1882
q 2001 The Society for the Study of Evolution. All rights reserved.
Evolution, 55(9), 2001, pp. 1882?1892
PHYLOGEOGRAPHY OF THE ASIAN ELEPHANT (ELEPHAS MAXIMUS)
BASED ON MITOCHONDRIAL DNA
ROBERT C. FLEISCHER,1,2 ELIZABETH A. PERRY,3 KASINATHAN MURALIDHARAN,4 ERNEST E. STEVENS,5 AND
CHRISTEN M. WEMMER5
1Molecular Genetics Laboratory, Conservation and Research Center, Smithsonian Institution, 3001 Connecticut Avenue, NW,
Washington, DC 20008
2E-mail: fleischer.robert@nmnh.si.edu
4Department of Pediatrics, Division of Genetics, Emory University, School of Medicine, Atlanta, Georgia 30322
5Conservation and Research Center, National Zoological Park, Smithsonian Institution, 1500 Remount Road,
Front Royal, Virginia 22630
Abstract. Populations of the Asian elephant (Elephas maximus) have been reduced in size and become highly frag-
mented during the past 3000 to 4000 years. Historical records reveal elephant dispersal by humans via trade and war.
How have these anthropogenic impacts affected genetic variation and structure of Asian elephant populations? We
sequenced mitochondrial DNA (mtDNA) to assay genetic variation and phylogeography across much of the Asian
elephant?s range. Initially we compare cytochrome b sequences (cyt b) between nine Asian and five African elephants
and use the fossil-based age of their separation (;5 million years ago) to obtain a rate of about 0.013 (95% CI 5
0.011?0.018) corrected sequence divergence per million years. We also assess variation in part of the mtDNA control
region (CR) and adjacent tRNA genes in 57 Asian elephants from seven countries (Sri Lanka, India, Nepal, Myanmar,
Thailand, Malaysia, and Indonesia). Asian elephants have typical levels of mtDNA variation, and coalescence analyses
suggest their populations were growing in the late Pleistocene. Reconstructed phylogenies reveal two major clades
(A and B) differing on average by HKY85/G-corrected distances of 0.020 for cyt b and 0.050 for the CR segment
(corresponding to a coalescence time based on our cyt b rate of ;1.2 million years). Individuals of both major clades
exist in all locations but Indonesia and Malaysia. Most elephants from Malaysia and all from Indonesia are in well-
supported, basal clades within clade A, thus supporting their status as evolutionarily significant units (ESUs). The
proportion of clade A individuals decreases to the north, which could result from retention and subsequent loss of
ancient lineages in long-term stable populations or, perhaps more likely, via recent mixing of two expanding populations
that were isolated in the mid-Pleistocene. The distribution of clade A individuals appears to have been impacted by
human trade in elephants among Myanmar, Sri Lanka, and India, and the subspecies and ESU statuses of Sri Lankan
elephants are not supported by molecular data.
Key words. Asian elephant, Elephas maximus, evolutionarily significant units, Loxodonta africana, mitochondrial
DNA, phylogeography, rate calibration.
Received August 3, 2000. Accepted May 9, 2001.
Molecular assessments of phylogeographic structure, in
concert with paleontological and geological information, are
highly valuable for the reconstruction of evolutionary sce-
narios. Molecular analyses of phylogeography can reveal ev-
idence for both ancient and recent demographic events such
as population size changes or dispersal (Slatkin and Mad-
dison 1989; Tajima 1989; Avise 1994; Harpending 1994).
Determination of phylogeographic structure can also be im-
portant for the purposes of conservation management (e.g.,
Dizon et al. 1992; Moritz 1994; Vogler and DeSalle 1994).
In the case of Asian elephants (Elephas maximus), humans
have been responsible both for fragmentation of populations
through habitat change and local extirpation and for poten-
tially high levels of gene flow through partial domestication
and transport of elephants for trade and war (Sukumar and
Santiapillai 1996). In this paper we assess mitochondrial
DNA (mtDNA) variation in Asian elephants from throughout
their remnant range. We combine the analysis with infor-
mation about geological, paleontological, and human history
to address alternative evolutionary scenarios that may explain
the phylogeographic patterns we find. Our results suggest that
most Asian elephant populations we studied are derived from
relatively recent mixing of two formerly allopatric lineages
3 Present address: Aqua Bounty Canada, Inc., 20 Hallett Crescent,
Box 21233, St. Johns, Newfoundland A1A 5B2, Canada.
(perhaps even forms described as different fossil species)
because of natural and human-assisted dispersal. Not all study
populations show evidence of this mixing, and we consider
the genetic evidence for subspecies and evolutionarily sig-
nificant units (ESUs) and discuss the conservation implica-
tions of our findings.
The Asian elephant is the most endangered extant prob-
iscidean, with fewer than 55,000 individuals remaining com-
pared to more than 500,000 African elephants (Loxodonta
africana). Their wild populations were greatly reduced and
fragmented in the late Holocene by the actions of humans
(Sukumar and Santiapillai 1996). The number of wild Asian
elephants was estimated at more than 200,000 as recently as
1900, but current estimates (Sukumar and Santiapillai 1996)
place the wild population at 37,000 to 55,000. An additional
15,000 are in captivity in a semidomesticated status, most of
which do not breed. Because of their long generation time
(minimum of 14?15 years and usually 18?20 years to female
breeding age; Shoshani and Eisenberg 1982; Sukumar 1989)
and low reproductive output (a calf every 2.5 to 8 years;
Shoshani and Eisenberg 1982), population recovery in the
wild will be a slow process (assuming that elephants and
their habitats are suitably protected).
Although subspecies taxonomy of Asian elephants has var-
ied among authors, the most recent treatment (Shoshani and
1883ASIAN ELEPHANT PHYLOGEOGRAPHY
Eisenberg 1982) recognizes three subspecies: E. m. indicus
on the mainland, E. m. maximus in Sri Lanka; and E. m.
sumatranus from Sumatra of Indonesia. These subspecies
designations are based primarily on body size and slight dif-
ferences in coloration and the fact that E. m. sumatranus has
relatively larger ears and one extra pair of ribs (Shoshani and
Eisenberg 1982). The Sri Lankan subspecies designation is
weakly supported by analysis of allozyme loci (Nozawa and
Shotake 1990), but not by analysis of mtDNA sequences
(Hartl et al. 1996; Fernando et al. 2000). No assessment of
the Sumatran subspecies was made in previous molecular
studies.
There is a remarkably complete and detailed fossil record
of the Elephantidae since the late Miocene (Maglio 1973;
Coppens et al. 1978; Todd and Roth 1996). Fossil remains
of the primitive elephantine Primelephas gomphotheroides
have been found from the late Miocene to early Pliocene of
Africa (Maglio 1973). This species likely gave rise to Lox-
odonta and Elephas in the early Pliocene (Maglio 1973; Tassy
and Shoshani 1988), although alternative evolutionary sce-
narios exist (see Todd and Roth 1996). The lineage of mam-
moths (genus Mammuthus) evolved shortly after, probably
from the Loxodonta lineage (Noro et al. 1998; but see Yang
et al. 1996; Ozawa et al. 1997). K-Ar dating of strata asso-
ciated with these fossils reveal that none date prior to about
4.5 million years ago, and Maglio (1973: fig. 15) estimates
the split of Loxodonta and Elephas occurred about 5 million
years ago. The Elephas lineage itself split into one lineage
that remained in Africa (E. recki and derivatives) and at least
one lineage that migrated into Asia about 3 million years ago
(E. planifrons and E. hysudricus and derivatives; Maglio
1973). Elephas maximus likely evolved directly from E. hy-
sudricus, perhaps during the last 0.2 million years (Maglio
1973). A Javan form, E. hysudrindicus, probably split from
the E. hysudricus lineage 0.8?1.0 million years ago.
Asian vegetation and biogeography were greatly impacted
by glaciation during the Pleistocene, and these changes likely
affected the distribution of Elephas. At glacial maxima much
of northern and western India was desert, and because of
lower sea levels, the Sunda peninsula was large and included
Sumatra, Java, and Borneo (Fig. 1). Elephas may have been
separated into refugia in southern India and the Sunda pen-
insula during glacial maxima.
Humans apparently began domestication of Asian ele-
phants about 4000?5000 years ago in the earliest settlements
of the Indus Valley (Tennant 1861; Carrington 1958; Lahiri-
Choudhury 1991). Until the early 19th century, large numbers
of captive animals were moved across national borders, pri-
marily as beasts for war or burden or, more recently, for
timber extraction. The geographic extent and scale of this
trade will never be completely known, but there are records
of transport of literally thousands of elephants, over thou-
sands of kilometers, especially during periods of war (Suk-
umar 1989). For example, one historical trade route existed
between Sri Lanka and India by about 300 B.C., and later, in
medieval times, another developed between Pegu in southern
Myanmar to Sri Lanka and Bengal, and then to the Sultanate
of Northern India (where wild elephants had been largely
extirpated by this time; Digby 1971). Because escapes of
domestic elephants were and continue to be relatively com-
mon occurrences (Ferrier 1947; Stracey 1963), translocations
of elephants are likely to contribute genes to wild popula-
tions. Thus, escapes from captivity may have had major im-
pacts on genetic structure. Natural dispersal occurs via move-
ment of males among matrilineal groups (Sukumar 1989;
Fernando and Lande 2000).
Here we report analyses of mtDNA sequences from Asian
elephant populations across most of their native range. We
calculate a divergence rate for the cytochrome b (cyt b) gene
based on an African and Asian elephant comparison and their
estimated date of ancestry (i.e., 5 million years ago; Maglio
1973; Coppens et al. 1978). We then apply this local rate
calibration to estimate how long ago the Asian elephant
mtDNA lineage split into two divergent clades (this study;
Hartl et al. 1996). We also assess variability within and
among Asian elephant populations and phylogeography for
a much larger sample of individuals by analysing a segment
of mtDNA that includes about half of the control region along
with a small part of cyt b, and two tRNA genes. Last, we
interpret the patterns in light of paleontological, biogeograph-
ical, and human history and address the implications of our
results for the conservation of Asian elephants.
MATERIALS AND METHODS
Samples
All elephants sampled were captives (Table 1), but for
nearly all individuals we know from where in the wild they
originated. Asian elephant samples were collected from el-
ephant work camps or orphanages in rural regions of Asia
(Nepal, Thailand, Indonesia, Sri Lanka), or from zoos, either
in Asia (Myanmar, Malaysia, Thailand) or North America
(i.e., Indian samples; Table 1, Fig. 1). The records for most
samples indicate that individuals are derived from local pop-
ulations. These records showed that (1) Indonesian samples
originated in three localities from across the island of Sumatra
(i.e., Aceh in the far north, Riau in the center, and Lampung
in the far south); (2) Malaysian samples came from five lo-
calities located throughout peninsular Malaysia (i.e., Perak,
Selangor, Johor, Pahang, and Taman Negara); (3) three
Myanmar samples came from Taungoo in Pegu; (4) all Nepal
samples came from Chitwan National Park; and (5) one In-
dian sample came from Assam in far northeastern India (Fig.
1). We do not have exact capture localities for samples from
Thailand or Sri Lanka, but based on the knowledge of local
experts these are also derived from local wild captures.
All but three of our original sequences are derived from
blood samples; two India-zoo sequences are from DNA iso-
lated from toenail clippings, and one African and some Gen-
bank sequences originated from salvaged hard tissues. Blood
was taken in syringes, either in the presence of heparin or
EDTA to prevent coagulation. White blood cells were sep-
arated by centrifugation. Samples were frozen and remained
so during transfer to the laboratory. In total, using both con-
trol region sequences (57 individuals) and cyt b sequences
(three different individuals) we determined the mitochondrial
clade membership for 60 individual Asian elephants from
seven countries (Table 1). We also obtained new cyt b se-
quences for two African elephants (Jacksonville Zoo and San
Diego Zoo).
1884 ROBERT FLEISCHER ET AL.
FIG. 1. Map of southern Asia showing the countries from which Asian elephant samples used in this study were collected. Sample sizes
of genotyped elephants are in parentheses. Dots indicate sampling localities of elephants, when known (see text). Pie diagrams depict
the relative frequencies of clade A (black) and clade B (white) in each sample (see text and Figs. 2 and 5 for explanation). Sample sizes
include all elephants designated by clade. This includes three individuals typed with cytochrome b only, for a total of 60. Pegu and the
Indus Valley are noted. Inset shows the approximate area of southern Asia at present (black) and at periods of glacial maximum (outline).
Molecular Protocols
DNA was isolated from blood samples and hard tissues by
digestion in saline buffer-SDS with proteinase k followed by
phenol-chloroform extractions and ethanol-Na-acetate pre-
cipitation. DNA was hooked from solution and placed in Tris-
EDTA buffer for storage and subsequent use.
We amplified two segments of mtDNA from elephant ge-
nomic DNA via the polymerase chain reaction (PCR). The
first segment, part of the cyt b gene, was amplified and se-
quenced from a subset of individuals, primarily to determine
the sequence divergence between African and Asian ele-
phants to calibrate a substitution rate. The second region,
which we here label as ??control region?? (CR), was amplified
from a much larger sample of Asian elephants (but could not
be amplified cleanly from our African elephant samples). CR
contains a small piece of the 39 end of the cyt b gene, tRNAthr,
and tRNApro, and the 59 end of the control region. The cyt
b product totaled 1001 bp and was amplified with primers
cyt b1 (39 base: L-14,608 of bovine mtDNA sequence, An-
derson et al. 1982; same primer as L-14,841 of human; from
Kocher et al. 1989) and cyt b-EE (H-15,800 of bovine
mtDNA, Anderson et al. 1982; 59-GTATAGAATT-
GAGGCTATTTG-39). The CR product totaled 721 bp and
was amplified with primers L-15,775 (bases 17?38 from the
sequence in table 1 of Irwin et al. 1991; L-15,540 of bovine
mtDNA, Anderson et al. 1982) and EW470 (H-16,265 of
bovine mtDNA, Anderson et al. 1982; 59-CCTGAAGTAG-
TAGGAACCAGATG-39).
Standard PCR components (at 1.5 mM MgCl2) and ap-
proximately 100 ng of template DNA (estimated by com-
parison of EtBr fluorescence to a known amount of size mark-
er on a minigel) were combined in a total volume of 50 ml.
Thermal cycling began with a hot start at 948C for 3 min,
and was followed by 25 cycles of 948C for 50 sec, 558 for
50 sec, and 728 for 1 min. This was followed by a period of
3 min at 728C for product extension. PCR products were
1885ASIAN ELEPHANT PHYLOGEOGRAPHY
TABLE 1. Sample sizes and sampling localities of elephants sequenced in this study (additional sequences for cytochrome b genes were obtained
from Genbank; see text).
Species/Region N Sampling Institutions
African elephant 2 San Diego Zoo (1), Jacksonville Zoo (1)
Asian elephant (60)
India
Indonesia
Malaysia
Myanmar
Nepal
Sri Lanka
Thailand
4
8
8
5
5
15
15
Birmingham Zoo (1), El Paso Zoo (1), Los Angeles Zoo (1), Malaka Zoo (1)
Sumatra: Taman Safari (from three locations in the wild)
Malaka Zoo (4), Singapore Zoo (2), Zoo Negara (2)
Rangoon Zoo
Royal Chitwan National Park
Malaka Zoo (2), Pinnewala Elephant Orphanage (13)
Khao Kheow Zoo (3), Dusit Zoo (4), Siam Park (7), Jacksonville Zoo (1)
FIG. 2. Phylogeny of Asian and African elephant cytochrome b
sequences. This midpoint-rooted tree is the shortest (D 5 0.107)
derived via heuristic searches using a minimum-evolution criterion.
The model of sequence change is the HKY85 model with G-cor-
rection (a 5 1.45). A cladistic parsimony criterion with a branch-
and-bound search results in four maximum-parsimony trees with
an almost identical consensus topology.
checked and quantified on an agarose gel and approximately
100 ng of Qiagen purified product was used for cycle se-
quencing reactions according to the manufacturer?s protocol
(Applied Biosystems, Inc., Foster City, CA). These reactions
were run on an ABI 373 automated sequencer. Sequence
chromatograms for light and heavy strands were aligned and
checked for correct base calls in SeqEd (ver. 1.0.3, Applied
Biosystems, Inc.) and/or Sequencher (ver. 3.0, GeneCodes
Corp., Ann Arbor, MI). For cyt b, sequence was generated
using the cyt b1 and cyt b-EE end primers along with cyt b2
(Kocher et al. 1989) and cyt b-DD (L-15,076 of bovine
mtDNA sequence, Anderson et al. 1982; 59-CATTTCA-
TCCTTCCATTTAC-39) internal primers. For CR we used
the PCR primers (above) for sequencing.
For cyt b we sequenced five Asian elephants (Indonesia2,
Malaysia2, Myanmar3, Thailand2, and studbook number 130
from the Jacksonville Zoo, but originally from Thailand; Fig.
2) and one African elephant (L. africana) from the Jackson-
ville Zoo (studbook number 394). This African sequence dif-
fered substantially from that of Irwin et al. (1991; Genbank
accession number X56285). Therefore we obtained tissue
from O. Ryder (San Diego Zoo) and resequenced cyt b from
the individual originally sequenced by Irwin et al. (1991).
Four additional Asian elephant (numbers D83048, Ozawa et
al. 1997; D50844, D50845, and AB002412, Noro et al. 1998;
collection localities unknown), and three African elephant
cyt b sequences were obtained from Genbank (numbers
D84150?D84152, Noro et al. 1998). Thus, we have five Af-
rican elephant and nine Asian elephant cyt b sequences for
the comparison and rate calibration. A dugong cyt b sequence
(Dugong dugong; U07564) was included to root the tree for
a likelihood ratio test of molecular rate homogeneity (see
below). Last, we obtained sequences from 57 Asian elephants
for the CR segment. When combined with the three individ-
uals for which we had cyt b sequences only, we could de-
termine clade membership for 60 individuals.
Analysis of Mitochondrial DNA Sequence Data
We initially aligned the elephant sequences to each other
and the published bovine one (Anderson et al. 1982) using
Sequencher. We estimated the number of haplotypes and seg-
regating sites, and the gene diversity (h) from both the aligned
cyt b and CR sequences. Nucleotide diversity (p, or the mean
of pairwise sequence difference) and its standard error were
estimated uncorrected as in Nei (1987) using the program
NucDiversity (C. E. McIntosh, unpubl.). p was calculated
for the entire sample for each species, for the two primary
clades (see below), and for each country sample of Asian
elephants. p estimates u (52Nem for haploid mtDNA) under
the assumptions of neutrality of substitutions and mutation
drift equilibrium. An alternative estimate of u, uP, can be
calculated from the proportion of segregating sites (pN of Nei
1987) and this can be compared via Tajima?s test (Tajima
1989) to u
p
to assess whether impacts of selection or pop-
ulation change can be detected. Tajima?s test was executed
1886 ROBERT FLEISCHER ET AL.
in Arlequin (ver. 2.000, Schneider et al. 1999). Haplotype
diversity (h) corrected for sampling error was calculated fol-
lowing Nei (1987).
To further assess demographic change indicated by the
mtDNA data, we constructed a mismatch distribution and
computed the raggedness index of Harpending (1994) for the
CR data using Arlequin (ver. 2.000, Schneider et al. 1999).
We compared the mismatch distribution to Poisson expec-
tation (Slatkin and Hudson 1991), and the raggedness index
to generalizable simulation results of Harpending (1994).
Last, we used the graphical coalescent approach of Nee et
al. (1995) to determine whether the pattern of increase in the
number of lineages over time (i.e., genetic distance) in a
phylogram matches the pattern expected for stable, growing
or decreasing populations. We used branch lengths (HKY85
and G-corrected; Sullivan et al. 1995; Swofford et al. 1997)
from a UPGMA (i.e., ultrametric) tree to estimate coalescence
time to a node. Both the mismatch and graphical approaches
were conducted using both the entire set of CR sequences
and each major clade separately.
Genetic distances among sequences of Asian and African
elephants were calculated in PAUP* (ver. 4.0b, Swofford
1997). To calculate unbiased 95% confidence limits of mean
distances we used the method of Steel et al. (1996). This
method requires a Jukes-Cantor correction (Swofford et al.
1997) and was conducted in CIProgram (ver. 1.01b, C. E.
McIntosh, unpubl.; available via http://www.si.edu/organiza/
museums/zoo/zooview/research/genetics/mgl.htm). For com-
parison to the Jukes-Cantor correction (Swofford et al. 1997),
we also used PAUP* to correct sequence divergence using a
more complex, but perhaps more realistic, HKY85 model
with a G-correction for rate variation among sites (Sullivan
et al. 1995). All mean divergences were also corrected for
potential lineage sorting by subtracting the weighted mean
of within-taxon distances from mean intertaxon distances and
their 95% confidence limits (Wilson et al. 1985; Nei 1987).
The corrected mean and 95% confidence limits of the Lox-
odonta-Elephas separation were divided by 5 million years,
the estimated date of cladogenesis from the fossil record, to
provide a local maximum rate of overall cyt b sequence di-
vergence.
We analyzed population structure by performing analysis
of molecular variance (AMOVA; Arlequin ver. 2.000, Ex-
coffier et al. 1992; Schneider et al. 1999) and constructing
phylogenetic trees for the CR dataset. For the AMOVA, we
make (perhaps unrealistic) assumptions that samples from
each country represent populations and that the populations
are in drift-migration equilibrium (see below). We calculate
FST-statistics that incorporate Tamura-Nei corrected se-
quence divergences among haplotypes as well as haplotype
frequencies. We use a nonparametric permutation procedure
to test whether FST-statistics are significantly different from
zero (Arlequin ver. 2.000, Excoffier et al. 1992; Schneider
et al. 1999).
We also assessed population structure by estimating the
genealogy of haplotypes via phylogenetic reconstruction. We
built trees (in PAUP*, Swofford 1997; Swofford et al. 1997)
and a minimum spanning network (in Arlequin ver. 2.0b2,
Schneider et al. 1999). Corrected distance matrices (HKY85
and G) were analysed using a parsimony criterion (minimum
evolution) in PAUP*: Starting trees obtained by the neighbor-
joining method were subjected to branch swapping in heu-
ristic searches until minimal length trees were obtained. If
more than one tree was obtained a 50% majority rule con-
sensus tree was calculated. Similar heuristic searches were
conducted using unweighted, character-based (cladistic) par-
simony (cyt b and CR). Distance and cladistic searches also
included 500 repetition bootstraps to estimate support for
particular nodes. Maximum-likelihood analysis was con-
ducted on the cyt b data using a F84 model of sequence
evolution with iterative or direct estimates of base frequen-
cies, transition to transversion ratio, and rate variation among
sites (PAUP*, Swofford 1997). Log-likelihoods were cal-
culated with and without a molecular clock from the cyt b
tree and compared using a likelihood-ratio test with (number
of terminal nodes 2 2) as the degrees of freedom (Felsenstein
1981).
RESULTS
Cytochrome b Sequences and Calibration
We sequenced between 896 and 941 bp from the 1001-bp
cyt b product. To this we added up to 110 bp of cyt b from
the CR segment to make a total cyt b sequence of up to 1020
bp (base 14,631 to base 15,650 of the bovine sequence; An-
derson et al. 1982). We aligned our sequences to the African
elephant cyt b sequence from Irwin et al. (1991). The Irwin
et al. African sequence has an inserted amino acid in residue
326 that is absent in all other elephantid sequences (Ozawa
et al. 1997; Noro et al. 1998) and sequences of other mammal
species (Irwin et al. 1991). When we aligned our resequence
of the Irwin et al. sample, we found the bases for residue
326 missing and 14 additional substitutions (seven transitions
and seven transversions). We believe that the Irwin et al.
(1991) sequence may contain sequencing errors or perhaps
represents a nuclear transposition of mtDNA sequence (e.g.,
Greenwood and Pa?a?bo 1999). Irwin et al. (1991) found, as
we did for both species, that African elephant cyt b has one
less amino acid than other mammalian cyt b sequences. Asian
elephant has a TAA termination codon, like African, rather
than AGA as in most other mammals. No other unusual fea-
tures were noted in our cyt b sequences.
The five African cyt b sequences contain four different
haplotypes (our Jacksonville Zoo and one Noro et al. [1998]
sequence were identical) and 11 variable sites; thus P 5 0.012
and up 5 0.0056 (SE 5 0.0034). All 11 substitutions involved
transitional changes (six A:G and five C:T); nine occurred
in third positions and the remaining two substitutions occured
in second positions. The nucleotide diversity (p of Nei 1987)
averaged 0.0064 (SE 5 0.0016) substitutions per site.
In contrast, the nine Asian cyt b sequences contained 31
variable sites and eight haplotypes (our zoo and Malaysian2
sequences were identical). All 31 substitutions were transi-
tions (six A:G and 25 C:T); 21 of the 31 substitutions oc-
curred in third positions, and seven occurred in first and three
in second positions. The u estimated from the mean number
of segregating sites (pN 5 0.033) was 0.0121 (SE 5 0.0054;
Nei 1987) and p was 0.0126 (SE 5 0.0022).
The Asian sequences fall clearly into two differentiated
clades (labeled ??A?? and ??B?? in Fig. 2) that differ at 16
1887ASIAN ELEPHANT PHYLOGEOGRAPHY
sites. The topology from an heuristic, minimum evolution
search is shown in Fig. 2 (HKY85 model with G-correction,
shape parameter a 5 1.45, shortest tree score is 0.107). Tree
topologies found by the cladistic parsimony and maximum-
likelihood approaches only differ in resolution of some of
the terminal branches. Support for the two clades is 100%
based on 500 repetition bootstraps of both minimum evo-
lution and cladistic trees. We expected this pattern of two
divergent clades from three earlier studies, one using a 335-
bp cyt b sequence (Hartl et al. 1996), one a restriction site
analysis of PCR-amplified ND5 and ND6 (Georgiadis et al.
1994), and a third based on control region sequences from a
limited geographical sampling (Fernando et al. 2000). We
compared the region of overlap of our cyt b sequences with
those of Hartl et al. (1996) and made the same clade des-
ignations (i.e., A or B).
The African and Asian cyt b sequences differ at 52 sites.
The mean Jukes-Cantor?corrected divergence between the
five African and the nine Asian sequences is 0.069. This
distance occurs in the lower end of the range of intrageneric
distances for mammals (Johns and Avise 1998). We evaluated
the dugong-rooted tree with and without a molecular clock
in PAUP* and found a marginally significant difference in
the log-likelihoods (G 5 22.5, df 5 13, P 5 0.05). Assess-
ment of relative branch lengths suggests that the marginal
rate heterogeneity is not caused by differences between the
African and Asian clades, so we proceeded with the rate
calibration. The HKY85 and G-corrected distance is 0.078,
or 1.14 times the Jukes-Cantor distance. The unbiased 95%
confidence limits (Steel et al. 1996) for the overall Jukes-
Cantor mean (0.069) are estimated to be 0.056?0.088. We
made an approximate HKY85 correction by multiplying the
95% confidence limit extremes by 1.14 (i.e., 0.064?0.100).
After subtracting the intraspecific distance (0.011; Wilson et
al. 1985; Nei 1987), we divided the three values by 5 million
years (the estimated Loxodonta-Elephas) to obtain a rate of
0.013 cyt b sequence divergence per million years (95% CI
5 0.011?0.018 per million years). The accuracy of the cal-
ibration relies on Loxodonta and Elephas sharing a common
ancestor about 5 million years ago. However, because the
two lineages are diagnosable in the fossil record for the past
4.5 million years (Maglio 1973), they must coalesce before
then, and thus the rate we have calibrated is likely a maximum
one.
We used the cyt b rate to estimate a date for the separation
of the two major Asian elephant clades. The mean Jukes-
Cantor corrected cyt b divergence between clades A and B
is 0.019 (95% CI 5 0.011?0.027) and the weighted mean
within-clade distance is 0.004. The mean HKY85 and G-
corrected distance is, as expected, only slightly larger, at
0.020. Applying the mean corrected rate for cyt b from above
(0.013 per million years) yields a mean coalescence time for
clades A and B of 1.2 million years (95% CI 5 0.5?1.7
million years). We also calculated a rate of sequence diver-
gence based on the mean African/Asian distance from the
Georgiadis et al. (1994) ND5/ND6 RFLP dataset and the
same split of 5 million years as 0.015 per million years. The
distance between their two Asian clades is 0.016 (less within-
clade distance) and the estimated coalescence time is thus
1.1 million years, very close to our independent cyt b estimate
of 1.2 million years. These results suggest the two mito-
chondrial lineages arose in the mid-Pleistocene, or perhaps
even earlier, given that we have calibrated maximum rates.
Control Region Variation
The 675 bp of the Asian elephant CR segment includes
110 bp of the 39 end of cyt b, 67 bp of each of tRNAthr and
tRNApro, and up to 431 bp of the 59 end of the control region
(including 354 bp of the left and 77 bp of the central do-
mains). The 110 bp of cyt b aligns well to other elephant
(Irwin et al. 1991; Ozawa et al. 1997; Noro et al. 1998) and
bovine (Anderson et al. 1982) sequences. Both tRNA genes
contain expected anticodon sequences and conserved stem
sequence complementarity in the stem leading to the anti-
codon loop (Anderson et al. 1982). The CR sequence does
not align to bovine sequence until 351 bp from its 59 start,
presumably at the beginning of the conserved central domain.
There is no evidence that our sequences represent transpo-
sitions of mtDNA sequences to the nuclear genome as found
by Greenwood and Pa?a?bo (1999). Our sequences differ by a
minimum of 18 substitutions (15 transitions and three trans-
versions) from the 281 bp of nuclear (Numt) CR sequence
they amplified from elephant hair. However, one of our se-
quences (NE1) differs by only a single base from their
mtDNA sequence amplified from DNA isolated from blood.
We found 47 variable sites and 36 haplotypes among the
57 CR sequences of up to 675 bp. This results in an overall
up of 0.018 (SE 5 0.003). Variable sites are not randomly
distributed among the different genes: cyt b has 10 variable
sites (0.091), tRNAthr has two (0.030), tRNApro has three
(0.045), the left domain of the CR has 29 (0.082), and the
central domain piece has three (0.039). Only three variable
sites contain more than two character states. Thus of 50 pos-
sible substitutions, only six involve transversions (a transi-
tion/transversion ratio of 7.33:1). The p over all 57 sequences
(0.019 6 0.001) is not significantly different from up by
Tajima?s (1989) test (D 5 1.40, P 5 0.090). Positive values
of D can indicate evidence for balancing selection or possibly
admixture (Tajima 1989). Haplotype diversity was high over-
all (H 5 0.972).
Genetic Structure and Phylogeography
As found for the cyt b data (this study; Hartl et al. 1996)
and for the CR data of Fernando et al. (2000), the CR phy-
logram (Fig. 3) and minimum spanning network (not shown)
was split into two divergent clades, with 22 individuals in
clade B and 35 in clade A. The two clades differ by a Jukes-
Cantor corrected mean of 0.029 (95% CI 5 0.018?0.041;
Steel et al. 1996). The mean HKY85 and G-corrected distance
(shape parameter a 5 0.063; method of Sullivan et al. 1995)
is much higher, at 0.050, indicating that there is greater rate
variation among sites for CR than for cyt b (as expected;
e.g., Wakeley 1993). This could account in part for the very
low rate (;0.01 per million years) of elephant CR sequence
evolution estimated by Fernando et al. (2000), because they
used a Jukes-Cantor correction rather than an HKY85/G-cor-
rection and may have underestimated divergence. We used
the cyt b?based interclade coalescence time (i.e., 1.2 million
years) to obtain a rough estimate of the rate of corrected
1888 ROBERT FLEISCHER ET AL.
FIG. 3. Phylogeny of Asian elephant control region sequences. The
tree was derived by criteria and methods as in Figure 2 (with a 5
0.063). Note the deep split into two major clades (labeled A and
B). The subclade A.1 contains all of our Indonesian (Sumatran)
sample, plus one Malaysian individual. IN, Indonesia; IZ, India
(zoo); ME, Malaysia; MY, Myanmar; NE, Nepal; SL, Sri Lanka;
TH, Thailand.
FIG. 4. Plot of mean nucleotide diversity (p) of control region
sequences within each sample and the two divergent clades (A and
B). Means (squares) and 95% confidence limits (vertical bars) are
calculated following equations from Nei (1987).
sequence divergence for the CR segment as 0.035 per million
years.
Nucleotide diversity is significantly smaller in clade B than
clade A (p 5 0.003 6 0.001 for B and p 5 0.012 6 0.001
for A; Fig. 4), and also for the Sumatran (Indonesian) versus
other populations (Fig. 4). In Sumatra and Malaysia we found
only sequences of clade A (Figs. 1, 3). All eight Indonesian
sequences and most Malaysian occur in well-supported clades
basal to the other sequences in clade A (Fig. 3, Indonesian
subclade A.1). This subclade is supported by unique bases
at three sites and a bootstrap of 72% and shows a Jukes-
Cantor distance (corrected by within-clade variation) from
the other members of clade A of 0.012 (95% CI 5 0.004?
0.020). Applying the very roughly-estimated CR rate (0.035
per million years) suggests that this Indonesian clade may
have begun diverging from the other clade A elephants about
0.34 (0.12?0.57) million years ago.
No significant differences exist between nucleotide diver-
sity (p) and up in the total sample (noted above), nor for each
clade analyzed separately: In clade A, p 5 0.004 6 0.001
and up 5 0.006 6 0.002 (Tajima?s D 5 20.65, n 5 22, P
5 0.28); in clade B, p 5 0.012 6 0.001 and up 5 0.016 6
0.004 (Tajima?s D 5 20.20, n 5 35, P 5 0.44). Haplotype
diversity is 0.905 in clade A and 0.963 in clade B.
The mismatch distribution for all 57 sequences is signif-
icantly different from Poisson expectation (Kolmogorov-
Smirnoff test, P 5 0.016) and has a flattened shape usually
suggestive of a stationary rather than expanding population
(Fig. 5a; Slatkin and Hudson 1991; Harpending 1994). How-
ever, the mismatch distributions for clades A and B analyzed
separately show no significant differences from Poisson ex-
pectation (P . 0.90; Figs. 5b, 5c) and have shapes usually
associated with expanding populations. Such unimodal pat-
terns could, however, be generated by variation in substi-
tution rates among sites (Aris-Brozou and Excoffier 1996),
and CR and tRNA sequences are especially prone to such
rate variation (e.g., Tamura and Nei 1993). Thus, other anal-
yses are needed to test whether the match to a Poisson dis-
tribution is due to population expansion.
The raggedness index was low for analysis of all 57 se-
quences (r 5 0.022) and for each clade analyzed separately
(clade A: r 5 0.082; clade B: r 5 0.014) and within the
range indicating a smooth distribution and thus a population
expansion (Harpending 1994). A plot of the log of the cu-
mulative number of lineages against their HKY85 distances
was made for each clade separately and combined. Analysis
of both clades results in a line concave up that, according to
the theoretical construct of Nee et al. (1995), should indicate
a long-term stable or declining population. Analysis of each
clade separately, however, reveals linear or slightly concave
down plots, which indicate a growing population for the pe-
riod during which each clade was evolving. Given the rate
at which coalescent events occur for mtDNA, it is not likely
that we can detect recent, human-caused population changes.
AMOVAs were conducted using Tamura-Nei corrections
in Arlequin (ver. 2.000, Tamura and Nei 1993; Schneider et
1889ASIAN ELEPHANT PHYLOGEOGRAPHY
FIG. 5. Mismatch distribution (Harpending 1994): histogram of pairwise differences for control region sequences. (a) Both clades; (b)
clade A only; (c) clade B only. The solid line in each graph is the Poisson expected given means of 11.5 (both clades), 7.35 (clade A),
and 2.05 (clade B) differences among sequences. The expected is significantly different by Kolmogorov-Smirnov test only in (a) (P 5
0.016).
al. 1999) to generate FST-values of the CR sequence variation.
Two AMOVAs were conducted: one across the seven country
samples and one across the three subspecies. A hierarchical
analysis was not possible because permutation tests require
more than one population per group. Both analyses revealed
significant genetic structuring. Among the seven populations,
FST 5 0.30, and the permutation test revealed this to be
significantly different from zero (P , 0.0001). Using equa-
tion 13.25 of Nei (1987), we calculate an average long-term
Nm (migration rate) of about 1.2 migrants per generation
among populations. FST for an AMOVA comparing the three
subspecies (mainland, Sri Lankan, and Indonesian) was also
high (FST 5 0.29, P , 0.0001). However, analyzing FST
pairwise revealed a lower level of divergence for Sri Lanka
versus the mainland subspecies (FST 5 0.14, P 5 0.009).
The divergence was considerably higher (FST 5 0.42, P ,
0.0001) for the mainland and Sri Lankan subspecies when
compared pairwise with the Indonesian one.
The two major clades (A and B) evident in the phylogenetic
analyses show an unexpected correlation with geography
(Fig. 1): The more southerly of our samples have a higher
frequency of individuals of clade A than the more northerly
ones. This relationship is borne out by a highly significant
(r 5 20.78, n 5 16, P 5 0.0004) regression between the
proportion (arcsine-transformed) of individuals from clade A
and the mean latitude of the sample. This regression includes
clade designations from the present study (based on 57 CR
and three cyt b sequences) as well as those of Hartl et al.
(1996; 53 cyt b sequences from six localities) and Fernando
et al. (2000; 118 CR sequences from three localities).
DISCUSSION
Phylogeography and History
Our results indicate that Asian elephants have typical levels
of within population mtDNA variation. Thus, there is no
evidence to suggest that historically recent range fragmen-
tation and population decreases have impacted within-pop-
ulation genetic variability to any great extent. In addition,
we found significant structuring of genetic variability among
localities, in spite of a potentially very high, human-assisted
dispersal level. Also, haplotypes are consistently separated
into two highly differentiated clades that differ in their rep-
resentation across localities and, to some extent, subspecies.
These two clades can also be identified in three earlier studies
of mtDNA in Asian elephants (Georgiadis et al. 1994; Hartl
et al. 1996; Fernando et al. 2000), however, clinal distribu-
tions of clade proportions (Fig. 1) were not noted in these
studies. The Georgiadis et al. (1994) study used North Amer-
ican zoo animals and their origins in Asia were not assessed,
and the Hartl et al. (1996) and Fernando et al. (2000) studies
dealt primarily with samples collected from central latitudes.
Their Sri Lankan and southern Indian samples were similar
in clade A frequencies to our Sri Lankan sample.
A split into two highly differentiated clades has been noted
for the African elephant (Georgiadis et al. 1994). In this case,
as with a number of other studies of natural populations (e.g.,
Wayne et al. 1990; Quinn 1992; Taberlet et al. 1992; Hoelzer
et al. 1994; Arctander et al. 1996; Thomaz et al. 1996; Wood-
ing and Ward 1997), highly divergent clades are found within
single populations. In some cases, the clades show a broader
geographic pattern suggesting that prior allopatry generated
them (e.g., Wayne et al. 1990; Quinn 1992; Arctander et al.
1996; Wooding and Ward 1997). Georgiadis et al. (1994)
favor the hypothesis that the two highly divergent clades in
African elephant populations result from long-term retention
of two ancestral, nonrecombining lineages rather than hy-
bridization between two differentiated allopatric populations
or species (e.g., Loxodonta africana and L. atlantica).
In support of the lineage retention hypothesis, Georgiadis
et al. (1994) calculated the expected time to coalescence (Tc)
for haplotypes in a population as Ne[1 2 (1/n)] generations.
With our estimate of 1.18 million years for Tc for the A and
B clades and a 20-year generation time, we used this equation
to estimate a long-term Ne sufficient to maintain the two
lineages to be approximately 29,000 females. This is only
1890 ROBERT FLEISCHER ET AL.
slightly larger than current population estimates of 18,500?
27,500 females (assuming at least half the population is fe-
male) and certainly smaller than the population of the last
century. Thus, unlike Fernando et al. (2000), we feel that we
cannot rule out from this analysis that our two clades could
have been maintained in a single large population purely by
stochastic lineage retention over the long term.
We also cannot rule out an alternative explanation for the
existence of the two divergent clades: that the clades evolved
in geographically separate populations, and after a period of
isolation, some barrier to dispersal was removed and hy-
bridization occurred. The cline in the proportion of clade A
individuals from Indonesia to Nepal somewhat supports this
allopatry hypothesis. However, the cline, and lack of clade
B in our Malaysian and Indonesian samples, could also have
been caused by loss of B through drift, perhaps caused by
the decreasing land mass of the Sunda in the Holocene. The
Tajima tests and the mismatch and other analyses on clades
A and B suggest that the populations of each clade have been
stable or expanding over time (but not bottlenecked). These
changes may have occurred at different rates given the sig-
nificant difference in p between the two clades. Different
histories of each clade seems incongruous with a single-pop-
ulation, lineage sorting hypothesis.
If clades A and B do represent formerly allopatric popu-
lations that have recently merged, then their initial separation
may have occurred more than 1 million years ago. Maglio
(1973) notes that fossil E. maximus, although rare overall, is
not found until the late Pleistocene (0.2 million years ago to
present). The Javanese endemic, E. hysudrindicus, considered
a sister taxon to E. maximus, appears to have separated from
the E. hysudricus/E. maximus lineage 0.8?1.0 million years
ago (Maglio 1973; Van den Bergh et al. 1996). It should be
noted, however, that comparisons of morphological vari-
ability in extant elephants have not yet been made, and ??the
variability observed in the modern forms thus may raise ques-
tions about the validity of some of the nominal species of
fossils?? (Todd and Roth 1996, p. 201). In addition, the basal
positions of the Malaysian and Indonesian individuals in
clade A (Fig. 3), suggest that clade A most likely arose on
the southern Sunda peninsula. Oddly, E. maximus does not
show up in the fossil record of this region until the last
interglacial, perhaps 0.05 million years ago, and only after
E. hysudrindicus has disappeared (Maglio 1973; Van den
Bergh et al. 1996). The estimated divergence time of about
0.34 million years between clade A.1 and other haplotypes
in clade A (Fig. 3) suggests that clade A was in situ on the
Sunda peninsula long before the putative arrival of E. max-
imus 0.05 million years ago. It is possible that the explosive
eruption of Toba 0.071 million years ago and the volcanic
winter that followed (Rose and Chesner 1990; Ambrose 1998)
may have initiated the major dispersal events that led to mix-
ing of the taxa.
One point possibly counter to the allopatry hypothesis is
the high proportion of clade A individuals found in Sri Lanka
and southern India. However, Sri Lanka (Ceylon) is unusual
in that it was heavily impacted by the elephant trade, having
been a major elephant import/export depot for more than 1000
years (Sukumar 1989). Strachan (in Tennant 1861) gives ac-
counts of Arab traders transporting Ceylonese elephants to
India in the 1600s and Jayewardene (1994, p. 60) calculates
that ??over 3,253 were exported in the second half of the 19th
century.?? Sukumar (1989, p. 5) points out that ??elephants
from certain regions, such as Sri Lanka, were imported be-
cause they were considered especially suited for war.?? War
between regions resulted in major movements and escapes
of elephants: of the Delhi Sultanate?s 3000 war elephants,
??most came as captures from enemies in South India, as
tribute from subordinate rulers, or as imports from various
regions, including East Bengal, Sri Lanka and Pegu in lower
Burma?? (Sukumar 1989, p. 5). Sri Lanka also imported large
numbers of elephants for breeding and training. The Pegu
region (Fig. 1) of southern Myanmar, in particular, was a
major source of elephants for Sri Lanka (Digby 1971).
If trade from Pegu brought clade A elephants to Sri Lanka
and southern India then the haplotypes of these elephants
should be like ones from southern Myanmar and adjacent
Thailand. Most Sri Lankan and Indian elephants in clade A
do have haplotypes more like ones from Thailand than like
ones from Malaysia and Indonesia (Fig. 3). Likewise, in the
study by Hartl et al. (1996), the most common clade A hap-
lotype (MAX V) in Sri Lanka (36%) and southern India (55%)
is also the only clade A haplotype found by them in southern
Myanmar. A recent study by Fernando et al. (2000) revealed
that wild elephant populations in Sri Lanka had a similar
frequency of haplotypes in clade A (only 65.3% of 81 ele-
phants) to the captive ones from our and the Hartl et al. (1996)
studies. However, they found only 24% of these clade A
haplotypes in Bhutan or Laos. These results suggest that trade
among non?Sunda Asian populations may have erased some
preexisting genetic structure in these areas (but apparently
did not on the lower Sunda). Haplotypes from Sri Lanka
deserve further study.
We speculate that clade A haplotypes could be descended
from haplotypes of the Sunda species E. hysudrindicus,
whereas clade B haplotypes are descended from those of the
northern E. maximus (formerly E. hysudricus). The geograph-
ic pattern we see today would thus be a result of hybridization
between the two forms, perhaps caused in the late Pleistocene
by a combination of habitat changes, the impacts of Toba?s
eruption, and, in later periods, by the elephant trade. The
timing based on our molecular clock estimates and the geo-
graphic pattern we found do not support the alternative sug-
gestion that the two haplotypes occur because of hybridiza-
tion between a much earlier Elephas form (i.e., E. namadicus)
and the lineage leading to E. maximus (Fernando et al. 2000).
Because of higher female philopatry (Fernando and Lande
2000) and maternal inheritance of mtDNA, and perhaps little
or no trade involving Indonesia and Malaysia, we still see
genetically divergent, relictual populations in these regions,
whereas other regions reflect very high admixture.
Implications for Conservation
Our results have implications for future in situ and ex situ
management of Asian elephants. Genealogical analysis of
mtDNA markers enables us to apply definitions of evolu-
tionarily significant units (e.g., Ryder 1986; Moritz 1994;
Vogler and DeSalle 1994). We confirmed earlier results of
Hartl et al. (1996) and Fernando et al. (2000) that the Sri
1891ASIAN ELEPHANT PHYLOGEOGRAPHY
Lankan subspecies is not supported as an ESU by current
patterns of mtDNA variation. Significant haplotype frequen-
cy divergence of Sri Lankan elephants from mainland ones
(also in Fernando et al. 2000) does suggest that Sri Lankan
elephants may be a management unit (Moritz 1994). How-
ever, if we ignore the single Malaysian individual in clade
A.1 (Fig. 3), the Sumatran subspecies, E. m. sumatranus, is
monophyletic and diagnosable. Thus, under both the phy-
logeographic definition of Moritz (1994) and the single di-
agnostic character definition of Vogler and DeSalle (1994)
we could define this taxon as an ESU and thus suggest that
Sumatran elephants be managed separately from other ele-
phants.
Before making this recommendation, however, we note
some caveats to the rationale for ESUs (Fleischer 1998). First,
no nuclear markers have been analyzed, and their analysis
may lead to different conclusions given the difference in
dispersal between the sexes (Fernando and Lande 2000). Sec-
ond, we found major genetic divergence between haplotypes
in clades A and B (more than four times that between In-
donesian clade A.1 and related sequences within clade A).
In most populations that we (and others) assessed, both hap-
lotypes occur at intermediate frequencies. If the high diver-
gence between clades A and B does not cause problems for
reproduction or survival, should the smaller divergences be-
tween Sumatran, Malaysian, and the other clade A haplotypes
matter for fitness in the event of crosses? We suspect out-
breeding depression is unlikely, but the success of crosses
in captive populations needs to be assessed directly. For ex-
ample, what would happen to the extra rib in hybrids?
Finally, beyond outbreeding depression is a philosophical
argument for retaining genetic structure, that is, to allow
evolution to proceed on its natural or historical path. How-
ever, given the long and complicated historical interaction
between elephants and humans, it may be difficult to deter-
mine what was the prehistoric situation for most populations,
let alone to reconstruct it. In addition, if certain populations
do become so small that reduced genetic variation and in-
breeding compromise fitness within them (an unfortunately
realistic possibility with the Sumatran population given re-
cent trends in the loss of Indonesian lowland forest; Jepson
et al. 2001; Whitten et al. 2001), then we may not have the
luxury of maintaining the ESU. At that point, as originally
suggested for ESUs by Ryder (1986), the question of crossing
should be reevaluated.
ACKNOWLEDGMENTS
Many people contributed generously to this project. For
assistance in the field, help with logistics, and obtaining sam-
ples we thank U. K. N. Lwin, F. Chit, and M. Chit of the
Rangoon Zoo (Myanmar); Widodo of PHPA (Indonesia); Dr.
J. Alahakoon, W. Senenayake, and Mr. Manitunga (Sri Lan-
ka); S. Dumnui and Dr. M. Limpoka (Thailand); Dr. T. Man
Maskey and Dr. R. Rudran (Nepal); and W. D. Ratnasooriya,
J. Lennhardt, O. Ryder, M. Shaw, D. Krishnamurthy, M.
Keele, C. Miller, and A. Teare. C. McIntosh, A. Neumann,
and I. Jones assisted with laboratory or other analyses. The
U.S. Agency for International Development (DHR-5600-G-
00-0062-00), the Smithsonian Institution Scholarly Studies
Program, the Lennox Foundation (via J. Lennhardt), and the
Friends of the National Zoo provided funding.
LITERATURE CITED
Ambrose, S. H. 1998. Late Pleistocene human population bottle-
necks, volcanic winter, and differentiation of modern humans.
J. Hum. Evol. 34:623?651.
Anderson, S., M. H. L. de Bruihn, A. R. Coulson, I. C. Eperon, F.
Sanger, and I. C. Young. 1982. Complete sequence of bovine
mitochondrial DNA: conserved features of the mammalian mi-
tochondrial genome. J. Mol. Biol. 156:683?717.
Arctander, P., P. W. Kat, R. A. Aman, and H. R. Siegismund. 1996.
Extreme genetic differences among populations of Gazella gran-
ti, Grant?s gazelle in Kenya. Heredity 76:465?475.
Aris-Brozou, S., and L. Excoffier. 1996. The impact of population
expansion and mutation rate heterogeneity on DNA sequence
polymorphism. Mol. Biol. Evol. 13:494?504.
Avise, J. C. 1994. Molecular markers, natural history and evolution.
Chapman and Hall, New York.
Carrington, R. 1958. Elephants, a short account of their natural
history, evolution and influence on mankind. Chatto and Windus,
London.
Coppens, Y., V. J. Maglio, C. T. Madden, and M. Bedden. 1978.
Proboscidea. Pp. 336?367 in V. J. Maglio and H. B. S. Cooke,
eds. Evolution of African mammals. Harvard Univ. Press, Cam-
bridge, MA.
Digby, S. 1971. War horse and elephant in the Delhi Sultanate.
Oriental Monographs, Oxford, U.K.
Dizon, A. E., C. Lockyer, W. F. Perrine, D. P. Demaster, and J.
Sisson. 1992. Rethinking the stock concept: a phylogeographic
approach. Conserv. Biol. 6:24?36.
Excoffier, L., P. E. Smouse, and J. M. Quattro. 1992. Analysis of
molecular variance inferred from metric distances among DNA
haplotypes: application to human mitochondrial DNA restriction
data. Genetics 131:479?491.
Felsenstein, J. 1981. Evolutionary trees and DNA sequences: a max-
imum likelihood approach. J. Mol. Evol. 17:368?376.
Fernando, P., and R. Lande. 2000. Molecular genetic and behavioral
analysis of social organization in the Asian elephant (Elephas
maximus). Behav. Ecol. Sociobiol. 48:84?89.
Fernando, P., M. E. Pfrender, S. E. Encalada, and R. Lande. 2000.
Mitochondria DNA variation, phylogeography and population
structure of the Asian elephant. Heredity 84:362?372.
Ferrier, A. J. 1947. The care and management of elephants in Bur-
ma. Government Printing and Stationery, Rangoon.
Fleischer, R. C. 1998. Genetics and avian conservation. Pp. 29?47
in J. Marzluff and R. Sallabanks, eds. Avian conservation: re-
search and management. Island Press, Washington, D.C.
Georgiadis, N., L. Bischof, A. Templeton, J. Patton, W. Karesh,
and D. Western. 1994. Structure and history of African elephant
populations. I. Eastern and Southern Africa. J. Hered. 85:
100?104.
Greenwood, A. D., and S. Pa?a?bo. 1999. Nuclear insertion sequences
of mitochondrial DNA predominate in hair but not in blood of
elephants. Mol. Ecol. 8:133?138.
Harpending, H. C. 1994. Signature of ancient population growth in
a low-resolution mitochondrial DNA mismatch distribution. Hu-
man Biol. 66:591?600.
Hartl, G. B., F. Kurt, R. Tiedemann, C. Gmeiner, K. Nadlinger, K.
U. Mar, and A. Rubel. 1996. Population genetics and systematics
of Asian elephant (Elephas maximus): a study based on sequence
variation at the cyt b gene of PCR-amplified mitochondrial DNA
from hair bulbs. Z. Saeugetierk. 6:285?294.
Hoelzer, G. A., W. P. Dittus, M. V. Ashley, and D. J. Melnick.
1994. The local distribution of highly divergent mitochondrial
DNA haplotypes in toque macaques, Macaca sinica, at Polon-
naruwa, Sri Lanka. Mol. Ecol. 3:451?458.
Irwin, D. M., T. D. Kocher, and A. C. Wilson. 1991. Evolution of
the cytochrome b gene of mammals. J. Mol. Evol. 32:128?144.
Jayewardene, J. 1994. The elephant in Sri Lanka. Wildlife Heritage
Trust Publications, Colombo, Sri Lanka.
1892 ROBERT FLEISCHER ET AL.
Jepson, P., J. K. Jarvie, K. MacKinnon, and K. A. Monk. 2001.
The end for Indonesia?s lowland forests? Science 292:859?861.
Johns, G. C., and J. C. Avise. 1998. A comparative summary of
genetic distances in the vertebrates from the mitochondrial cy-
tochrome b gene. Mol. Biol. Evol. 15:1481?1490.
Kocher, T. D., W. K. Thomas, A. Meyer, S. V. Edwards, S. Pa?a?bo,
F. X. Villablanca, and A. C. Wilson. 1989. Dynamics of mito-
chondrial DNA evolution in animals: amplification and sequenc-
ing with conserved primers. Proc. Natl. Acad. Sci. USA. 86:
6196?6200.
Lahiri-Choudhury, D. K. 1991. Indian myths and history. Pp 130?
147 in S. K. Eltringham, ed. The illustrated encyclopedia of
elephants. Crescent Books, New York.
Maglio, V. J. 1973. Origin and evolution of the Elephantidae. Trans.
Am. Philos. Soc. 63(3):1?149.
Moritz, C. 1994. Defining ??evolutionarily significant units?? for
conservation. Trends Ecol. Evol. 9:373?375.
Nee, S., E. C. Holmes, A. Rambaut, and P. H. Harvey. 1995. In-
ferring population history from molecular phylogenies. Philos.
Trans. R. Soc. Lond. B 349:25?31.
Nei, M. 1987. Molecular evolutionary genetics. Columbia Univ.
Press, New York.
Noro, M., R. Masuda, I. A. Dubrovo, M. C. Yoshida, and M. Kato.
1998. Molecular phylogenetic inference of the woolly mammoth,
Mammuthus primigenius, based on complete sequences of mi-
tochondrial cytochrome b and 12S ribosomal RNA genes. J. Mol.
Evol. 46:314?326.
Nozawa, K., and T. Shotake. 1990. Genetic differentiation among
local populations of Asian elephant. Z. Zool. Syst. Evolution-
forsch. 28:40?47.
Ozawa, T., S. Hayash, and V. M. Mikhelson. 1997. Phylogenetic
position of mammoth and Steller?s sea cow within Tethytheria
demonstrated by mitochondrial DNA sequences. J. Mol. Evol.
44:406?413.
Quinn, T. W. 1992. The genetic legacy of mother goose: phylo-
geographic patterns of lesser snow goose, Chen caerulescens
caerulescens, maternal lineages. Mol. Ecol. 1:105?117.
Rose, W. I., and C. A. Chesner. 1990. Worldwide dispersal of ash
and gases from Earth?s largest known eruption, Toba, Sumatra
75 ka. Global Planet. Change 89:269?275.
Ryder, O. A. 1986. Species conservation and systematics: the di-
lemma of subspecies. Trends Ecol. Evol. 1:9?10.
Schneider, S., D. Roessli, and L. Excoffier. 1999. Arlequin: a soft-
ware for population genetics data analysis. Ver. 2.000. Univ. of
Geneva, Geneva, Switzerland.
Shoshani, J., and J. F. Eisenberg. 1982. Elephas maximus. Mamm.
Species, 182:1?8.
Slatkin, M., and R. R. Hudson. 1991. Pairwise comparison of mi-
tochondrial DNA sequences in stable and exponentially growing
populations. Genetics 129:555?562.
Slatkin, M., and W. P. Maddison. 1989. A cladistic measure of gene
flow inferred from the phylogenies of alleles. Genetics 123:
603?613.
Steel, M. A., A. C. Cooper, and D. Penny. 1996. Confidence in-
tervals for the divergence time of two clades. Syst. Biol. 45:
127?134.
Stracey, P. D. 1963. Elephant gold. Weidenfeld and Nicolson, Lon-
don.
Sukumar, R. 1989. The Asian elephant: ecology and management.
Cambridge Univ. Press, Cambridge, U.K.
Sukumar, R., and C. Santiapillai. 1996. Elephas maximus: status
and distribution. Pp. 327?331 in J. Shoshani and P. Tassy, eds.
The Proboscidea: evolution and palaeoecology of elephants and
their relatives. Oxford Univ. Press, Oxford, U.K.
Sullivan, J., K. E. Holsinger, and C. Simon. 1995. Among-site rate
variation and phylogenetic analysis of 12S rRNA in sigmodon-
tine rodents. Mol. Biol. Evol. 12:988?1001.
Swofford, D. L. 1997. PAUP*. Ver. 4.0d52-57. Sinauer Associates,
Sunderland, MA.
Swofford, D. L., G. J. Olsen, P. Waddell, and D. M. Hillis. 1997.
Methods of phylogenetic inference. Pp. 407?514 in D. M. Hillis,
C. Moritz, and B. K. Mable, eds. Molecular systematics. 2d ed.
Sinauer Associates, Sunderland, MA.
Taberlet, P., A. Meyer, and J. Bouvet. 1992. Unusual mitochondrial
DNA polymorphism in two local populations of blue tit, Parus
caeruleus. Mol. Ecol. 1:27?36.
Tajima, F. 1989. The effect of change in population size on DNA
polymorphism. Genetics 123:597?601.
Tamura, K., and M. Nei. 1993. Estimation of the number of nu-
cleotide substitutions in the control region of mitochondrial
DNA in humans and chimpanzees. Mol. Biol. Evol. 10:512?526.
Tassy, P., and J. Shoshani. 1988. The Tethytheria: elephants and
their relatives. Pp. 283?315 in M. J. Benton, ed. The phylogeny
and classification of tetrapods. II. Mammals. Clarendon Press,
Oxford, U.K.
Tennant, J. E. 1861. Sketches of the natural history of Ceylon.
Longman, London.
Thomaz, D., A. Guiller, and B. Clarke. 1996. Extreme divergence
of mitochondrial DNA within species of pulmonate land snails.
Proc. R. Soc. Lond. B 263:363?368.
Todd, N. E., and V. L. Roth. 1996. Origin and radiation of the
Elephantidae. Pp. 93?202 in J. Shoshani and P. Tassy, eds. The
Proboscidea: evolution and palaeoecology of elephants and their
relatives. Oxford Univ. Press, Oxford, U.K.
Van Den Bergh, G. D., P. Y. Sondaar, J. De Vos, and F. Aziz.
1996. The proboscideans of the South-East Asian islands. Pp.
240?248 in J. Shoshani and P. Tassy, eds. The Proboscidea:
evolution and palaeoecology of elephants and their relatives.
Oxford Univ. Press, Oxford, U.K.
Vogler, A. P., and R. DeSalle. 1994. Diagnosing units of conser-
vation management. Conserv. Biol. 8:354?363.
Wakeley, J. 1993. Substitution rate variation among sites in hy-
pervariable region 1 of human mitochondrial DNA. J. Mol. Evol.
37:613?623.
Wayne, R. K., A. Meyer, N. Lehman, B. Van Valkenburgh, P. W.
Kat, T. K. Fuller, D. Girman, and S. J. O?Brien. 1990. Large
sequence divergence among mitochondrial DNA genotypes
within populations of eastern African black-backed jackals.
Proc. Natl. Acad. Sci. USA. 87:1772?1776.
Whitten, T., D. Holmes, and K. MacKinnon. 2001. Conservation
biology: a displacement behavior for academia? Conserv. Biol.
15:1?3.
Wilson, A. C., R. L. Cann, S. M. Carr, M. George Jr., U. B. Gyl-
lensten, K. Helm-Bychowski, R. C. Higuchi, S. R. Palumbi, E.
M. Prager, R. D. Sage, and M. Stoneking. 1985. Mitochondrial
DNA and two perspectives on evolutionary genetics. Biol. J.
Linn. Soc. 26:375?400.
Wooding, S., and R. Ward. 1997. Phylogeography and pleistocene
evolution in the North American black bear. Mol. Biol. Evol.
14:1096?1105.
Yang, H., E. M. Golenberg, and J. Shoshani. 1996. Phylogenetic
resolution within the Elephantidae using fossil DNA sequence
from the American mastodon (Mammut americanum) as an out-
group. Proc. Natl. Acad. Sci. USA 93:1190?1194.
Corresponding Editor: B. Bowen