntiacinae at 8 §1
MYA was combined with the node (Rangifer + American
genera) at 5 §1 MYA. The results show that the dates
obtained with this third combination of calibration points
are much younger than in the two Wrst analyses (Fig. 3).
For instance, the origin of the family Cervidae was esti-
mated between 7.7 and 9.6 MYA instead of 16.5–18 MYA,
and the origin of American deer at around 4.2–5.7 MYA,
instead of 9.1–12.9 MYA.
The dates obtained with each marker analyzed sepa-
rately were very close to those obtained in the combined
analysis (data not shown). However, we observed great
variations in the intervals of credibility: they were very
large for nuclear genes, shorter for mitochondrial genes,
and very short for the combined analysis. These results con-
Wrm that the intervals of credibility decrease when more
informative sites are included in the analyses. This observa-
tion is in favor of the combination of as many markers as
possible, as emphasized by Yang and Yoder (2003).
In our phylogenetic analyses, the genus Alces was found
as either the sister group to the clade composed of Rangifer
and American genera, or as the sister group to Hydropotes
Fig. 3. Divergence times resulting from the three diVerent molecular dat-
ing analyses. This graph illustrates the 95% credibility intervals (in ordi-
nates) for the dates obtained for the nodes of the combined tree presented
in Fig. 2 (in abscissa) by using three diVerent combinations of calibration
points: DCervidae (20 §2 MYA) + Muntiacini (D Muntiacinae)
(8 §1 MYA); DCervidae (20 §2 MYA) + Odocoileini ( D
(Rangifer + American genera) (5 §1 MYA); DOdocoileini (5 §1
MYA) + Muntiacini (8 §1 MYA). This shows that only the combination
of the calibration points Odocoileini (5 §1 MYA) + Muntiacini (8 §1
MYA) provides date estimates that are consistent with the fossil record of
Muntiacini and American deer (see text for more details). The lines
between the date estimates are indicated only to facilitate the comparisons
between the diVerent analyses and palaeontological/geological dates.
1
3
5
7
9
11
13
15
17
OldestMuntiacini(7-9)
IsthmusofPanama(3-3.5)
OldestAmericanfossils(5)
Dates
MYA
Nodes
Muntiacini
SouthAmericandeer
Americandeer
Cervidae
Dama
Cervus
Mazama+Odocoileus
Axis+Rucervus
Cervini
Capreolini
Odocoileini
Capreolini+Alceini
Capreolinae
Cervinae
C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117 109
and Capreolus. The molecular dating analyses were there-
fore performed on these two topologies. The dates obtained
with Alces sister to the clade (Rangifer + American genera)
are all slightly older than those obtained with Alces sister to
(Hydropotes +Capreolus), but they diVer only by 0.15 mil-
lion years in average (data not shown). In particular, these
two sets of analyses provided very close dates for the com-
mon ancestor of Alces +(Capreolus +Hydropotes) (7.4
MYA) and the one of Alces +(Rangifer + American genera)
(7.5 MYA). These results suggest therefore that these three
lineages diverged from each other during a very short time
frame.
3.3. Ancestral morphotype of Cervidae
Fig. 4 shows the distributions of diVerent character
states corresponding to antlers, upper canines, body size,
sexual weight dimorphism, and habitat. These inferences
Fig. 4. Synthetic tree of the family Cervidae. The tree is a consensus derived from the Bayesian and maximum likelihood analyses of the matrix combining
all four markers (CO2, Cyb, LAlb and PRKCI). The nodes supported by a Bootstrap value below 70 in the combined analysis are not represented. The
date estimates were calculated by using the tree shown in Fig. 2 and are detailed in Table 2. The time scale comes from the Geological Society of America
(1999, available at http://www.geosociety.org/science/timescale/timescl.pdf). The symbols indicate the distributions of diVerent character states corre-
sponding to antlers ( D one tine; D two tines; D three tines; D four tines or more), D tusk-like upper canines, body size (S D Small, or
minimum shoulder size <650 mm; T D Tall or minimum shoulder size >650 mm), sexual dimorphism in weight ( D monomorphism; D dimor-
phism), and habitat type (O D open; C D closed). As detailed in Section 2, the evolution of these characters was inferred using PAUP 4.0b10
(SwoVord, 2003).
10 (MYA) 5.3 1.8 0.01
MIOCENE PLIOCENE PLEISTOCENE
Muntiacus reevesi
Elaphodus cephalophus
Dama dama
Cervus unicolor
Cervus elaphus
Cervus nippon
Cervus albirostris
Axis porcinus
Axis axis
Rangifer tarandus
Mazama americana
Odocoileus hemionus
Odocoileus virginianus
Blastocerus dichotomus
Mazama gouazoubira
Pudu puda
Hippocamelus antisensis
Hydropotes inermis
Capreolus capreolus
MUNTIACINI
C
E
R
V
ALCEINI
CAPREOLINI
C
A
P
R
E
O
L
I
N
A
E
CERVINI
C
O
I
L
E
I
N
I
O
D
O
I
N
A
E
C
E
R
V
I
D
A
E
Cervus timorensis
Cervus eldi
Cervus davidianus
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
SC
C
S
TO
O
SC
TO
TO
C
SO
C
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
TO
SC
SC
SC
SC
SC
Rucervus duvauceli
Alces alces
Dama mesopotamica
110 C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117
suggest that in the ancestor of Cervidae, the males were
large (shoulder height> 650 mm), bigger than females, with
three-tined antlers, without upper canines (although only
the tusk-like upper canines are illustrated on Fig. 4, the
presence/absence of upper canines was also tested), and
lived in open habitats. The same combination of character
states is also found for most nodes, but for the ancestors of
(Muntiacus +Elaphodus) and (Capreolus +Hydropotes), our
analyses suggest a completely diVerent pattern: they lived in
closed habitats, and the males were small (shoulder
height < 650 mm), similar in body mass to females, with
two-tined antlers. In addition, tusk-like upper canines were
acquired independently in Hydropotes and in the common
ancestor of (Muntiacus +Elaphodus). More generally, all
other species found in dense habitats (Mazama americana,
M. gouzoubira, Pudu) are characterized by a reduction of
antler size, a small body size, and sexual weight monomor-
phism.
4. Discussion
4.1. Phylogeny and taxonomy of the family Cervidae
In this study, we combine both nuclear and mitochon-
drial markers and the largest sample of genera so far pub-
lished to provide a good phylogenetic resolution spanning
the whole evolutionary history of the family Cervidae.
The results show that the family is divided into two main
clades corresponding to the Plesiometacarpalia and Tele-
metacarpalia proposed by Brooke (1878). Plesiometacar-
palia possess only the proximal part of the lateral
metacarpals II and V, whereas Telemetacarpalia possess
only the distal part of these metacarpals. This basal
dichotomy was previously found in the molecular study of
Hassanin and Douzery (2003) based on a much smaller
taxonomic sample. As the telemetacarpal condition is also
observed in musk deer (Moschidae) (Scott and Janis,
1987) and in the fossil ruminants Merycodontidae (Frick,
1937), it is likely that this character state is plesiomorphic
within Cervidae. If it is true, only the plesiometacarpal
condition should be considered as a synapomorphy,
which would actually be exclusive of the Plesiometacarpa-
lia. However, this hypothesis is not supported by Bouv-
rain et al. (1989), who argue that the plesiometacarpal
condition cannot derive from the telemetacarpal condi-
tion. They considered these two character states to have
evolved independently from an ancestral morphotype
possessing complete metacarpals (holometacarpal condi-
tion). As the telemetacarpal condition is observed in
Moschidae, which is the sister family of Bovidae (Hassa-
nin and Douzery, 2003), the hypothesis of Bouvrain et al.
(1989) implies that the evolution of this character is
homoplasic within ruminants. Bouvrain et al. (1989) also
observed that the temporal canal of all Telemetacarpalia
is broadly opened in its medial region, whereas it is closed
in other ruminants. Our results conWrm that this character
constitutes a synapomorphy of Telemetacarpalia.
Our analyses clearly divide the Plesiometacarpalia into
two main clades corresponding to the subfamilies Cervinae
and Muntiacinae in the classiWcation of Grubb (1993), or to
the tribes Cervini and Muntiacini in the classiWcation of
McKenna and Bell (1997). This dichotomy was previously
proposed on the basis of molecular data (e.g., Cronin et al.,
1996; Randi et al., 1998), but our study is the Wrst one incor-
porating all genera of Plesiometacarpalia. The grouping of
Muntiacus and Elaphodus in the subfamily Muntiacinae (or
in the tribe Muntiacini) has never been seriously questioned
based on morphological data, even if Groves and Grubb
(1987) noted that the general skull shape of Muntiacus
resembles that of Axis porcinus more than that of Elapho-
dus. Our study is the Wrst one conWrming the monophyly of
this taxon on a molecular basis: it is recovered in the com-
bined analysis (PP D1; BPB/ML D77/72), and independently
by two markers (Cyb and PRKCI). The grouping of Mun-
tiacus and Elaphodus is morphologically coherent, as it is
supported by one osteological synapomorphy, i.e., the
fusion of the large cuneiform and cubonavicular in the tar-
sus (Garrod, 1877). However, this character is not exclusive
to the muntjacs, as these tarsal bones are also fused in the
South American Pudu (Pocock, 1910). Since both muntjacs
and Pudu share similar habitats in dense forests, the fusion
of tarsal bones may be interpreted as being a convergent
adaptation for locomotion resulting from the colonization
of forested habitats.
The monophyly of the subfamily Cervinae sensu Grubb
(1993) (or tribe Cervini sensu McKenna and Bell, 1997)
is strongly supported in the combined analysis (PP D1;
BPB/ML D100), and is recovered by two markers (Cyb and
PRKCI). This result is in agreement with previous studies
based on Cyb or D-loop sequences (Bonnet, 2001; Douzery
and Randi, 1997; Pitra et al., 2004; Randi et al., 1998, 2001).
From the morphological point of view, Pocock (1910) pro-
posed that the hairy tuft on the upper half of the metapo-
dials could be used to diagnose this clade.
The genus Axis contains four species, and two of them
are included in our sample: the spotted deer (A. axis) and
the hog deer (A. porcinus). Both species are found in Nepal,
Sri Lanka and India, where their distributions sometimes
overlap (see for example Biswas, 1999); in addition, the
distribution of A. axis includes Bangladesh, while that of
A. porcinus covers Indochina, Pakistan, and China (Grubb
and Gardner, 1998). These two species are quite diVerent in
size and coat color (Nowak, 1999). Geist (1998) is even
doubtful of their close aYnity, because they exhibit diver-
gent behaviours during sexual arousal. The analyses of Cyb
sequences seem to conWrm this hypothesis, as Axis was
found paraphyletic in the studies of Liu et al. (2003) and
Pitra et al. (2004), with A. axis related to C. duvauceli, and
A. porcinus allied with C. timorensis. By contrast, the analy-
ses of D-loop support the monophyly of the genus Axis, and
its grouping with Cervus duvauceli (Bonnet, 2001). Our
analyses agree with this study: the monophyly of Axis is
strongly supported by three independent markers, includ-
ing the Cyb gene, and two nuclear introns (LAlb and
C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117 111
PRKCI), and Axis is grouped with C. duvauceli. All previ-
ous Cyb studies were actually performed by using the
sequence of A. porcinus produced by Ludt et al. (2004)
(Accession No. AY035874). Our results suggest therefore
that this Cyb sequence was obtained either from carry-over
contamination, or by using tissue from a misidentiWed spec-
imen. This hypothesis is also corroborated by our interpre-
tation of cytogenetic data. All species of Cervinae sensu
Grubb (1993) possess a fundamental number of 70, (Fon-
tana and Rubini, 1990). According to Bonnet (2001), A.
porcinus (2nD68) has retained the ancestral karyotype of
Cervinae, and A. axis (2nD66) diVers from this ancestral
karyotype by only one autapomorphic centric fusion. In
addition, C. timorensis (2nD60) and C. unicolor (2nD58 or
56) share four centric fusions. These data therefore favor
the grouping of C. timorensis with C. unicolor rather than
with A. porcinus, as proposed by Ludt et al. (2004).
Previous molecular studies based on D-loop and Cyb
sequences have concluded that the genus Cervus is polyphy-
letic, because C. eldi was allied with Elaphurus (Liu et al.,
2003; Pitra et al., 2004 and Randi et al., 2001), whereas C.
duvauceli and C. schomburgki were grouped with Axis
(Pitra et al., 2004). Our analyses conWrm the polyphyly of
Cervus. Highlighting the crucial need for a taxonomic revi-
sion of the subfamily Cervinae (or tribe Cervini), Randi
et al. (2001) proposed the recognition of only the three gen-
era Cervus, Axis, and Dama, with Elaphurus synonymized
with the genus Cervus. Here, we show that Cervus duvauceli
is grouped with the genus Axis, as previously found by
Pitra et al. (2004). We suggest therefore that C. duvauceli
should be placed in the genus Rucervus, as deWned by
Hodgson (1838).
The clade Telemetacarpalia is well supported in the com-
bined analysis (PP D1; BPB/ML D83/70), but it is only
retrieved with one marker independently (PRKCI). It is
however deWned by a strong molecular synapomorphy, i.e.,
a deletion of 12 nucleotides in LAlb, conWrming the Wnd-
ings of Hassanin and Douzery (2003) based on only four
Telemetacarpalia rather than 11 in the present study.
All members of Telemetacarpalia belong to the sub-
family Odocoileinae except the Chinese water deer
(Hydropotes inermis), which is the only species of the sub-
family Hydropotinae. Indeed, the subfamily Odocoileinae
sensu Grubb (1993) is found to be paraphyletic because
Hydropotes is robustly grouped with the roe deer (Capreo-
lus) in the combined analysis (PP D1; BPB/ML D100) and
in all independent analyses of the four genes (Fig. 1). In
contrast to Groves and Grubb (1987), who argue that
Hydropotes is divergent from all other deer, this grouping
conWrms previous molecular investigations based on vari-
ous mitochondrial and nuclear markers (Douzery and
Randi, 1997; Hassanin and Douzery, 2003; Randi et al.,
1998). In addition to the clade uniting Hydropotes and
Capreolus, two other lineages emerge from our analyses:
the Wrst one includes Rangifer and all American genera
(Blastocerus, Hippocamelus, Mazama, Odocoileus, and
Pudu), and the second one is composed only of the genus
Alces. The relationships between these three lineages
remain unclear. This lack of resolution may be explained
by the fact that they diverged from each other in a very
short amount of time.
To reconcile our phylogenetic results with taxonomy, we
propose a subdivision of the family Cervidae into the two
subfamilies Cervinae and Capreolinae, as recognized by
Pocock (1910). This arrangement is equivalent to the Ple-
sio-/Telemetacarpalia dichotomy proposed by Brooke
(1878). The subfamily Cervinae is composed of the two
tribes Cervini and Muntiacini. Within the subfamily Capre-
olinae, we propose to retain three tribes: the tribe Capreo-
lini includes Capreolus and Hydropotes, the tribe Alceini is
composed of Alces alone, and the tribe Odocoileini con-
tains Rangifer and all American genera, as previously pro-
posed by McKenna and Bell (1997). The tribe Odocoileini
is recovered in analyses of three independent markers (Cyb,
LAlb, and PRKCI), and is characterized by a deletion of a
triplet ATT in PRKCI. Brooke (1878) previously noted that
all Odocoileini share a vomerine septum dividing the cho-
ana into two chambers.
Within the tribe Odocoileini, the South American genera
constitute a monophyletic group which is divided into two
clades by the mitochondrial genes. Surprisingly, each of
these clades includes one species of Mazama. The Wrst clade
links Mazama americana to the genus Odocoileus (PP D1;
BPB/ML D100 in the combined analysis), i.e., two taxa found
in South and North America. The second clade includes all
taxa restricted to South America, i.e., Mazama gouazoubira,
Blastocerus, Hippocamelus, and Pudu (PP D1; BPB/ML D99/
98).
This topology is in disagreement with the morphologi-
cal phylogeny of Webb (2000), which proposes the exis-
tence of two tribes: the tribe Rangiferini, which includes
Rangifer, Hippocamelus, and Pudu; and the tribe Odoco-
ileini, which groups together all other American genera,
i.e., Mazama, Ozotoceros, Blastocerus, and Odocoileus.
However, the conclusions of Webb (2000) are critical for
three reasons: (1) only Capreolus was used as outgroup to
polarize the transformations of character states, (2) most
of the characters supporting the two tribes are quantita-
tive (antlers laterally compressed, stylohyoid cup
enlarged, reduced P2, enlarged bullae), and (3), the loss of
upper canines is considered to be a synapomorphy of the
tribe Odocoileini sensu Webb (2000), whereas our analy-
ses show that this character state is plesiomorphic in the
family Cervidae (see paragraph 3.3).
As discussed in Eisenberg (2000) and Medellin et al.
(1998), the taxonomy of the genus Mazama is very con-
fusing and no comprehensive study has been published
on inter-speciWc relationships. Strikingly, the mitochon-
drial genes analyzed in this study strongly suggest that M.
americana and M. gouazoubira are not closely related,
thereby implying the polyphyly of the genus Mazama
(Figs. 1 and 2). The monophyly of Mazama has never
been questioned on the basis of morphological charac-
ters. However, M. americana and M. gouazoubira possess
112 C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117
several diVerences concerning coat color (reddish brown
versus grayish brown), antler characteristics (rugose ver-
sus ridged) (Medellin et al., 1998), and body size, as M.
gouazoubira is smaller than M. americana (shoulder
height: 55–60 versus 60–70 cm; weight: 22 (average) ver-
sus 29–35 kg) (Eisenberg, 2000). In addition, their karyo-
types show large diVerences in diploid numbers: 2nD42–
53 in M. americana, and 2nD70 in M. gouazoubira
(Eisenberg, 2000). However, the two species share a very
similar morphotype characterized by small body size
(minimum shoulder height <65 cm) and the possession of
simple one-tined antlers, among other similarities.
Assuming that Mazama is polyphyletic, our morphologi-
cal analyses show that this morphotype must have
evolved convergently in the two taxa (Fig. 4). If it is true,
the polyphyly of Mazama would constitute a striking case
of morphological convergence within mammals. None-
theless, before drawing deWnitive conclusions, this
hypothesis needs to be corroborated by a more compre-
hensive analysis including all the six species of the genus
Mazama (Grubb, 1993), and new nuclear markers provid-
ing more information on the relationships between Amer-
ican species.
4.2. New classiWcation (Fig. 4)
Family Cervidae Goldfuss, 1820:374. Deer
Subfamily Cervinae Goldfuss, 1820:374 [ D Plesiometa-
carpalia Brooke, 1878].
Tribe Cervini Goldfuss, 1820:374.
-Cervus Linnaeus, 1758:66
[including Elaphurus Milne-Edwards, 1866:1090–1091].
-Axis C.H. Smith in GriYth, Smith and Pidgeon,
1827:312–313.
- Rucervus Hodgson, 1838:154.
-Dama Frisch, 1775:table.
Tribe Muntiacini Pocock, 1923:207.
- Muntiacus RaWnesque, 1815:56.
- Elaphodus Milne-Edwards, 1871:93.
Subfamily Capreolinae Brookes, 1828:62 [ D Telemeta-
carpalia Brooke, 1878].
Tribe Alceini Brookes, 1828:61.
- Alces Gray, 1821:307.
Tribe Capreolini Brookes, 1828:62.
- Capreolus Gray, 1821:307.
- Hydropotes Swinhoe, 1870:264.
Tribe Odocoileini Pocock, 1923:206.
- Blastocerus Gray, 1850: 68
- Hippocamelus Leuckart, 1816:24.
- Mazama RaWnesque, 1817:363.1
- Odocoileus RaWnesque, 1832:109.
- Ozotoceros Ameghino, 1891:243.2
- Pudu Gray, 1852:242.
- Rangifer, C. H. Smith, in GriYth, Smith and Pidgeon,
1827: 304.
4.3. Calibration points and date estimates
In a Wrst approach, divergence times were estimated by
using two calibration points corresponding to the Wrst
appearance of Muntiacini and Cervidae in the fossil record,
at 8 §1 MYA and 20 §2 MYA, respectively. The calibration
point Cervidae (20§2 MYA) was previously used in several
studies (Douzery and Randi, 1997; Hassanin and Douzery,
2003; Ludt et al., 2004; Randi et al., 1998, 2001), and refers to
the oldest antlers found in the lower Miocene of Eurasia,
with the genera ÐLagomeryx and ÐProcervulus (Ginsburg,
1988). The dates obtained are inconsistent with the paleobi-
ogeographic data (Fig. 3): 11.5–15.5 MYA for the tribe
Odocoileini, 9.1–12.8 MYA for the American clade, and 7.3–
10.8 MYA for the South American clade, whereas the oldest
fossils of Cervidae in the New World were found in North
America at the Miocene/Pliocene boundary, at around 5
MYA, with ÐBretzia and ÐEocoileus (Fry and Gustafson,
1974; Webb, 2000). Thus, these dates would imply a huge gap
in the fossil record of North America. Such a gap seems
unlikely given that the fossil mammalian fauna of North
America is one of the best documented in the world (Stehli
and Webb, 1985). Similarly, the divergence times obtained
for the tribe Cervini are older than the ages suggested by the
fossil record (Di Stefano and Petronio, 2002): 8–11.9 MYA
for Cervini, whereas they Wrst appear at the Mio-Pliocene
boundary; 4.3–6.8 MYA for Cervus and 5.5–8.7 MYA for
Axis, whereas these two genera appear in the Early Pleisto-
cene. In the same way, coupling the calibration points Cervi-
dae (20 §2 MYA) and Odocoileini (5 §1 MYA) yields a
date between 10.6 and 15.2 MYA for the origin of Muntia-
cini, which is older than the Wrst fossils of Muntiacini (7–9
MYA) (Dong et al., 2004) (Fig. 3). The dates obtained by cal-
ibrating the age of the family Cervidae at 20 MYA would
thus imply a large gap in the fossil record or numerous misin-
terpretations of older fossils. These results suggest therefore
that the fossils used as calibration point for the node Cervi-
dae, i.e., ÐLagomeryx and ÐProcervulus, are not closely
related to extant Cervidae. Interestingly, their taxonomic sta-
tus is controversial, Wrst because the shape of their antlers is
singular among fossil deer, and second because it is not sure
whether they were deciduous or not (Azanza, 1993). These
two issues have caused controversy regarding their system-
1More data are needed to draw deWnitive conclusion about the genus
Mazama. Pending a comprehensive revision of Mazama, this classiWcation
provisionally regards the genus as it is currently deWned, i.e., including six
species (Grubb, 1993).
2Although the genus Ozotoceros was not included in the present study,
three arguments suggest that it belongs to the tribe Odocoileini: Wrst, its
forefoot is typically telemetacarpalian; second, its skull shows the synapo-
morphy of Odocoileini, i.e., the choana is divided by an extension of the vo-
mer; and third, its geographic distribution limited to Argentina, Bolivia,
Brazil, Paraguay and Uruguay (Grubb and Gardner, 1998) suggests that it
belongs to the South American clade, as proposed in previous classiWcations
(Brooke, 1878; Grubb, 1993; McKenna and Bell, 1997; Simpson, 1945).
C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117 113
atic position within ruminants (Vislobokova et al., 1989):
they have been placed in the family Cervidae (Chow and
Shih, 1978; McKenna and Bell, 1997), in their own family
ÐLagomerycidae (Pilgrim, 1941), and they have been related
to the family GiraYdae (e.g., Simpson, 1945; Stirton, 1944).
In contrast to ÐLagomeryx and ÐProcervulus, our two
other calibration points refer to fossils whose taxonomic
positions are based on unambiguous morphological charac-
ters: the synapomorphy of the tribe Odocoileini, i.e., the
presence of a vomerine septum dividing the choana, appears
in the Early Pliocene with ÐPavlodaria (Vislobokova, 1980)
and the oldest antlers of muntjacs appear in the Late Mio-
cene (between 7 and 9 MYA) (Dong et al., 2004). The use of
these two calibration points produced date estimates that
agree with the palaeontological data available for Cervidae
(Fig. 3). For instance, we found 4.2–5.7 MYA for the Ameri-
can clade; 3.4–4.9 MYA for the South American clade; 4.2–
6.0 MYA for the tribe Cervini; and 1.9–3.4 MYA and 2.3–
3.7 MYA for the genera Axis and Cervus (as deWned in the
classiWcation proposed in paragraph 4.2) (Table 2). We
therefore base the rest of the discussion on the date estimates
obtained with this combination of calibration points.
4.4. Origin and diversiWcation of deer in the Late Miocene of
Asia
Both the fossil record and geographic distribution of
Cervinae suggest that this group arose in the Miocene of
Asia. The oldest remains of Cervinae were found in Central
Asia, at the Mio/Pliocene boundary for the tribe Cervini
with ÐCervocerus novorossiae (Di Stefano and Petronio,
2002), and during the Late Miocene for the tribe Muntia-
cini with ÐMuntiacus leilaoensis (Dong et al., 2004). In addi-
tion, all extant species of Cervinae are found only in Asia,
with the exception of Cervus elaphus and Dama dama.
However, three arguments support the hypothesis that Cer-
vus and Dama recently dispersed to Europe (and, in the case
of Cervus elaphus, North America): (1) these two species are
also distributed in Asia; (2) they appear within a clade of
exclusively Asian species; (3) the phylogeographic study of
Ludt et al. (2004) concluded that Cervus originated in Cen-
tral Asia.
Although the subfamily Capreolinae is now found in Eur-
asia and America, the fossil record suggests that it diversiWed
in Central Asia, and dispersed thereafter to America. Indeed,
the three tribes of Capreolinae emerged in Central Asia with
ÐProcapreolus during the Miocene/Pliocene boundary for
Capreolini (Di Stefano and Petronio, 2002), ÐPavlodaria dur-
ing the Early Pliocene of North Eastern Kazakhstan for
Odocoileini (Vislobokova, 1980), and ÐCervalces and Alces
during the Pliocene for Alceini (Breda and Marchetti, 2005;
Heintz and Poplin, 1981; Kahlke, 1990). As both the subfam-
ilies Cervinae and Capreolinae have an Asian origin, our
data support the hypothesis that the family Cervidae origi-
nated in Asia. Our dating estimates favor a Late Miocene
origin for deer (7.7–9.6 MYA), which reduces the age usually
assumed for the family Cervidae by more than half (Douzery
and Randi, 1997; Hassanin and Douzery, 2003; Ludt et al.,
2004; Randi et al., 1998, 2001).
The origin of the Cervidae and their tribal diversiWcation
occurred during the Late Miocene of Asia. This period was
characterized by dramatic changes in the environment and
landscapes of Asia. The pulse in the uplift of the Tibetan pla-
teau began around 11 MYA, had peak values at 9 MYA, and
lasted until 7.5 MYA (Amano and Taira, 1992). It coincides
with a global increase in seasonality and aridity (Flower and
Kennett, 1994), which resulted in the spread of grasslands in
Asia and East Africa (Cerling et al., 1997; Morgan et al.,
1994; Quade and Cerling, 1995). It is striking to note that
many other groups of ruminants diversiWed in Asia during
the Late Miocene, including Caprini sensu lato (goats, sheep,
and allies), Boselaphini (species related to the nilgai and four-
horned antelope), and Bubalina (buValoes) (Hassanin and
Ropiquet, 2004; Ropiquet and Hassanin, 2005a,b). As all
these groups include browser/grazer species, the competition
resulting from their overlapping diversiWcations in Asia must
have played a key role in the evolution of Cervidae.
4.5. Colonization of America
Given that the two subfamilies of Cervidae originated in
Asia, their presence in America must be interpreted by one
or more dispersal events from Asia across Beringia. The
fossil record indicates that deer did not enter North Amer-
ica until the latest Miocene (Webb, 2000). Two distinct
odocoileine genera appeared at around 5 MYA: ÐEocoileus
in Florida, and ÐBretzia in northeastern Nebraska. Their
close aYnities with the Early Pliocene ÐPavlodaria from
North Eastern Kazakhstan (Vislobokova, 1980) suggest
that they were recent immigrants to the New World (Webb,
2000). This scenario is supported by our molecular dating,
as the common ancestor of the American Odocoileini is
dated between 4.2 and 5.7 MYA. This implies that it was
possible to cross Beringia during the latest Miocene. The
paleontological data support this hypothesis, as Camelidae
dispersed from North America to Asia between 6.3 and 5.8
MYA (Van Der Made et al., 2002). These dispersals could
T
a
bl
e
2
Divergence times estimated within the family Cervidae
Age MYA (SD) 95% Cred. Int.
M. americana + Odocoileus 2.2 (§0.3) 1.6–2.8
Cervus 2.9 (§0.3) 2.3–3.7
Dama 3 (§0.4) 2.2–4
Axis 2.6 (§0.4) 1.9–3.4
South American clade 4.1 (§0.4) 3.4–4.9
American clade 4.9 (§0.4) 4.2–5.7
Cervini 5 (§0.5) 4.2–6
Capreolini 5.6 (§0.6) 4.5–6.9
Odocoileini 5.8 (§0.2) 5.3–6
Muntiacini 7.3 (§0.3) 7–8
Alceini+Capreolini 7.4 (§0.5) 6.4–8.4
Odocoileinae 7.8 (§0.4) 6.9–8.7
Cervinae 7.9 (§0.4) 7.2–8.9
Cervidae 8.5 (§0.5) 7.7–9.6
114 C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117
have been favored by the spread of increasingly open and
drier habitats during the Late Miocene (Cerdeño, 1998).
Since Odocoileini occurred early in America, they may
have had time to diversify in diVerent habitats, and some
species may have developed tolerance to warm and damp
environments, allowing their dispersal into the Neotropics.
The deer found in South America exhibit a wide range of
morphological variation and live in diVerent ecological
habitats, which is characteristic of an adaptive radiation:
Mazama occurs in tropical forests, Blastocerus lives in
marshy areas, Hippocamelus and Pudu dwell in the Chilean
Andes and Ozotoceros inhabits the Pampas (Nowak, 1999).
Moreover, a great diversity of fossils suddenly appears at
the Plio-/Pleistocene boundary of South America with at
least four diVerent genera, i.e., ÐAntifer, ÐEpieuryceros,
ÐMorenelaphus, and ÐParaceros (Castellanos, 1945; Krag-
lievich, 1932). This radiation is traditionally explained by
the arrival of their common ancestor in South America at
3–3.5 MYA, i.e., after the formation of the Isthmus of Pan-
ama (Eisenberg, 1987; Marshall et al., 1979; Stehli and
Webb, 1985). Our molecular analyses suggest, however,
that the evolution of deer in America during the Pliocene
was more complex. Indeed, the genus Mazama is found to
be polyphyletic: M. americana is allied with Odocoileus,
whereas M. gouazoubira is grouped with the genera
endemic to South America, i.e., Blastocerus, Hippocamelus,
and Pudu. Our interpretation is that South America was
colonized at least twice: a Wrst time by the ancestor of the
South American clade in the Early Pliocene, and a second
time, by both M. americana and O. virginianus at the Plio-
Pleistocene boundary. In agreement with this scenario, the
common ancestor of the species endemic to South America
is dated between 3.4–4.9 MYA, which is compatible with
the completion of the Pliocene land bridge. The fossil
record indicates that the dwarf forms related to the genera
Pudu and Mazama evolved as small-bodied Neotropic
forms during the Plio-Pleistocene, and descended from
larger Nearctic ancestors possessing more developed ant-
lers (Eisenberg, 2000; Webb, 2000). Our morphological
inferences support this hypothesis, but they suggest that
similar adaptations were acquired independently in M.
americana, and in the brockets and pudus of the South
American clade (Fig. 4).
Although Alces and Rangifer have a wide Holarctic dis-
tribution today, their presence in America probably
resulted from a recent dispersal event during the Early
Pleistocene. The diversiWcation of Alceini in Asia and the
subsequent Pleistocene dispersal of Alces to North America
are well documented by the fossil record (Breda and
Marchetti, 2005; Heintz and Poplin, 1981; Kahlke, 1990).
Such evidence, however, is lacking for Rangifer, as no fos-
sils are known that predate the Pleistocene. But as Rangifer
is also a specialist of the Arctic tundra, we suppose that it
dispersed to America during the Pleistocene, i.e., at the
same epoch than Alces and the three bovid genera Oream-
nos, Ovibos, and Ovis. According to our analyses, the
lineages leading to Alces and Rangifer diverged from other
Capreolinae at around 6.4–8.4 and 5.3–6.0 MYA, respec-
tively. Because fossils of these two genera are not found
before the Pliocene for Alces, and the Pleistocene for Rang-
ifer, an important gap is inferred for the fossil record in
Asia.
4.6. Evolution of sexual dimorphism
In the various species of deer, males can be generally dis-
tinguished from females by the presence of antlers and/or
tusk-like upper canines, larger body size and mass, and/or
diVerences in coat coloration. Antlers are made of a decidu-
ous bony core covered by velvet skin, which fully regener-
ates each year from the permanent pedicles (Lie and Suttie,
2001). Structurally, they diVer from all other cranial
appendages found in ruminants (Scott and Janis, 1987) and
can thus be considered as an autapomorphy of the family
Cervidae. Three main reasons have been proposed for
explaining the origin of antlers in male deer: defense against
predators, display structures to be appreciated by females,
or weapons that serve during intermale Wghts for territory
and/or access to mating with more than one female (Jar-
man, 2000). Our analyses suggest that the common ancestor
of Cervidae lived in open habitats, that females were antler-
less, and that males were large (shoulder height > 650 mm),
bigger than females, with three-tined antlers, but without
enlarged upper canines. Interestingly, closed habitats have
been independently colonized by Capreolini, Muntiacini,
Mazama americana, Mazama gouazoubira, and Pudu, and
all these taxa have developed similar morphological adap-
tations, including the reduction and simpliWcation of ant-
lers in males (absent in Hydropotes, or with one or two tines
rather than three or four in other species), and the acquisi-
tion of a small body size, accompanied by sexual weight
monomorphism. We can infer that these reductions were
selected positively because (1) large males with long and
ramiWed antlers are expected to move much more slowly in
closed habitats, and (2) the display function of antlers is
expected to be much less eVective in habitats where visibil-
ity is considerably reduced. In parallel with these reduc-
tions, males of Hydropotes and Muntiacini have acquired
tusk-like upper canines, which are used as display orna-
ments and weapons for Wghting with their congeners during
the rut (Cooke and Farrell, 1998; Hutchins et al., 2004).
Thus, canines clearly replace antlers in the sexual behaviour
of these species. This supports the hypothesis that the pri-
mary role of antlers is for use in sexual competition during
the breeding season, when males Wght each other to gain
access to females in estrus. This important function may
explain why antlers have been maintained in all cervids
except Hydropotes. InterspeciWc diVerences in the morphol-
ogy of antlers may have occurred because of the evolution
of divergent Wghting behaviors (Caro et al., 2003; Geist,
1966; Lundrigan, 1996). As the reproductive success of
males is supposed to be directly correlated with strength
and weapon size (Clutton-Brock, 1989), it is likely that
large males with important armament have been positively
C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117 115
selected during the evolution of Cervidae. This would
explain why antlers are particularly developed and ramiWed
in males of species dwelling in open habitats (Fig. 4).
The only species in which females have acquired antlers
is Rangifer tarandus. As antlers may serve as weapon
against predators (Geist, 1968), and because a high percent-
age of mortality in reindeer is due to wolves and bears
(Crête et al., 2001), predation may have constituted a suY-
cient force for the selection of antlers in females. However,
whereas this predation pressure is even more pronounced
for the genus Alces (Wayne Kuzyk, 2002) and also exists in
the genus Odocoileus, females of these genera do not pos-
sess antlers.
As an alternative hypothesis, we suggest that antlered
females may have been positively selected because of gregari-
ousness. In fact, intraspeciWc competition for food is expected
to be important in the large mixed-sex herds of reindeer,
especially during harsh winters when the snow level is high
for extended periods of time. Schaefer and Mahoney (2001)
actually found that through a 1000km range, the percentage
of antlered females was correlated positively with average
annual snowfall and mean snow depth at the end of March.
These results support the hypothesis that antlers on females
provide functional advantages in interference competition
for winter food (Schaefer and Mahoney, 2001). Antlers in
female reindeer could thus have evolved as an adaptation
linked to intraspeciWc competition for food during winter,
but not to anti-predator defense.
Acknowledgments
We thank Jean-Luc Berthier, Céline Canler, François
CatzeXis, Philippe Chardonnet, Raphaël Cornette, Jacques
Cuisin, Emmanuel Douzery, Mathieu Fritz, Françoise Her-
gueta-Claro, Jacques Rigoulet, William Robichaud, Michel
Tranier, and Vitaly Volobouev for providing tissues or
DNA samples. We also acknowledge Evelyne Bremond-
Hoslet, Jean-Marc Bremond, Pedro Cordeiro and Woody
Cotterill for their help with the bibliography. We acknowl-
edge the two anonymous reviewers for comments on the
manuscript.
Appendix A. Supplementary data
Supplementary data associated with this article can be
found, in the online version, at doi:10.1016/
j.ympev.2006.02.017.
References
Amano, K., Taira, A., 1992. Two-phase uplift of Higher Himalayas since
17 Ma. Geology 20, 391–394.
Amato, G., Egan, M., Schaller, G.B., 2000. Mitochondrial variation in
muntjac: evidence for discovery, rediscovery, and phylogenetic rela-
tionships. In: Vrba, E., Schaller, G.B. (Eds.), Antelopes, Deer, and Rela-
tives. Fossil Record, Behavioral Ecology, Systematics, and
Conservation. Yale University Press, New Haven and London, pp.
285–295.
Azanza, B., 1993. Sur la nature des appendices frontaux des cervidés (Arti-
odactyla, Mammalia) du Miocène inférieur et moyen. Remarques sur
leur systématique et leur phylogénie. C.R. Acad. Sci. Paris Série II. 316,
1163–1169.
Biswas, T., 1999. Habitat utilization by hog deer (Axis porcinus) in relation
to other sympatric species at Jaldapara wildlife sanctury, West Bengal,
India. M.Sc. dissertation abstract. In: Deer Specialist Group, Newslet-
ter 15, available at <http://iibce.edu.uy/citogenetica/deer/DSG-
News14.pdf/>.
Bonnet, A., 2001. Etude chromosomique et moléculaire de la sous famille des
Cervinae. Application des résultats à la gestion des troupeaux en capti-
vité d’espèces menacées. Unpublished PhD thesis. Université de Tours.
Bouvrain, G., Geraads, D., Jehenne, Y., 1989. Nouvelles données relatives
à la classiWcation des Cervidae (Artiodactyla, Mammalia). Zool. Anz.
223, 82–90.
Breda, M., Marchetti, M., 2005. Systematical and biochronological review
of Plio-Pleistocene Alceini (Cervidae; Mammalia) from Eurasia. Qua-
ternary Sci. Rev. 24, 775–805.
Brooke, V., 1878. On the classiWcation of the Cervidae, with a synopsis of
the existing species. Proc. Zool. Soc. Lond. 1878, 883–928.
Caro, T.M., Graham, C.M., Stoner, C.J., Flores, M.M., 2003. Correlates of
horn and antler shape in bovids and cervids. Behav. Ecol. Sociobiol. 55,
32–41.
Castellanos, A., 1945. Notas sobre algunas cornamentas de ciervos de los
géneros Epieuryceros, Antifer y Paraceros y descripcion de la de Epieu-
ryceros proximus n. sp. Publicaciones del Instituto de WsiograWa y geo-
logia. 24, 4–21.
Cerdeño, E., 1998. Diversity and evolutionary trends of the family Rhinoc-
erotidae (Perissodactyla). Palaeogeogr. Palaeoclimatol. Palaeoecol.
141, 13–34.
Cerling, E., Harris, J.M., MacFadden, B.J., Leakey, M.G., Quade, J., Eisen-
mann, V., Ehleringer, J.R., 1997. Global vegetation change through the
Miocene/Pliocene boundary. Science 389, 153–158.
Chow, B., Shih, M., 1978. A skull of Lagomeryx from the middle Miocene
of Linchu, Shangtung. Vert. Palasiatica 16, 111–122.
Clutton-Brock, T.H., 1989. Mammalian mating systems. Proc. R. Soc.
Lond. 236, 339–372.
Cooke, A., Farrell, L., 1998. Chinese Water Deer. The Mammal Society
and British Deer Society, Fordingbridge, UK.
Crête, M., Ouellet, J.P., Lesage, L., 2001. Comparative eVects on plants of
caribou/reindeer, moose and white-tailed deer herbivory. Artic. 54 (4),
407–417.
Cronin, M.A., Stuart, R., Pierson, B.J., Patton, J.C., 1996. K-casein gene
phylogeny of higher Ruminants (Pecora, Artiodactyla). Mol. Phyloge-
net. Evol. 6, 295–311.
Di Stefano, G., Petronio, C., 2002. Systematics and evolution of the Eur-
asian Plio-Pleistocene tribe Cervini (Artiodactyla, Mammalia). Geol.
Romana. 36, 311–334.
Dong, W., Pan, Y., Liu, J., 2004. The earliest Muntiacus (Artiodactyla,
Mammalia) from the late Miocene of Yuanmou, southwestern China.
C.R. Palevol. 3, 379–386.
Douzery, E.J.P., Randi, E., 1997. The mitochondrial control region of Cer-
vidae: evolutionnary patterns and phylogenetic content. Mol. Biol.
Evol. 14, 1154–1166.
Eisenberg, J.F., 1987. The evolutionary history of the Cervidae with special
reference to the South American radiation. In: Wemmer, C.M. (Ed.),
Biology and Management of the Cervidae. Smithsonian Institution
Press, Washington, London, pp. 60–64.
Eisenberg, J.F., 2000. The contemporary Cervidae of South America. In:
Vrba, E., Schaller, G.B. (Eds.), Antelopes, Deer, and Relatives. Fossil
record, Behavioral Ecology, Systematics, and Conservation. Yale Uni-
versity Press, New Haven and London, pp. 189–202.
Felsenstein, J., 2004. PHYLIP (PHYLogeny Inference Package) version
3.6b. Distributed by the author. Department of Genome Sciences, Uni-
versity of Washington, Seattle.
Flower, B.P., Kennett, J.P., 1994. The middle Miocene climatic transition:
East Antarctic ice sheet development, deep ocean circulation and
global carbon cycling. Palaeogeogr. Palaeoclimatol. Palaeoecol.
116 C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117
Fontana, F., Rubini, M., 1990. Chromosomal evolution in Cervidae. Bio-
systems 24, 157–174.
Frick, C., 1937. Horned ruminants of North America. Bull. Amer. Mus.
Nat. Hist. 69, 1–669.
Fry, W.E., Gustafson, E.P., 1974. Cervids from the Pliocene of central
Washington. J. Paleontol. 48, 375–386.
Garrod, A.H., 1877. Notes on the visceral anatomy and osteology of
the ruminants, with a suggestion regarding a method of expressing
the relation of species by means of formulae. Proc. Zool. Soc. Lond.
2–19.
Geist, V., 1966. The evolution of horn-like organs. Behaviour 27, 175–214.
Geist, V., 1968. Horn-like structure as rank symbols, guards and weapons.
Nature 220, 813–814.
Geist, V., Bayer, M., 1988. Sexual dimorphism in the Cervidae and its rela-
tion to habitat. J. Zool. Lond. 214, 45–53.
Geist, V., 1998. Deer of the world. Their evolution, behavior and ecology.
Stackpole books.
Gentry, A.W., 1994. The Miocene diVerentiation of old world Pecora
(Mammalia). Hist. Biol. 7, 115–158.
Ginsburg, L., 1988. La faune des mammifères des sables Miocènes du syn-
clinal d’Esvres (Val de Loire). C.R. Acad. Sci. Paris. Série II 307, 319–
322.
Groves, C.P., Grubb, P., 1987. Relationships of living deer. In: Wemmer,
C.M. (Ed.), Biology and Management of the Cervidae. Smithsonian
institution press, Washington and London, pp. 21–59.
Grubb, P., 1993. Mammal Species of the World: A Taxonomic and Geo-
graphic Reference. In: Wilson, D.E., Reeder, D.M. (Eds.), Smithsonian
Institution Press, Washington and London, pp. 384–392.
Grubb, P., Gardner, A.L., 1998. List of species and sub-species of the fami-
lies Tragulidae, Moschidae, and Cervidae. In Wemmer C., IUCN/SSC
Deer Specialist Group (Eds.), Deer: status survey and conservation
action plan. Chapter 2. Available at: <http://www.iucn.org/themes/ssc/
pubs/sscaps.htm/>.
Guindon, S., Gascuel, O., 2003. A Simple, Fast, and Accurate Algorithm to
Estimate Large Phylogenies by Maximum Likelihood. Syst. Biol. 52,
696–704.
Hall, T., 2001. BioEdit version 5.0.6. Available at: <http://www.mbio.
ncsu.edu/BioEdit/bioedit.html/>.
Hassanin, A., Douzery, E., 1999. Evolutionary aYnities of the enigmatic
Saola (Pseudoryx nghetinhensis), in the context of the molecular phy-
logeny of Bovidae. Proc. R. Soc. Lond. 266, 893–900.
Hassanin, A., Douzery, E.J.P., 2003. Molecular phylogeny of Ruminan-
tia and the alternative position of the Moschidae. Syst. Biol. 52,
206–228.
Hassanin, A., Pasquet, E., Vignes, J.-D., 1998. Systematic relationships
within the subfamily Caprinae (Artiodactyla, Bovidae) as determined
from cytochrome b sequences. J. Mammal. Evol. 5, 217–236.
Hassanin, A., Ropiquet, A., 2004. Molecular phylogeny of the tribe Bovini
(Bovinae, Bovini) and the taxonomic status of the kouprey, Bos sauveli
Urbain, 1937. Mol. Phylogenet. Evol. 36, 154–168.
Heintz, E., Poplin, F., 1981. Alces carnutorum (Laugel, 1862) du Pléis-
tocène de Saint-Prest (France). Systématique et evolution des Alcinés
(Cervidae, Mammalia). Quartärpaläontologie 4, 105–122.
Hodgson, B.H., 1838. Ann. N.H. 1, 154.
Honeycutt, R.L., Nedbal, M.A., Adkins, R.M., Janecek, L.L., 1995. Mam-
malian mitochondrial DNA evolution: a comparison of the cyto-
chrome b and cytochrome c oxidase II genes. J. Mol. Evol. 40, 260–272.
Huelsenbeck, J.P., Ronquist, F., 2001. MrBayes: Bayesian inference of
phylogenetic trees. Bioinformatics 17, 754–755.
Hutchins M., Thoney D.A. and McDade M.C. 2004. Eds. Grzimek’s Ani-
mal Life Encyclopedia. 2nd ed. Detroit: Gale, pp. 344–355; 373–377.
Janis, C.M., Scott, K.M., 1987. The interrelationships of higher ruminant
families with special emphasis on the members of the Cervoidea. Bull.
Am. Mus. Nat. Hist. 2893, 1–85.
Jarman, P.J., 2000. Dimorphism in social Artiodactyla: selection upon
females. In: Vrba, E., Schaller, G.B. (Eds.), Antelopes, Deer, and Rela-
tives. Fossil Record, Behavioral Ecology, Systematics, and Conserva-
tion. Yale University Press, New Haven and London, pp. 171–179.
Kahlke, H.D., 1990. On the evolution, distribution and taxonomy of fossil
Elk/Moose. Quartärpaläontologie 8, 83–106.
Kraglievich, L., 1932. Contribucion al conocimiento de los ciervos fosiles
del Uruguay. An. Mus. Hist. Nat. Montevideo. 2, 355–439.
Lie, C., Suttie, J.M., 2001. Deer antlerogenic periosteum: a piece of
postnatally retained embryonic tissue? Anat. Embryol. 204,
375–388.
Liu, X.-H., Wang, Y.-Q., Liu, Z.-Q., Zhou, K.-Y., 2003. Phylogenetic rela-
tionships of Cervinae based on sequence of mitochondrial cytochrome
b gene. Zool. Res. 24, 27–33.
Loison, A., Gaillard, J.M., Pélabon, C., Yoccoz, N.G., 1999. What factors
shape sexual size dimorphism in ungulates? Evol. Ecol. Res. 1, 611–633.
Lundrigan, B., 1996. Morphology of horns and Wghting behavior in the
family Bovidae. J. Mammal. 77, 462–475.
Ludt, C.J., Schroeder, W., Rottmann, O., Kuehn, R., 2004. Mitochondrial
DNA phylogeography of red deer (Cervus elaphus). Mol. Phylogenet.
Evol. 31, 1064–1083.
Marshall, L.G., Butler, R.F., Drake, R.E., Curtis, G.H., Tedsford, R.H., 1979.
Calibration of the great American Interchange. Science 204, 272–279.
Matthee, C.A., Davis, S.K., 2001. Molecular insights into the evolution of
the family Bovidae: a nuclear DNA perspective. Mol. Biol. Evol. 18,
1220–1230.
McKenna, M.C., Bell, S., 1997. ClassiWcation of Mammals Above the Spe-
cies Level. Columbia University press, New York.
Medellin, R.A., Gardner, A.L., Aranda, J.M., 1998. The Taxonomic status
of the Yucatan brown brocket, Mazama Pandora (Mammalia: Cervi-
dae). Proc. Biol. Soc. Washington. 111, 1–14.
Meijaard, E., Groves, C.P., 2004. Morphometrical relationships between
South-east Asian deer (Cervidae, tribe Cervini): evolutionary and bio-
geographic implications. J. Zool. Lond. 263, 179–196.
Merino, M.L., Milne, N., Vizcaino, S.F., 2005. A cranial morphometrics
study of deer (Mammalia, Cervidae) from Argentina using three-dime-
sional landmarks. Act. Ther. 50 (1), 91–108.
Miyamoto, M.M., Allard, M.W., Adkins, R.M., Janecek, L.L., Honeycutt,
R.L., 1994. A congruence test of reliability using linked mitochondrial
DNA sequences. Syst. Biol. 43, 236–243.
Mooring, M.S., Blumstein, D.T., Stoner, C.J., 2004. The evolution of para-
site-defence grooming in ungulates. Biol. J. Lin. Soc. 81, 17–37.
Morgan, M.E., Kingston, J.D., Marino, B.D., 1994. Carbon isotopic evi-
dence for the emergence of C4 plants in the Neogene from Pakistan
and Kenya. Nature 367, 162–165.
Nowak, R.M., 1999. Walker’s, Mammals of the World, sixth Ed. Johns
Hopkins University Press, Baltimore and London. pp. 1091–1132.
Nylander, J.A.A., 2004. MrModeltest v2. Program distributed by the
author. Evolutionary Biology Centre, Uppsala University.
Pilgrim, G., 1941. The dispersal of artiodactyla. Biol. Rev. 16, 134–163.
Pitra, C., Fickel, J., Meijaard, E., Groves, C.P., 2004. Evolution of old
world deer. Mol. Phylogenet. Evol. 33, 880–895.
Pocock, R.I., 1910. On the specialized cutaneous glands of ruminants.
Proc. Zool. Soc. Lond. 840–946.
Pocock, R.I., 1923. On the external characters of Elaphurus, Hydropotes,
Pudu and other Cervidae. Proc. Zool. Soc. Lond. 181–207.
Pocock, R.I., 1933. The homologies between the branches of the antlers of
the Cervidae based on the theory of dichotomous growth. Proc. Zool.
Soc. Lond. 377–406.
Quade, J., Cerling, T.E., 1995. Expansion of C4 grasses in the Late Mio-
cene of Northern Pakistan: evidence from stable isotopes in paleosols.
Palaeogeogr. Palaeoclimatol. Palaeoecol. 115, 91–116.
Randi, E., Mucci, N., Pierpaoli, M., Douzery, E.J.P., 1998. New phylo-
genetic perspectives on the Cervidae (Artiodactyla) are provided
by the mitochondrial cytochrome b gene. Proc. R. Soc. Lond. 265,
793–801.
Randi, E., Mucci, N., Claro-Hergeta, F., Bonnet, A., Douzery, E.J.P., 2001. A
mitochondrial DNA control region phylogeny of the Cervinae: specia-
tion in Cervus and implications for conservation. Anim. Conserv. 4, 1–11.
Ropiquet, A., Hassanin, A., 2005a. Molecular evidence for the polyphyly
of the genus Hemitragus (Mammalia, Bovidae). Mol. Phylogenet. Evol.
36, 154–168.
C. Gilbert et al. / Molecular Phylogenetics and Evolution 40 (2006) 101–117 117
Ropiquet, A., Hassanin, A., 2005b. Molecular phylogeny of caprines
(Bovidae, Antilopinae): the question of their origin and diversiWcation
during the Miocene. J. Zool. Syst. Evol. Res. 43, 49–60.
Schaefer, J.A., Mahoney, S.P, 2001. Antlers on female caribou: biogeogra-
phy of the bones of contention. Ecology 82, 3556–3560.
Scott, K.M., Janis, C.M., 1987. Phylogenetic relationships of the Cervidae,
and the case for a superfamily “Cervoidea”. In: Wemmer, C.M. (Ed.),
Biology and Management of the Cervidae. Smithsonian Institution
Press, Washington, London, pp. 3–20.
Simpson, G.G., 1945. The principles of classiWcation and a classiWcation of
mammals. Bull. Am. Mus. Nat. His. 85, 1–350.
Stehli, F.G., Webb, D.S., 1985. The Great American Biotic Interchange.
Plenum Press, New York and London.
Stirton, R., 1944. Comments and relationships of the Cervoid family Pala-
eomerycidae. Am. J. Sci. 242, 643–645.
SwoVord, D.L., 2003. PAUP*. Phylogenetic Analysis Using Parsimony
(* and Other Methods). Version 4. Sinauer Associates. Sunderland,
Massachussetts.
Thorne, J.L., Kishino, H., 2002. Divergence time and evolutionary rate
estimation with multilocus data. Syst. Biol. 51, 689–702.
Thorne, J.L., Kishino, H., Painter, I.S., 1998. Estimating the rate of molecu-
lar evolution. Mol. Biol. Evol. 15, 1647–1657.
Van Der Made, J., Morales, J., Sen, S., Aslan, F., 2002. The Wrst camel from
the Upper Miocene of Turkey and the dispersal of the camels into the
Old World. C.R. Palevol. 1, 117–122.
Vislobokova, I., 1980. The systematic position of a deer from Pavlodar and
the origin of neocervinae. Paleontology J. 3, 97–111.
Vislobokova, I., Changkang, H., Sun, B., 1989. On the systematic position
of the Lagomerycinae. Vert. Palasiatica 27, 130–132.
Wang, W., Lang, H., 2000. Rapid and Parallel Chromosomal Number
Reductions in Muntjac Deer Inferred from Mitochondrial DNA Phy-
logeny. Mol. Biol. Evol. 17, 1326–1333.
Wayne Kuzyk, G., 2002. Wolf distribution and movements on caribou
ranges in West-Central Alberta. MSc. Thesis. University of Alberta.
Webb, D.S., 2000. Evolutionary history of New World Cervidae. In: Vrba,
E., Schaller, G.B. (Eds.), Antelopes, Deer, and Relatives. Fossil Record,
Behavioral Ecology, Systematics, and Conservation. Yale University
Press, New Haven and London, pp. 38–64.
Winnepenninckx, B., Backeljau, T., Dewachter, R., 1993. Extraction of
high molecular weight DNA from mollusc. Trends Genet. 9, 407.
Yang, Z., Yoder, A., 2003. Comparison of likelihood and Bayesian meth-
ods for estimating divergence times using multiple gene loci and cali-
bration points, with application to a radiation of cute mouse lemur
species. Syst. Biol. 52, 705–716.
