Demographic and evolutionary history
We identified a Late Pleistocene origin for all five major phylogeographical groups of the modern lion (Figure 2), thus providing additional support for a single-African-origin model. This model was first proposed by Yamaguchi et al. based on morphology, distribution, and parietal art and has been reinforced by aDNA studies of extinct Late Pleistocene lions [23, 24, 35]. Additional analyses of ancient DNA from historical examples of modern lions suggest that sub-Saharan Africa, with a tentatively identified focal point in eastern-southern Africa, was the likely evolutionary cradle of the modern lion . The eastern-southern origin has been independently supported by Antunes et al., who based their findings on mtDNA and lion feline immunodeficiency virus (FIV) sequences amplified from 357 modern East African, South African and Indian lions. Our results, showing longer branches amongst eastern-southern African lions (Figure 2), also support the idea that the evolutionary cradle of the modern lion was in eastern-southern Africa.
Our results suggest a very late date for the most recent common ancestor of all five phylogeographical groups. Burger et al. previously concluded that the most recent common ancestor of the modern lion lived 74–203 thousand years BP based on mitochondrial cytb sequences calibrated with fossil data from P. leo fossilis. Antunes et al. suggested c. 145–502 thousand years BP for a major range expansion of the modern lion, including the colonization of south-western Eurasia based on mtDNA (12S and 16S rRNA genes). Although our estimate of the divergence time overlaps with both previous estimates, we demonstrate that the modern lion exodus from Africa took place only ~21,000 years ago. This analysis has not previously been possible due to the extinction of sister populations in North Africa and Middle East, now accessible for study through ancient DNA.
Our use of statistical phylogeography and DNA sequences from purported refugial regions allows us to reconstruct the past movements of lion populations and place them in the context of palaeoclimatic evidence. Late Pleistocene Africa experienced massive, sudden fluctuations in hydrology caused by the glacial/interglacial cycle, with concomitant effects on vegetation and biome distribution [51–53]. More specifically, several episodes of rapid cycling between high humidity and aridity have been identified from the palynological record [51, 53, 54], with the three most recent humid phases having occurred during Marine Isotope Stage 5 (MIS5, 120-110Ka), Marine Isotope stage 3 (MIS3, 50-45Ka) and the early Holocene (10-6Ka) [55–57]. During periods of high humidity, tropical rainforest expanded across the equatorial region, and the Sahara became savanna [52, 55, 56]. Conversely, during periods of high aridity, tropical rainforest contracted to form isolated refugia as the Sahara expanded [51, 52, 54, 56]. The timing and nature of these changes correlate well with inferred patterns of lion diversification and suggest a mechanism to explain the phylogeographical patterning of lions in Africa and Asia.
During the Middle Pleistocene, prior to MIS5, lions were probably widespread over Africa, occupying regions of savanna/scrub-woodland. During the humid period of MIS5, tropical rainforest expansion eastward from the Gulf of Guinea to the Great Rift Valley would have isolated southern and eastern African populations from western and northern populations, corresponding to the basal divergence among lion lineages ~124,200 years BP (95% HPD: 81,800-183,500). As aridity increased, leading into MIS4, the Sahara expanded and separated lion populations in North Africa and West Africa. This climatic phase corresponds to the bifurcation of these two populations, which is estimated to have occurred ~51,000 years (95% HPD: 26,600-83,100).
The contraction of tropical rainforests, as the continent dried, permitted the expansion of lions from West Africa into newly open biomes in central Africa, and a signal of this movement is recovered by the statistical phylogeographical analysis. At the same time, there is evidence for a complicated interaction between populations in East and South Africa, with the Rift Valley potentially acting as a partial barrier to dispersal. Lastly, our phylogenetic and demographic reconstructions provide evidence for two separate excursions into Asia by lions from North Africa, initially during the end-Pleistocene ~21,000 years BP (95% HPD: 8,300-38,800). The most recent population movement involves Iranian lions that appear to be descended from North African lions dispersing during the mid-late Holocene. The recent dates for emigration to Asia are interesting because lion remains have been found from throughout the Late Pleistocene in the Middle East [58, 59]. However, the genetic data presented here only reflect the most recently arrived populations in the region. Populations that resulted from earlier expansions are unlikely to be detected since their genetic signatures would have been overwritten by those of later expansions.
The identification of well-supported phylogeographical groups of lions naturally prompts the question of what has maintained these divisions. Several obvious dispersal barriers have been proposed in previous publications, including the Great Rift Valley , tropical rainforests , and the Sahara , though other topographic obstacles are also likely to have played a role. The separation between West African and Central African lions occurs in the region between Benin and Cameroon, an area bisected by the Niger River (Figure 3). Similarly, the Central African lions appear to be bounded by the Nile on their eastern front, separating them from the Eastern African group (Figure 3), though the sparse sampling from this region makes identifying barriers difficult. The influence of large rivers on African mammal phylogeography has been demonstrated for chimpanzees and bonobos (Pan spp.), where the Niger River may act as a subspecies barrier . Other African savannah ungulates, such as giraffes (Giraffa camelopardalis) also show a congruent pattern . Similarly, large rivers appear to act as biogeographic boundaries to a certain extent, even for the jaguar (Panthera onca) which is renowned for its ability to swim . It is probable that deserts, valleys, and rivers and watershed boundaries have all acted to maintain the phylogeographical structure of lions.
Our analyses recovered five major phylogeographical groups in the modern lion: North African/Asian, West African, Central African, South African, and East-South African. All of these could be designated as Evolutionarily Significant Units (ESUs) in the absence of conflicting morphological or nuclear DNA data  (Figure 2). This pattern is consistent with previous studies based on control region data from across the species range , and studies using other mitochondrial regions to examine detailed phylogeographical patterns in sub-Saharan Africa [27, 28] and western-central Africa .
International bodies currently recognize only two lion conservation units: African and Asian lions  on the basis of early attempts to categorize lions using crude allozyme separation [63, 64]. DNA sequence studies have questioned these widely accepted legislative conservation units because the current dichotomy does not coincide with the intraspecific phylogeny estimated using a wider sampling regime [19, 20, 23, 28].
Our results further suggest that this dichotomy requires revision (Figure 2). The mitochondrial data clearly show that Asian lions are nested within the diversity present in Central, West, and North Africa. Perhaps Panthera leo persica should be treated as consubspecific with Panthera leo leo, or alternatively the other phylogeographical groups could be considered for elevation to the same status. Of particular concern are the central African and western African populations, which may be close to extinction, with estimates of ~800 lions in West Africa and ~900 lions in Central Africa . Our data confirm the distinct nature of western African lions and the need to afford them appropriate protection . At the same time, we encourage a careful approach when discussing ESUs based on mtDNA, rather than morphology or nuclear DNA, due to the scale-dependent and static nature of many units (see  for detailed arguments).
The close phylogenetic relationships among Barbary, Iranian, and Indian lion populations are noteworthy given their considerable geographical separation (Figures 1 and 2). Nonetheless, the extinct North African Barbary lion harboured appreciable genetic diversity prior to extirpation, including unique cytb haplotypes (Table 1, Figure 2A). Individuals PL3 and PL12 (Table 1) differ from the majority haplotype, though neither sample is associated with a specific provenance. PL3 was collected in “Barbary” according to museum notes, whereas PL12 was kept in the Tower of London during the 15th Century and its North African origin has only recently been identified . In comparison to the Barbary lions from Tunisia and Algeria, the divergent haplotypes might reflect the historical presence of additional population subdivisions within North African lions (e.g., Atlas Mountains and Mediterranean coast populations), evidence of a large, diverse, panmictic population, or incomplete lineage sorting. Barbary lion samples with precise provenance data would be needed to resolve the issue.
The restoration of the extinct North African Barbary lion has attracted the attention of conservationists both inside and outside North Africa [19, 31, 67, 68]. Although circumstantial evidence suggested that the Barbary lion could have survived in captivity [67, 68], the most likely descendants of wild Barbary lions from the Moroccan Royal Menagerie do not appear to be (maternally) Barbary [19, this study]. However, there is a close mitochondrial relationship between the Barbary lion and the extant Indian lion, and this has been tentatively (but independently) supported by non-molecular studies [26, 30].
In the tiger, another charismatic felid species, studies of ancient mitochondrial DNA have suggested a close relationship between the extinct central Asian Caspian tiger (Panthera tigris virgata) and the extant Amur tiger (P. t. altaica) . This has allowed conservationists to discuss the translocation of Amur tiger stock to occupy the former range of the Caspian tiger , with support from the World Tiger Summit . Similarly, if no examples of purebred Barbary lions can be found within the zoo population, there might be scope for restoration of the North African lion population using the closely related Indian lion.