The taxonomic status of the African elephant (Loxodonta africana
) has been debated since the turn of the 20th
] and up to 22 subspecies have been described [2
]. However, modern taxonomy refers to two types, with their names reflecting the habitat in which they are found, namely the larger savannah (Loxodonta africana africana
) (Blumenbach 1797) and the smaller forest (Loxodonta africana cyclotis
) (Matschie 1900) elephants. It has become increasingly established in the literature that forest and savannah elephants are distinct species (L. Africana and L. cyclotis
], with recent publications considering their datasets in the light of this concept. The most persuasive genetic basis for a two-taxon model originates from a series of studies exploring patterns of differentiation at nuclear loci, culminating in a study using male inherited Y-chr
, andbi-parentally inherited X-chr
] that concluded "there was a deep and almost complete separation between African forest and African savannah elephants."
In this study, divergent nuclear DNA sequences segregated with either forest or savannah elephant morphological types. There were, however, a number of exceptions, including a forest elephant from Garamba in the Democratic Republic of Congo (DRC, where forest and savannah populations are sympatric) that had nuclear sequences typical of savannah elephants and two savannah elephants from Cameroon (at the limit of the forest-savannah transition zone) that had nuclear sequences typical of forest elephants [6
]. The study estimated the divergence between the savannah and forest elephants to be 3 million years. The two-taxon argument has also been used to explain data from two nuclear microsatellite DNA [5
] and one morphological study [8
]. However, subsequently Debruyne [10
] performed a morphometric analysis of museum elephant skulls, and found evidence for a continuum between two morphotypes, suggesting that, despite historical separation that promoted subdivision, these two forms freely interbreed wherever their ranges intersect.
Molecular studies using mitochondrial (mt) DNA [10
] including data from the study by Roca et al
] have pointed to a more complex scenario for African elephants. Debruyne [10
] examined several thousand base pairs of mtDNA from elephants across Africa and although he also reported two highly divergent molecular clades, these did not conform to the morphological delineations of cyclotis
. He interpreted these results as a consequence of incomplete isolation between forest and savannah African elephant populations, followed by recurrent and ongoing introgression between the two forms. Roca et al
] obtained very similar mitochondrial results but explained the non-concordance between mitochondrial and nuclear markers as a result of cytonuclear genomic disassociation such that the mitochondrial tree did not reflect the species tree. The mtDNA results observed were explained as having arisen during episodes of backcrossing between successive generations of savannah males with forest females, leading to half of extant savannah elephants surveyed possessing 'forest' typical mitochondrial haplotypes but almost exclusively 'savannah' nuclear X and Y-chromosomal DNA. Eggert et al
] (in addition to Nyakaana et al.'s mitochondrial sequences [12
]) included samples from west Africa and found a more complex picture using mtDNA and nuclear microsatellites, suggesting that western savannah and forest elephants formed a potential third Loxodonta
taxonomic unit. Finally, Roca et al.
]recently revisited the question with a statistical re-analysis of eight morphological and genetic datasets (nuclear and mitochondrial) including their own and those of Eggert et al.
] and Debruyne [10
] and reconfirmed their initial interpretation of a two taxon model with cyto-nuclear genomic dissociation.
The above-mentioned studies largely share a pronounced lack of forest elephant data. The nuclear DNA studies [4
] featured limited sampling from central African forest elephants. Despite describing a narrow hybrid zone between the two elephant types, only one population located in this zone (Garamba, (DRC)) was included and none from elsewhere in DRC or from west central Africa were examined. Elsewhere, Debruyne [10
] included elephants from across DRC in his study but was again limited by sample size. The study by Eggert et al
] was limited by the inclusion of only two populations of Central African forest elephants, both from the edge of the forest range in Cameroon which may conceivably have influenced their conclusion of the genetic uniqueness of forest and western elephants. To date, no study has addressed the partitioning of genetic diversity in the equatorial forests of Africa. Further, the potential effect of Pleistocene forest refugia was partially addressed by Eggert et al.
] and also previously reported as having a major influence on large mammal distribution and range dynamics [14
] has yet to be addressed in African elephants. Here we report results from the most extensive sample of forest elephants to date, from the core of their range, and compare these results with previously published mitchondrial DNA sequences for savannah elephants from east and southern Africa and populations from west Africa and DRC.
We examined the phylogeographic history, population structure and past demography of African elephants using patterns of molecular diversity for the mtDNA control region and cytochrome b gene. Since mtDNA is maternally inherited, this marker provides a female-biased view of population history and structure. We included the most variable mtDNA segment, the hypervariable region 1 (HVR1) of the control region since it has a high rate of nucleotide change, allowing recently diverged lineages to be distinguished [19
]. This segment is equivalent to data previously published by Eggert et al
] and Debruyne [10
], allowing us to examine forest elephant sequences within the context of a sample set with the largest geographic coverage. We could not use Roca's mtDNA sequences as he studied a different fragment (ND5 instead of control region).