Journal of Genetics (2019) 98:83 © Indian Academy of Sciences https://doi.org/10.1007/s12041-019-1125-y

PERSPECTIVES

African genetics: enigmas and anomalies†

ALFRED L. ROCA1,2∗

1Department of Sciences, and 2Carl R. Woese Institute for Genomic Biology, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA *E-mail: [email protected].

Received 5 November 2018; revised 28 March 2019; accepted 27 May 2019

Keywords. ; mito-nuclear; ; effective population size; glacial refugia.

During the last two decades, our understanding of the Introduction genetics of populations has greatly increas- ed. Strong evidence, both morphological and genetic, sup- Morphological analyses of skull dimensions of African ele- ports recognition of two African elephant : the phants from widespread locations in Africa have revealed savanna elephant (Loxodonta africana) and the forest ele- complete separation morphologically between forest and phant (L. cyclotis). Among elephantids, phylogeographic savanna elephants, with a few intermediaries primarily patterns for mitochondrial DNA are highly incongruent with in habitat transition zones (Groves and Grubb 2000a, b; those detected using nuclear DNA markers, and this incon- Grubb et al. 2000). Various nuclear genetic studies have gruence is almost certainly due to strongly male-biased gene also reported that forest and savanna elephants are genet- flow in elephants. As our understanding of elephant popula- ically distinct and deeply divergent (Roca et al. 2001; tion genetics has grown, a number of observations may be Comstock et al. 2002; Rohland et al. 2010; Palkopoulou considered enigmatic or anomalous. Here, several of these et al. 2018), with a limited degree of ongoing hybridization are discussed. (i) There are a number of within-species mor- between the two species (Roca et al. 2001; Comstock et al. phological differences purported to exist among elephants 2002; Mondol et al. 2015). Genetic polymorphisms includ- in different geographic regions, which would be difficult to ing indels (insertion-deletion variants) that are common reconcile with the low genetic differentiation among popula- in one species may be completely or almost completely tions. (ii) Forest elephants have a higher effective population absent from the other, further demonstrating the lack of size than savanna elephants, with nuclear genetic markers nuclear gene flow between forest and savanna elephants much more diverse in the forest elephants than savanna ele- (figure 1)(Roca et al. 2001). Hybrid elephants near the for- phants, yet this finding would need to be reconciled with the est edge carry genetic markers typical of both species, yet life history of the two species. (iii) The savanna and forest the genetic variants in ones species do not introgress into elephants hybridize and produce fertile offspring, yet full- the other species (figure 1)(Roca et al. 2001, 2005), empha- genome analysis of individuals distant from the hybrid zone sizing the genetic isolation of savanna elephants from suggests that gene flow has been effectively sterilized for at forest elephants, and strongly supporting their status as least ∼500,000 years. (iv) There are unexplored potential distinct species (Mayr 1942, 1963, 1969; Roca et al. 2001, ramifications of the unusual mito–nuclear patterns among 2005, 2007). This separation is also supported by differ- elephants. These questions are considered in light of high ences in behaviour and life history traits (Grubb et al. 2000; male and low female dispersal in elephants, higher variance Turkalo et al. 2017, 2018). Further, millions of years of of reproductive success among males than females, and of genetic divergence separate forest and savanna elephants, a habitat changes driven by glacial cycles and human activity. separation almost as deep as that between Asian elephants ( maximus) and woolly (Mammuthus †This is one of the articles of collections on ’Conservation Genetics’. primigenius)(Rohland et al. 2010; Palkopoulou et al. 2018).

0123456789().: V,-vol 83 Page 2 of 12 Alfred L. Roca

Figure 1. Differences in the phylogeographic patterns of nuclear and mitochondrial markers among forest and savanna elephants. Sets of pie charts indicates for each locality the frequencies of species-typical genetic markers including: maternally-inherited mtDNA (left), paternally-inherited Y chromosome (right) or biparentally-inherited markers (centre). In each pie chart, the green colour indicates markers within clades typical of forest elephants (including mitochondrial ‘F clade’), while the blue colour indicates markers within clades typical of savanna elephants (including mitochondrial ‘S clade’). Total indicate the number of individuals (for mtDNA, Y chromosomes) or combined number of chromosome segments (biparentally inherited markers) examined. At locales with an asterisk, the number of forest-typical and savanna-typical markers was significantly different (P < 0.05) between mtDNA and biparentally inherited markers. Inset beneath the map are the expected pattern for a recent hybrid (left), while the term ‘conplastic’ is more typically used for laboratory mice in which mtDNA from one strain is bred into a second strain by backcrossing hybrid females to males only from the second strain (Yu et al. 2009). This pattern is analogous to those found in localities in which savanna elephants carry mtDNA derived from forest elephants, but show no evidence of forest elephant nuclear gene introgression, and implies a similar pattern of backcrossing. The biparentally inherited markers are three unlinked X-chromosome intronic segments; similar patterns are present in diploid autosomal markers (Roca et al. 2001; Ishida et al. 2011, 2013; Mondol et al. 2015). Forest localities: DS, Dzanga Sangha; LO, Lope; OD, Odzala. Savanna localities: AB, Aberdares; AM, Amboseli; BE, Benoue; CH, Chobe; HW, Hwange; KE, central Kenya; KR, Kruger; MA, Mashatu; MK, Mount Kenya; NA, Namibia; NG, Ngorongoro; SA, Savuti; SE, Serengeti; SW, Sengwa; TA, Tarangire; WA, Waza; ZZ, Zambezi. Garamba (GR) is in a transition zone of vegetation in Congo that includes both habitats and both elephant species (White 1983; Groves and Grubb 2000a, b). Dark green is tropical forest habitat; light green is the forest-savanna transition zone of vegetation (White 1983). Figure is from my own publication which permits reuse of figures with this reference to the original publication (Roca et al. 2005).

Analyses of nuclear genomes indicate that the forest O’Brien 2005). In many cases of interspecies hybridization, and savanna elephants have been genetically isolated for mitochondrial but not nuclear DNA crosses the species at least 500,000 years. By contrast, mtDNA patterns barrier when females are the nondispersing sex (Petit and show frequent introgression of forest elephant mtDNA Excoffier 2009). in savanna elephant populations (figure 1)(Eggert et al. As the number of elephant population genetic studies 2002; Nyakaana et al. 2002; Debruyne 2005; Johnson has increased, a number of findings have been reported et al. 2007; Ishida et al. 2011, 2013). Male elephants dis- that appear to be anomalous or enigmatic. Those that perse from their natal social group, while females do not may be considered anomalous appear at first glance to (Douglas-Hamilton 1972; Moss 2001). Thus, unlike all the be inconsistent with what may be expected, especially in other genetic markers for which gene flow can be medi- light of other known aspects of elephant biology. Others ated by males, there is little gene flow for mtDNA due may be considered enigmatic because they are puzzling to nondispersal of females that remain with core social or difficult to explain, often because there may not be groups (Ishida et al. 2011). The discordant mtDNA pat- enough information to infer strong conclusions about terns detected in elephants are thus attributable to sex the causes or consequences of some facet of elephant differences in dispersal (Roca et al. 2005, 2007; Roca and genetics. This article considers several different aspects of African elephant genetics Page 3 of 12 83 elephant genetics that may be considered to fall into these categories.

Regional morphological differences within African elephant species are purported to exist, although within-species genetic differentiation is low

There are a number of reputed geographic differences in the morphology of forest and of savanna elephants within their respective species ranges. For example, the desert dwelling elephants of Namibia are said to be taller and leaner, with longer legs and larger feet than other popula- tions of savanna elephants, and they have been proposed as a distinct subspecies (Ishida et al. 2018). Several taxa of pygmy elephants have been said to inhabit the tropical forests of Africa (Frade 1955; Groves and Grubb 2000a, b; Debruyne et al. 2003). Other regional differences have been reported among both forest and savanna elephants, with over 20 subspecies proposed, largely based on purported morphological differences (Frade 1955). Reports of geographic differences in the morphology of different populations of African elephants are difficult to reconcile with genetic data. Among savanna elephants, male dispersal and gene flow are quite high, and this would Figure 2. Genetic differentiation is low within forest and within tend to limit the degree of genetic differentiation among savanna elephants. Forest and savanna elephants are distinct populations of savanna elephants from various regions species, isolated from nuclear gene flow for at least ∼500,000 of the continent (Sukumar 2003). F is a commonly years (Palkopoulou et al. 2018). By contrast, within each species, ST high levels of gene flow limits genetic differentiation among pop- used estimate of genetic substructure among populations, ulations. (a) Savanna elephant are reputed to show geographic which ranges in value from zero (no substructure) to variation in morphology. For example, the Namibian desert one. Low FST values have been estimated for savanna elephant is said to be taller and leaner, with longer legs and elephants across their range; for example, between ele- larger feet than other savanna elephants, and they have been phants in eastern and southern Africa the value of F proposed as a distinct subspecies (Ishida et al. 2018). However, ST genetic analyses comparing desert elephants to other elephants using microsatellite markers has been reported as just 0.016 within Namibia found no evidence for genetic differentiation (Comstock et al. 2002). Likewise, there is evidence from using either microsatellites (e.g. the principal co-ordinate analy- the distribution of indels (insertion-deletion variants) that sis shown) or mitochondrial DNA sequences (not shown) (Ishida gene flow has been extensive among the forest elephants et al. 2016). (b) For forest elephants, genetic differentiation is also across central Africa (Roca et al. 2001, 2015). The esti- quite limited across central Africa. Analysis of microsatellites using a Bayesian clustering approach almost completely parti- mated FST between forest elephant populations in the tioned forest from savanna elephants (not shown). When only eastern Congolian forest block and those in the western forest elephants were analysed, an incomplete pattern of parti- Congolian forest block is just 0.035 (Ishida et al. 2018), tioning was present between eastern and western localities within and populations across the central African do the Congolian forest block (Ishida et al. 2018), consistent with not display great genetic distinctiveness (figure 2)(Ishida low levels of genetic structure and high levels of gene flow across the tropical forests of central Africa (Roca et al. 2001). Both et al. 2018). These low estimates for limited geographic dif- images are reproduced under the terms of the Creative Com- ferentiation within each species are supported by various mons Attribution License; panel A is from (Ishida et al. 2016); studies using different types of nuclear DNA data (Roca panel B is from (Ishida et al. 2018). et al. 2001; Ishida et al. 2011; Ahlering et al. 2012). A number of hypotheses may be proposed to explain the purported regional morphological differences among which are said to differ in morphology from other savanna elephants. The first would be that many of the reputed elephants, no genetic differentiation was found when they regional differences are anecdotal and may not stand up were compared with other Namibian savanna elephants to scrutiny. Many subspecies of African elephants were (figure 2)(Ishida et al. 2016). In a study that measured proposed based on the features of a single specimen, and the shoulder heights of savanna elephants using a digital may represent individual variation that was taken to be photogrammetric method (Shrader et al. 2006), savanna meaningful when typological considerations dominated elephants were found to attain similar asymptotic shoulder . For the desert dwelling elephants in Namibia, heights across the 10 populations (Shrader et al. 2006). For 83 Page 4 of 12 Alfred L. Roca a population in eastern Africa and another in southern By contrast, genetic similarity between west and central Africa for which the age of individuals was known, ele- African forest elephants has been emphasized by other phants in both localities were found to grow at the same studies (Debruyne et al. 2003; Capelli et al. 2006; Ishida rate (Shrader et al. 2006). et al. 2011). The forest elephants of west and central Another possible explanation for reported difference in Africa are genetically similar enough for genetic methods morphology is that age structure among different pop- that estimate provenance to have cross-assigned elephants ulations may vary, especially in areas where elephants between the two regions (Ishida et al. 2011). By con- have been heavily hunted for the trade, and puta- trast, such misassignment never occurs between forest tive morphological differences may reflect this (Jones and savanna elephants when similar methods are used et al. 2018). As an cautionary note, some specimens (Wasser et al. 2004). Thus, west African forest elephants assigned as belonging to proposed taxa of pygmy ele- are unlikely to represent a distinct species from forest phants turned out to be young specimens of the forest elephants in central Africa, even should forest elephants elephant once their skulls were examined and compared between the two regions be genetically differentiated. to those of other L. cyclotis and the age of the indi- Overall, the taxonomic status of forest elephants in west viduals was estimated using their eruption stage Africa remains something of an enigma (Groves 2000; (Groves and Grubb 2000a, b). Finally, if morphological Roca et al. 2015; Ishida et al. 2018). features can be shown to vary geographically within the Other points are worth noting when discussing west forest elephant or within the savanna elephant species, it African elephants. First, there is no support for the claim is possible that environmental factors such as nutrition or that forest and savanna habitats across contain pathogens may play a role. These factors may differ region- a single elephant morphological type that is intermediate ally and may have an influence on elephant growth rates or between the morphology of the forest elephant and that patterns. of the savanna elephant (Roca et al. 2015). This assertion (Johnson et al. 2007) appears to have been based on a book chapter by Fernando Frade (Frade 1955). Frade in his chapter shows a map reporting the subspecies desig- Are west and central African forest elephants nations of elephant museum specimens collected across genetically distinctive? Africa (Frade 1955; Roca et al. 2015). The map listed west African specimens as unknown, i.e. not assigned by any There may be one major exception to the lack of dif- previous authors to a putative subspecies. This map may ferentiation within African elephant species. The dense have been misunderstood to suggest that elephants in west tropical humid forest of west and central Africa, which Africa had been judged to be of unknown or interme- comprise most of the home range of the forest elephant, diate morphology between forest and savanna elephants. is fragmented in two by the Benin (or Dahomey) Gap, a This would be a misinterpretation of Frade’s work, which ca. 200 km wide corridor dominated by open vegetation unequivocally states that Africa’s elephants fall into only (Demenou et al. 2018). The Benin gap may have caused two distinct species, with a range extending across west the isolation of forest elephant populations west and east Africa for each of the two species (Frade 1934, 1955; Roca of the Gap, perhaps sufficiently to allow them to differen- et al. 2015). More recently, analyses of the full genome tiate genetically. Genetic differentiation between west and sequences of forest and savanna elephants found no evi- central African forest elephants would be of interest for dence of admixture between the two species in the genome conservation management, since in west Africa numbers of a west African forest elephant (Palkopoulou et al. 2018). of elephant are quite low with their range very fragmented, A final point is that, as elephant populations are iso- more so than in other African regions (Thouless et al. lated, they would be increasingly subject to the effects of 2016). genetic drift and inbreeding. This could lead to genetic Although the possibility needs to be better established, and morphological differences among the isolated pop- some genetic analyses have suggested that forest elephants ulations, especially among forest elephants, as the high in west Africa have a different evolutionary history than level of forest elephant genetic diversity would enable those of central Africa (Eggert et al. 2002; Ishida et al. greater differentiation among populations following iso- 2018; Palkopoulou et al. 2018). Any genetic differences lation. The elephants of west Africa may be more greatly between forest elephants in the two regions may have affected, because and decimation been augmented by hybridization of west (but not cen- of elephant populations occurred earlier there and to a tral) African forest elephants with Paleoloxodon, an extinct greater degree than in other regions (Roth and Douglas- African elephantid, as suggested by genomic comparisons Hamilton 1991; Barnes 1999). In west Africa, the range of (Meyer et al. 2017; Palkopoulou et al. 2018). There have the forest elephant currently extends into savanna habitats also been assertions that forest elephants in the two for- (Groves 2000; Mondol et al. 2015) (although it is not clear est blocks may be somewhat different morphologically if this was true historically), and the degree of hybridiza- (Bosman and Hall-Martin 1989). tion between forest and savanna elephant species appears African elephant genetics Page 5 of 12 83 to be somewhat greater than in other regions (Roca et al. A second hypothesis is that the variance in reproductive 2005). Increased opportunities for interspecies hybridiza- success in male forest elephants may be lower than that tion could also potentially enable greater differentiation of savanna elephants. Among savanna elephants there is a after isolation. high degree of male–male competition, with older larger males having much greater reproductive success than younger smaller males (Slotow et al. 2000; Hollister-Smith et al. 2007; Rasmussen et al. 2008). Because many males The higher nuclear genetic polymorphism in forest do not reproduce, their genetic diversity is lost to the next than in savanna elephants would need to be reconciled generation, reducing the effective population size relative with their life histories to census size. Under this alternative hypothesis, forest elephant males may have less variance in reproductive Genetic diversity has been examined within the two success, with a larger proportion of males contributing African elephant species using nuclear sequences and to the next generation than is the case for savanna ele- microsatellite genotypes (Roca et al. 2001, 2005; Com- phants. The genetic contribution to the next generation stock et al. 2002; Rohland et al. 2010; Ishida et al. 2011; by a larger proportion of forest elephant males would Palkopoulou et al. 2018). Nuclear genetic diversity was tend to increase Ne relative to that of the savanna ele- higher among forest elephants than among savanna ele- phant for a given census size. If this hypothesis was true, phants in all of these studies, indicating that the two species it would be unclear whether innate social behaviours (e.g. have different patterns in the generation or loss of genetic reduced male–male competition, or differences in female diversity (Roca et al. 2015). Further, the higher nuclear choice) or environmental factors (e.g. different reproduc- genetic diversity in forest elephants is evident even in stud- tive patterns due to the dense forest habitats) would be ies that included a larger number of savanna elephants responsible. from a broader geographic range when compared with the A third hypothesis is based on the fossil record of forest elephants. proboscideans. For most of the Pliocene and Pleistocene By contrast, a higher effective population size (Ne) has Epochs, taxa assigned to the Paleoloxodon (although been estimated for the forest elephant than for the savanna sometimes assigned to another genus) predominated in elephant (figure 3)(Rohland et al. 2010). The higher Ne the savannas of Africa (Maglio 1973; Kingdon 1979; among forest elephants would be need to be considered Sanders et al. 2010; Palkopoulou et al. 2018). The most in light of a recently conducted study of life history and recent of these taxa, Paleoloxodon (recki) iolensis, dis- reproductive patterns in the forest elephant (Turkalo et al. appears from the fossil record towards the end of the 2017, 2018). Forest elephants were found to have a later age Pleistocene (Sanders et al. 2010), after which the extant of first calving, and a much longer inter-calving interval savanna elephant is believed to have expanded in range than had been reported for various savanna populations (Kingdon 1979). If the current population of Loxodonta (Turkalo et al. 2017, 2018). When r/K selection theory is africana derives from a smaller founding population, considered, K-strategists tend to have relatively reduced then the lower Ne of this species may at least in part effective population sizes relative to r-strategists (Romigu- reflect a smaller ancestral population size before the ier et al. 2014). Although all elephant species would be con- extinction of Paleoloxodon (Kingdon 1979)(Roca et al. sidered to be K-selected, an older age of first calving and 2001). greater intercalving interval among forest elephants would One aspect of the high diversity of nuclear genetic be need to be reconciled with the larger Ne and higher sequences among forest elephants is that when sequences nuclear genetic diversity among forest elephants than are aligned, a high number of indels have been detected savanna elephants (Roca et al. 2001, 2005; Comstock et al. among forest elephants, relative to the number of indels in 2002; Rohland et al. 2010; Ishida et al. 2011; Palkopoulou savanna elephants (Roca et al. 2001, 2005). Among forest et al. 2018). At least three hypotheses may be reasonable. elephants known to be male that are Sanger sequenced, Forest elephant range likely retreated into tropical forest no polymorphisms are detected, since males would only refugia during glacial cycles (Ishida et al. 2018). Thus one carry one X chromosome. On the other hand, polymor- possibility may be that high diversity of forest elephants phisms are often detected among female forest elephants, may in part be due to the long-term isolation of pop- including many indels, since females carry two X chromo- ulations, which can preserve regional genetic differences somes. Small indels of the type detected in forest elephant (Ishida et al. 2018). The overall diversity of the species sequence alignments have been attributed to replication could then have increased once the populations became slippage, recombination, unequal crossing over, or imper- contiguous and panmictic, due to reversal of the Wahlund fect repair of double strand breaks (Messer and Arndt effect (Ishida et al. 2018). Under this hypothesis, the forest 2007). Several forest elephant indels are found across cen- elephant has a temporarily heightened genetic diversity, tral African forest populations, and thus have been subject which would be gradually reduced by drift (Ishida et al. to gene flow across the east–west expanse of the Con- 2018). golian forest block (Roca et al. 2001, 2005, 2015). The 83 Page 6 of 12 Alfred L. Roca

Figure 3. Relationships, genetic diversity and gene flow among proboscidean species. The black horizontal arrows represent gene flow inferred from the genomes of living and extinct species, with arrow thickness corresponding to inferred gene flow (Palkopoulou et al. 2018). Note that although gene flow has been detected across elephant species and genera, there has been no detectable gene flow between the extant forest and savanna African elephant species for ∼500,000 years (Palkopoulou et al. 2018). The tree depicts more ancient relationships that can be inferred despite subsequent gene flow (Meyer et al. 2017; Palkopoulou et al. 2018). Theta (θ), a gauge of effective population size and genetic diversity, is indicated for the three living species (Rohland et al. 2010). Note that forest elephant nuclear genetic diversity is more than three times higher than that of the savanna elephant. Branch lengths and splits are not drawn to scale. The dotted line segments and shaded areas represent limited periods of gene flow between incipient species (Palkopoulou et al. 2018). The labels consisting of N and subscripts refer to common ancestors at internal nodes are not relevant to this review. Values of θ are from Rohland et al. (2010), while the tree figure is from (Palkopoulou et al. 2018) and may be reproduced in a review article (http://www.pnas.org/page/about/rights-permissions) provided that this full journal reference is cited (Palkopoulou et al. 2018). high number of indels is consistent with the higher Ne There is a dearth of nuclear gene flow even in the detected for forest elephants than for other elephantids presence of a hybrid zone between forest and savanna (Rohland et al. 2010). The high number of indels may elephants suggest that the forest elephant species is a combina- tion of different ancient lineages that evolved separately As noted above, elephant mtDNA phylogeographic pat- for long periods of time, before combining into a sin- terns do not conform to those revealed using nuclear gle population. If the indels had previously occurred markers, due to highly sex-biased dispersal and gene flow. in locally restricted forest refugia, sufficient time has The mito–nuclear incongruence detected between African occurred since the central African forests became con- elephant species is consistent with patterns detected in tiguous for gene flow to carry the indels across this other cases of interspecies hybridization involving taxa region. in which females are the nondispersing sex (Petit and One may also consider that the high current genetic Excoffier 2009). The discordant mtDNA and nuclear diversity of forest elephants, and the detection of high genetic patterns have unfortunately been a source of con- numbers of indels would be more likely if ancient iso- fusion in genetic studies, some of which have wrongly lated populations had similar sizes at the point that they concluded that mtDNA results are novel or contradictory became contiguous due to expansion of forest habitats. to other studies that relied on nuclear DNA markers (Roca Equally large contributions from different refugia would et al. 2007; Ishida et al. 2011). have helped to equalize the frequencies of genetic vari- Phylogenies based on mtDNA have detected two lin- ants after the forests became contiguous, fostering the eages among African elephants, designated the ‘S’ and ‘F’ persistence of nuclear genetic diversity and increasing the clades (Debruyne 2005), with an estimated divergence of amount of time required for the loss of alleles to drift. 5.5 Mya (Brandt et al. 2012). The ‘S’ (or ‘savanna’) clade Such a hypothesis would, of course, be quite specula- mtDNA haplotypes are only carried by savanna elephants, tive. whereas no forest elephant in the deep tropical forest African elephant genetics Page 7 of 12 83 has been found to carry a haplotype within the S clade laboratory mouse strains undergo deliberate inbreeding, (figure 1)(Ishida et al. 2011). By contrast, F (or ‘forest- unlike the elephants. A second, and critical difference is derived’) clade mtDNA haplotypes are carried by all forest that conplastic mice are generated through human iso- elephants, and for that reason the clade is believed to lation of strains in the laboratory and manipulation of originate in forest elephant populations (figure 1)(Ishida mating patterns. In the case of elephants, the inferred et al. 2011). Yet F clade mtDNA has regularly introgressed patterns of mating would require a natural explanation. into savanna elephant populations. Because many savanna The two species are parapatric, with ranges overlapping elephant populations across eastern and southern Africa at the edges of the Congolian forest block where vegeta- carry F clade mtDNA but do not show evidence of nuclear tion zones change from tropical forest to savanna habitats genes originating in forest elephants (figure 1), a process of (figure 1)(Roca et al. 2005). Thus geographic isolation largely unidirectional hybridization and backcrossing of cannot be used to explain why the two types of elephants forest and hybrid females to savanna males can be inferred should remain genetically distinctive (Grubb et al. 2000; (figure 4)(Roca et al. 2005, 2007; Ishida et al. 2011, 2013). Roca et al. 2015; Stoffel et al. 2015). Because the two The genetic patterns present in African elephant species, African elephant types are parapatric and appear to readily in which mitochondrial but not nuclear introgression hybridize where their habitats overlap, the absence of gene between species can be detected (Cahill et al. 2013; Li flow between the forest and savanna elephants for at least et al. 2016), are somewhat analogous to those in ‘conplas- the last ∼500,000 years (figure 3)(Palkopoulou et al. 2018) tic’ inbred mouse strains (Boyse 1977; Yu et al. 2009). A requires an explanation that would make sense in light of ‘conplastic’ mouse is one in which the mtDNA is the only the widespread hybridization. A number of observations locus transferred from one strain to a second strain. A con- may be considered when framing potential explanations. plastic mouse is generated after a female from one strain is First, because mitochondrial gene flow from forest hybridized to a male from a second strain. In subsequent elephants into savanna elephant populations is readily generations, female offspring are always backcrossed to detected (figure 1)(Roca et al. 2005; Lei et al. 2008, 2009), males of the second strain. This gradually converts the female elephant hybrids are successfully passing on their nuclear genetic background of the mice to that of the sec- mtDNA, i.e. successfully reproducing. Thus a consider- ond strain, but the mice retain the mtDNA present in able female–male difference in hybrid reproductive success the original female of the first strain (Yu et al. 2009). An would be critical for understanding the discrepant mito– analogous process appears to have occurred in the case of nuclear pattern in savanna elephants (figure 4)(Roca et al. elephants and other species in which the mtDNA appears 2007). As in the case of conplastic mice, the mito–nuclear to derive from one species or lineage, but the nuclear mark- pattern detected in savanna elephants requires not only ers no longer show evidence of hybridization (figure 1) hybridization between forest and savanna elephants, but (Roca et al. 2007; Cahill et al. 2013; Li et al. 2016). specifically requires for female forest elephants to have In many savanna elephant populations, the numbers of hybridized with savanna males. That backcrossing had to individuals examined for mtDNA and nuclear sequences occur repeatedly between hybrid females and nonhybrid is quite high (figure 1)(Roca et al. 2005). For example, savanna males is supported by the absence of detectable across eastern and southern Africa, savanna elephants levels of forest elephant nuclear alleles in these savanna at 15 localities have been sequenced for the presence elephant populations (figure 1)(Roca et al. 2005; Lei of F clade (forest-elephant-derived) mtDNA or S-clade et al. 2009). To have removed the forest elephant nuclear mtDNA (present only among savanna elephants), and contribution from populations with high frequencies of were also sequenced for three X-linked gene segments forest-elephant-derived F-clade mtDNA, generations of (BGN, PHKA2 and PLP) that distinguish between for- repeated backcrossing to savanna elephant males can be est and savanna elephants. Considering the nuclear gene inferred, since this would be necessary to dilute out the segments, among eastern and southern African savanna forest elephant contribution from the nuclear gene pool elephant populations there was not a single nuclear (figure 4)(Roca et al. 2007). sequence of forest elephant origin among 1623 nuclear Second, at least some male hybrids between forest gene segments sequenced, even though 38 of 214 ele- and savanna elephants are not physiologically sterile. At phants carried F clade (forest-derived) mtDNA (figure 1) least one individual hybrid has been shown to carry a (Roca et al. 2005). For the populations that carried F Y-chromosome sequence from a forest elephant, an indi- clade mtDNA, the difference in forest elephant-derived cation that the forest elephant contributed the paternal lin- vs savanna elephant-derived mtDNA and forest elephant- eage for this individual (Roca et al. 2005). Additionally,ele- derived vs savanna elephant-derived nuclear sequences was phants in some parts of the hybrid zone have been reported very strongly supported statistically (figure 1), and the pat- to show evidence of hybridization between the two species tern is inconsistent with random mating (Roca et al. 2005). in both directions (Mondol et al. 2015). However, this Of course, there are differences to consider when com- would not detract from the very strong statistical support paring the analogous mito–nuclear patterns detected in showing many populations carry forest elephant-derived savanna elephants to those in conplastic mice. One is that mtDNA but do not carry forest elephant-derived nuclear 83 Page 8 of 12 Alfred L. Roca

Groves and Grubb 2000a, b). Although, the role of body size in reproductive success in hybrid zones has not been examined, it is plausible that the difference in body size between male savanna elephants and forest- or hybrid- ele- phant males may play a role in the backcrossing of hybrid females to unhybridized savanna males (figure 4)(Roca et al. 2005, 2007). However, it would take many generations for the gene pool of hybrid elephants to be replaced by backcrossing only to savanna elephant males. In the production of con- plastic strains of mice, typically 10 generations have been used to ensure conversion of the gene pool of hybrid mice to one that matches just with one of the parental strains (Markel et al. 1997). Yet in elephants, before 10 generations of backcrossing of hybrid females to savanna males have occurred, it would seem plausible that some hybrid males would begin to approach nonhybrid savanna elephant Figure 4. Hybridization and backcrossing between the two males in body size, and thus, that some level of forest species of African elephant. In this schema, the forest ele- phant component of the nuclear genome would be diluted and elephant nuclear contribution would be detectable in pop- completely replaced in herds that retained residual maternally ulations of savanna elephants that carry F-clade mtDNA. inherited forest-typical mtDNA haplotypes (Roca et al. 2005; The absence of detectable levels of forest elephant contri- Roca and O’Brien 2005). This pattern of backcrossing can be bution to these herds would suggest that perhaps other inferred from mito-nuclear patterns in some savanna elephant species-isolation mechanisms (Coyne and Orr 2004)may populations that carry forest elephant-derived mtDNA but do not carry traces of forest elephant nuclear DNA (figure 1)(Roca be involved in keeping the two elephant species distinct, et al. 2005). This pattern is also supported when one consid- i.e., in precluding reproductive success in hybrid males even ers the distribution of morphological types and nuclear genetic after generations of backcrossing. patterns among African elephants, and compares them to the pat- Finally, there has been no genetic introgression detected terns seen for mitochondrial DNA (figure 1)(Groves and Grubb at all from savanna elephants into forest elephants other 2000a, b; Roca et al. 2001, 2005, 2007; Debruyne 2005; Ishida et al. 2011, 2013). Where savanna and forest elephant ranges than at the edge of their range. Even though many savanna meet, hybridization occurs (Groves and Grubb 2000a, b; Roca elephant populations carry F-clade mtDNA derived from et al. 2001; Mondol et al. 2015). Given that reproductive success forest elephants, no S-clade mtDNA has ever been detected among male elephants depends largely on body size (Slotow et al. in populations deep within the tropical forest, even after 2000; Hollister-Smith et al. 2007; Rasmussen et al. 2008), and that extensive sampling and mtDNA sequencing of forest the deleterious effects of hybridization may differentially harm male hybrids (Haldane 1922), large savanna males are likely to elephants across central and west Africa (Ishida et al. have greater reproductive success than smaller forest or hybrid 2011, 2013). In a geographically more limited examina- males. However, other species isolation mechanisms (Coyne and tion of nuclear gene sequences among forest elephants, Orr 2004) may be involved. savanna elephant-typical nuclear gene sequences have not been detected among elephants in the deep tropical for- est (figure 1), although both forest and savanna elephant sequences (figure 1)(Roca et al. 2005). Thus, even in the nuclear (and mtDNA) sequences are present in the hybrid absence of physiological sterility, the reproductive success zone of Garamba (Roca et al. 2001, 2005), where both of male hybrids across many generations is sufficiently types of elephants are found (Groves and Grubb 2000b). reduced relative to nonhybrid males that they effectively The lack of nuclear introgression from savanna elephants do not contribute to the gene pool of savanna elephants in to populations in the tropical forest for at least ∼500,000 the long term (Roca et al. 2007; Palkopoulou et al. 2018). years (figure 3)(Palkopoulou et al. 2018)suggeststhatgene What causes a relative lack of reproductive success flow of any type from savanna elephants to forest elephants among hybrid forest-savanna male elephants? One argu- is effectively sterilized. ment has been that male–male competition may be in part Thus, additional species-isolation mechanisms would be responsible (Roca et al. 2005, 2007). Reproductive success needed to explain why savanna elephant nuclear gene flow among savanna elephant males is known to depend on into forest elephant populations, deep in the tropical forest body size and age, mediated by periods of in which has never been detected, and a genomic level appears to testosterone increases and males become more aggres- be have not occurred for at least ∼500,000 years (figure 3) sive towards other males (Poole 1989; Slotow et al. 2000; (Palkopoulou et al. 2018). Since hybrid males would pre- Hollister-Smith et al. 2007; Rasmussen et al. 2008). Fully- sumably be larger than forest elephant males, factors other grown forest elephant males are only half as massive as than body size must play a role in precluding introgression fully-grown savanna elephant males (Groves et al. 1993; of savanna elephant nuclear or mtDNA into the forest African elephant genetics Page 9 of 12 83 elephant gene pool. Such factors would also help explain et al. 2018). Nuclear haplotypes that are completely absent the inferred lack of reproductive success of hybrid males among forest elephants predominate among savanna ele- in the savannas. It would be speculative to discuss which phants south, east and north of the tropical forest (figure 1) species isolation mechanisms (Coyne and Orr 2004)may (Roca et al. 2001, 2005). A high level of nuclear gene flow play a role African elephants. There is no evidence for any is evident from the low genetic differentiation detected particular mechanism in elephants. But some factor bey- among savanna elephant populations across these regions ond body size may play a role in maintaining the species of Africa (Comstock et al. 2002). By contrast, mtDNA dis- barrier to nuclear gene flow from forest to savanna ele- plays a very different phylogeographic pattern in savanna phants, or to any gene flow in the reverse direction, even elephants. The S clade, which is found only among in the presence of a hybrid zone. savanna elephants, is further subdividided into three well- supported subclades (Ishida et al. 2011, 2013). While the ‘savanna-wide’clade is found among savanna elephants Do the very different phylogeographic patterns for across most of their range, the northern-central savanna nuclear and mitochondrial DNA in elephantids have subclade is found across north-central African Sudanian an effect on cellular mitochondrial function? savanna belts into east Africa, while haplotypes within the southeast savanna clade are carried by elephants in south- A final consideration regards the potential effect of ern and part of eastern Africa (Ishida et al. 2013). The the quite different phylogeographic patterns for mtDNA mtDNA phylogeography of savanna elephants is further and nuclear markers among African elephants (figure 1) complicated by the introgression of forest elephant derived and elephantids in general (Ishida et al. 2011; Roca F clade haplotypes into savanna populations. Savanna ele- 2015; Meyer et al. 2017). While in forest elephants, phants carry haplotypes that resemble (or are identical to) nuclear genetic diversity is quite high, this diversity is not the haplotypes carried by forest elephants to which they geographically structured, at least across central Africa are geographically proximate (Johnson et al. 2007; Ishida (figure 2)(Ishida et al. 2018). Also, nuclear gene flow is et al. 2011, 2013); thus savanna elephants in northern Tan- high between the eastern and western ends of the Congo- zania (which otherwise show no evidence of hybrid origin) lian forest block, given that populations carry the same tend to carry haplotypes within the east-central subclade set of indels (Roca et al. 2001, 2005, 2015). By contrast, of the F clade, which is common in forest elephants of the while all forest elephants carry mtDNA haplotypes that are eastern Congolian forest (figure 1)(Ishida et al. 2013). part of the ‘F clade,’ this clade is further subdivided into It may be possible for these quite distinctive nuclear five distinct ‘subclades,’ each with a geographically limited and mitochondrial patterns to have consequences for the distribution within the range of the forest elephant (Ishida fitness of elephant populations. Proteins encoded by the et al. 2013). For example, the ‘western subclade’ is found mitogenome form complexes that are involved in cellu- only in west Africa, in the Guinean forest block, and has lar metabolism. The proteins encoded by the mitogenome not been detected in central African populations (Eggert form these complexes with proteins encoded by the et al. 2002; Ishida et al. 2013). The west-central subclade nuclear genome. One may consider whether high levels is found both in west Africa, and also in the western part of nuclear gene flow and lack of geographic structure of the Congolian forest block. The three other subclades among nuclear genes intraspecies in elephants would affect (designated north-central, east-central and south-central) mitochondrial function and organismal fitness, given that tend to be found in greater frequency within the Congolian nuclear-encoded proteins interact with proteins coded for forest block in the ordinal direction indicated by their des- by the mitochondrial genome. Variants of the mitogenome ignation, with all three subclades completely absent from may be geographically constrained due to limited female west Africa (Ishida et al. 2013). The pattern for mtDNA dispersal, and thus may be quite distinctive locally or may reflect the isolation of forest elephant populations in regionally. Three potential consequences may be worth forest refugia during glacial cycles (Ishida et al. 2018). The considering. pattern would also reflect limited gene flow of females, First, there is likely to be a great persistence of mtDNA which would allow the local formation and local persis- haplotypes across generations, even if these haplotypes tence of mtDNA haplotypes while preventing the long may reduce the fitness of the organism by having a neg- distance movement of mtDNA haplotypes across forest ative effect on mitochondrial function. Whereas purifying regions (Ishida et al. 2018). By contrast, gene flow due selection would tend to remove genetic variants that lower to dispersal of males from natal social groups for genera- the fitness of organisms, such selection would have to tion after generation would have mediated the transfer of be mediated through a selective sweep in which the vari- nuclear markers across the length of the central African ants that provide greater fitness replace those that lead to forest (Ishida et al. 2018). lower fitness. However, a selective sweep across the range Savanna elephants also display differences between of a species would require gene flow across that range mitochondrial and nuclear phylogeographic patterns (Petit and Excoffier 2009). Since mtDNA is only trans- (figure 1)(Roca et al. 2005; Ishida et al. 2011; de Flamingh mitted by females, and female elephants are philopatric, a 83 Page 10 of 12 Alfred L. Roca selective sweep across the range of the elephants could be phylogeographic patterns has been reported not just thwarted by nondispersal of females. Thus it is possible between the two species of African elephant, but also that mitochondrial function may be negatively affected by within-species in savanna elephants, forest elephants and the mtDNA haplotypes carried by some elephants, even Asian elephants (Fernando et al. 2000; Fleischer et al. if not sufficiently compromised for selection to readily 2001; Eggert et al. 2002; Nyakaana et al. 2002; Debruyne remove the haplotypes in the absence of female-mediated 2005; Roca et al. 2005; Johnson et al. 2007; Lei et al. 2009; mitochondrial gene flow (Petit and Excoffier 2009). Ishida et al. 2013; de Flamingh et al. 2018); and, in analy- A somewhat speculative corollary may be that selec- ses involving extinct taxa, between Columbian tion could become more effective at removing haplo- (Mammuthus columbi) and species (Enk types should elephants migrate across long distances. et al. 2011; Palkopoulou et al. 2018), between Paleolox- The mtDNA-encoded proteins form complexes that are odon and the extant forest elephant (Meyer et al. 2017; involved in bioenergetics. The mtDNA variants that Palkopoulou et al. 2018), and among woolly mammoth reduce fitness may persist while females remain in a region. populations (Debruyne et al. 2008; Gilbert et al. 2008). However, the migration of females across long distances Thus among elephantids, the typical phylogeographic pat- may enable selection favouring females in which mito- tern is for mtDNA to be discordant with morphology nuclear protein complexes have not been compromised by or with nuclear genetic structure (Roca 2015; Roca et al. slightly harmful mutations in the mtDNA. It is possible 2015). Analyses of their mito-nuclear sequences and of that long distance migration of females would enable selec- how their genetic diversity may have affected mitochon- tion against mtDNA that carries mutations that negatively drial protein complexes may provide insights into the impact mitochondrial bioenergetics. evolution of proboscidean lineages. A final possibility is that there may come a point at In conclusion, we have considered several enigmatic or which the level of fitness is compromised sufficiently for anomalous findings detected by genetic studies of African purifying selection to remove the mitochondrial variants elephants. These relate primarily to sex differences among even in a matrilocal species. For example, a mutation in a elephants, which are central to understanding elephant nuclear gene that codes for a protein in a mitochondrial population genetics. These sex differences distort genetic complex could conceivably increase fitness among males, patterns, and these genetic patterns in elephantids have potentially driving a selective sweep across the range of the been prone to repeated misinterpretation. One theme of species. Potentially, the novel variant could be compatible this study is that, relative to females, male elephants have with proteins encoded by some haplotypes of mtDNA, much greater variance in reproductive success, leading to but incompatible with proteins encoded by other haplo- distortions in effective population sizes among markers, types. If this were the case, then the selective sweep of a away from ratios that would be expected if the sexes had nuclear-encoded variant could potentially lower the fit- similar variance in reproductive success. One can speculate ness of females with certain mitochondrial haplotypes to that differences between the two extant African species such a degree that their lineage may not survive. Thus in the degree of reproductive variance may potentially in principle mutations in nuclear genes involved in mito- account for higher effective population size among for- chondrial function could, during a selective sweep of a est elephants. Sex differences in the dispersal of elephants nuclear-encoded variant, drive the disappearance of haplo- have strongly distorted mtDNA phylogeographic patterns types or haplogroups of mtDNA that proved incompatible relative to the patterns inferred for any other genetic with the novel variant. The incompatibility could occur marker (since even the X chromosome, though present if their mtDNA encoded for proteins that failed to form in a single copy in males, is transmitted by males across effective complexes with the novel nuclear-encoded pro- the landscape, unlike mtDNA). Thus, mtDNA which is teins undergoing a selective sweep through male-mediated only transmitted by females, has a phylogeographic pat- dispersal. It is notable that experiments involving the trans- tern highly discrepant from those of other genetic markers, fer of mitochondria from one species into cell lines of both within and across species. Since mtDNA is one of a related species have found that the function of mito- the most widely used genetic markers, the distortion of nuclear protein complexes may be disrupted, most likely its phylogeographic patterns relative to all other mark- due to the inability of nuclear and mtDNA-encoded pro- ers has unfortunately led to misleading conclusions about teins to properly interact (Barrientos et al. 1998; McKenzie population genetic patterns. The discrepant mito–nuclear et al. 2003). patterns do allow inferences to be made regarding the role Although, the disappearance of one of two major mito- of males and females in the genetic isolation of the two chondrial clades in the woolly mammoth occurred long African elephant species (even in the face of hybridiza- before the extinction of the species (Debruyne et al. 2008; tion), and may suggest potential ramifications for protein Gilbert et al. 2008), there is no evidence to indicate complexes involved in mitochondrial function. This study that mito-nuclear interactions played a role. Nonethe- highlights the anomalous or enigmatic outcomes of genetic less, consistent with the strongly male-biased disper- research, so that these may be further considered in future sal across elephantid taxa, discordance in mito-nuclear studies. African elephant genetics Page 11 of 12 83

Acknowledgements from mitochondrial DNA sequence and nuclear microsatel- lite markers. Proc. R. Soc. London, B Biol. Sci. 269, 1993– For enabling part of the primary research that forms the basis 2006. for this study, we thank the US Fish and Wildlife Service African Enk J., Devault, A., Debruyne, R., King, C. E., Treangen, T., Elephant Conservation Fund, and specifically grants AFE1816- O’Rourke D. et al. 2011 Complete F18AP00819 and AFE1606-F16AP00909. mitogenome suggests interbreeding with woolly mammoths. Genome Biol. 12, R51. Fernando P., Pfrender M. E., Encalada S. E. and Lande R. 2000 Mitochondrial DNA variation, phylogeography and References population structure of the . Heredity 84, 362– 372. Ahlering M. A., Eggert L. S., Western D., Estes, A., Munishi Fleischer R. C., Perry E. A., Muralidharan K., Stevens E. E. and L., Fleischer R. et al. 2012 Identifying source populations and Wemmer C. M. 2001 Phylogeography of the Asian elephant genetic structure for savannah elephants in human-dominated (Elephas maximus) based on mitochondrial DNA. Evol. Int. J. landscapes and protected areas in the Kenya-Tanzania border- Org. Evol. 55, 1882–1892. lands. PLoS One 7, e52288. Frade F. 1934 Sur l’existence en Afrique de deux espéces Barnes R. F. W. 1999 Is there a future for elephants in West d’eléphants. Bull. Soc. Portug. Sci. Nat. 11, 135. Africa? Rev. 29, 175–199. Frade F. 1955 Ordre des proboscidiens ( Illiger, Barrientos A., Kenyon L. and Moraes C. T. 1998 Human 1811). Traité de zoologie: anatomie, systématique, biologie. In xenomitochondrial cybrids. Cellular models of mitochondrial Tome XVII Mammifères, vol. 17 (ed. P.-P. Grassé), pp. 715– complex I deficiency. J. Biol. Chem. 273, 14210–14217. 875. Masson, Paris. Bosman P. and Hall-Martin A. 1989 Elephants of Africa. Safari Gilbert M. T., Drautz, D. I., Lesk, A. M., Ho, S. Y.,Qi, J., Ratan Press, Long Beach, California. A. et al. 2008 Intraspecific phylogenetic analysis of Siberian Boyse E. A. 1977 The increasing value of congenic mice in woolly mammoths using complete mitochondrial genomes. biomedical research. Lab. Anim. Sci. 27, 771–781. Proc. Natl. Acad. Sci. USA 105, 8327–8332. Brandt A. L., Ishida Y., Georgiadis N. J. and Roca A. L. 2012 Groves C., Grubb P. and Bricknell S. 1993 Taxonomy of African Forest elephant mitochondrial genomes reveal that elephan- elephants. Sixth International Theriological Congress, Sydney. tid diversification in Africa tracked climate transitions. Mol. Groves C. P. 2000 What are the elephants of West Africa? Ele- Ecol. 21, 1175–1189. phant 2, 7–8. Capelli C., MacPhee R. D., Roca A. L., Brisighelli F., Georgiadis Groves C. P. and Grubb P. 2000a Are there pygmy elephants? N., O’Brien S. J. et al. 2006 A nuclear DNA phylogeny of the Elephant 2, 8–10. woolly mammoth (Mammuthus primigenius). Mol. Phylogenet. Groves C. P. and Grubb P. 2000b Do Loxodonta cyclotis and L. Evol. 40, 620–627. africana interbreed? Elephant 2, 4–7. Cahill J. A., Green R. E., Fulton T. L., Stiller M., Jay F., Grubb P.,Groves, C. P.,Dudley, J. P.and Shoshani J. 2000 Living Ovsyanikov N. et al. 2013 Genomic evidence for island African elephants belong to two species: Loxodonta africana population conversion resolves conflicting theories of polar (Blumenbach, 1797) and Loxodonta cyclotis (Matschie, 1900). bear evolution. PLoS Genet. 9, e1003345. Elephant 2, 1–4. Comstock K. E., Georgiadis N., Pecon-Slattery J., Roca A. L., Haldane J. B. S. 1922 Sex ratio and unisexual sterility in hybrid Ostrander E. A., O’Brien, S. J. et al. 2002 Patterns of molecular . J. Genet. 12, 101–109. genetic variation among African elephant populations. Mol. Hollister-Smith J. A., Poole J. H., Archie E. A., Vance E. A., Ecol. 11, 2489–2498. Georgiadis N. J., Moss C. J. et al. 2007 Age, musth and Coyne J. A. and Orr H. A. 2004 Speciation. Sinauer Associates, paternity success in wild male African elephants, Loxodonta Sunderland. africana. Ani. Behav. 74, 287–296. de Flamingh A., Roca A. L. and van Aarde R. J. 2018 Origin and Ishida Y., Oleksyk T. K., Georgiadis N. J., David V. A., Zhao phylogeography of African savannah elephants (Loxodonta K., Stephens R. M. et al. 2011 Reconciling apparent conflicts africana) in Kruger and nearby parks in southern Africa. Con- between mitochondrial and nuclear phylogenies in African ele- serv. Genet. 19, 155–167. phants. PLoS One 6, e20642. Debruyne R. 2005 A case study of apparent conflict between Ishida Y., Georgiadis N. J., Hondo T. and Roca A. L. 2013 molecular phylogenies: the interrelationships of African ele- Triangulating the provenance of African elephants using mito- phants. Cladistics 21, 31–50. chondrial DNA. Evol. Appl. 6, 253–265. Debruyne R., Van Holt A., Barriel V. and Tassy P. 2003 Status Ishida Y., Van Coeverden de Groot P. J., Leggett K. E., Putnam of the so-called African pygmy elephant (Loxodonta pumilio A.S.,FoxV.E.,LaiJ.et al. 2016 Genetic connectivity across [NOACK 1906]): phylogeny of cytochrome b and mitochon- marginal habitats: the elephants of the Namib Desert. Ecol. drial control region sequences. C. R. Biol. 326, 687–697. Evol. 6, 6189–6201. Debruyne R., Chu G., King C. E., Bos K., Kuch M., Schwarz Ishida Y., Gugala N. A., Georgiadis N. J. and Roca A. L. 2018 C. et al. 2008 Out of America: ancient DNA evidence for a Evolutionary and demographic processes shaping geographic new world origin of late quaternary woolly mammoths. Curr. patterns of genetic diversity in a keystone species, the African Biol. 18, 1320–1326. forest elephant (Loxodonta cyclotis). Ecol. Evol. 8, 4919– Demenou B. B., Doucet J. L. and Hardy O. J. 2018 History of the 4931. fragmentation of the African rain forest in the Dahomey Gap: Johnson M. B., Clifford S. L., Goossens B., Nyakaana S., Curran insight from the demographic history of Terminalia superba. B.,WhiteL.J.et al. 2007 Complex phylogeographic history of Heredity (Edinb) 120, 547–561. central African forest elephants and its implications for taxon- Douglas-Hamilton I. 1972 On the ecology and behaviour of the omy. BMC Evol. Biol. 7, 244. African elephant: the elephants of lake Manyara. Ph.D., Uni- Jones T., Cusack J. J., Pozo R. A., Smit J., Mkuburo L., Baran P. versity of Oxford. et al. 2018 Age structure as an indicator of pressure: Eggert L. S., Rasner C. A. and Woodruff D. S. 2002 The evo- Insights from rapid assessments of elephant populations across lution and phylogeography of the African elephant inferred space and time. Ecol. Indicators 88, 115–125. 83 Page 12 of 12 Alfred L. Roca

Kingdon J. 1979 East African : an atlas of evolution in Roca A. L. and O’Brien S. J. 2005 Genomic inferences from Africa. Volume III Part B (Large mammals). Academic Press, and the evolution of elephants. Curr. Opin. Genet. London. Dev. 15, 652–659. Lei R., Brenneman R. A. and Louis E. E. 2008 Genetic diversity Roca A. L., Georgiadis N. and O’Brien S. J. 2005 Cytonu- in the North American captive African elephant collection. J. clear genomic dissociation in African elephant species. Nat. Zool. 275, 252–267. Genet. 37, 96–100. Lei R., Brenneman R. A., SchmittD. L. and Louis E. E., Jr. 2009 Roca A. L., Georgiadis N., Pecon-Slattery J. and O’Brien S. J. Detection of cytonuclear genomic dissociation in the North 2001 Genetic evidence for two species of elephant in Africa. American captive African elephant collection. J. Hered. 100, Science 293, 1473–1477. 675–680. Roca A. L., Georgiadis, N. and O’Brien, S. J. 2007 Cyto-nuclear Li G., Davis B. W., Eizirik E. and Murphy W. J. 2016 Phyloge- genomic dissociation and the African elephant species ques- nomic evidence for ancient hybridization in the genomes of tion. Quat. Int. 169–170, 4–16. living cats (Felidae). Genome Res. 26, 1–11. Roca A. L., Ishida, Y., Brandt, A. L., Benjamin, N. R., Zhao, Maglio V. J. 1973 Origin and evolution of the . K. and Georgiadis, N. J. 2015 Elephant natural history: Trans. Am. Phil. Soc. Philad., New Ser. 63, 1–149. a genomic perspective. Annu. Rev. Anim. Biosci. 3, 139– Markel P., Shu P., Ebeling C., Carlson G. A., Nagle D. L., Smutko 167. J. S. et al. 1997 Theoretical and empirical issues for marker- Rohland N., Reich D., Mallick S., Meyer M., Green R. E., assisted breeding of congenic mouse strains. Nat. Genet. 17, Georgiadis N. J. et al. 2010 Genomic DNA sequences from 280–284. and woolly mammoth reveal deep speciation of for- Mayr E. 1942 Systematics and the origin of species, from the view- est and savanna elephants. PLoS Biol. 8, e1000564. point of a zoologist. Columbia University Press, New York. Romiguier J., Gayral P., Ballenghien M., Bernard A., Cahais Mayr E. 1963 Animal species and evolution. Harvard University V., Chenuil A. et al. 2014 Comparative population genomics Press, Cambridge, Cambridge. in animals uncovers the determinants of genetic diversity. Mayr E. 1969 Principles of systematic zoology. McGraw-Hill, Nature 515, 261–263. New York. Roth H. H. and Douglas-Hamilton I. 1991 Distribution and sta- McKenzie M., Chiotis M., Pinkert C. A. and Trounce I. A. 2003 tus of elephants in west Africa. Mammalia 55, 489–527. Functional respiratory chain analyses in murid xenomitochon- Sanders W. J., Gheerbrant E., Harris J. M., Saegusa H. and drial cybrids expose coevolutionary constraints of cytochrome Delmer C. 2010 Proboscidea. In Cenozoic mammals of Africa. b and nuclear subunits of complex III. Mol. Biol. Evol. 20, (ed. L. Werdelin and W. J. Sanders). University of California 1117–1124. Press, Berkeley. Messer P. W. and Arndt P. F. 2007 The majority of recent short Shrader A. M., Ferreira S. M., McElveen M. E., Lee P. C., Moss DNA insertions in the human genome are tandem duplica- C. J. and van Aarde R. J. 2006 Growth and age determination tions. Mol. Biol. Evol. 24, 1190–1197. of African savanna elephants. J. Zool. (London) 270, 40–48. Meyer M., Palkopoulou E., Baleka S., Stiller M., Penkman Slotow R., van Dyk G., Poole J., Page B. and Klocke A. 2000 K. E. H., Alt K. W. et al. 2017 Palaeogenomes of Eurasian Older bull elephants control young males. Nature 408, 425– straight-tusked elephants challenge the current view of ele- 426. phant evolution. Elife 6. Stoffel C., Dufresnes C., Okello J. B., Noirard C., Joly P., Mondol S., Moltke I., Hart J., Keigwin M., Brown L., Stephens Nyakaana S. et al. 2015 Genetic consequences of population M. et al. 2015 New evidence for hybrid zones of forest and expansions and contractions in the common hippopotamus savanna elephants in Central and West Africa. Mol. Ecol. 24, (Hippopotamus amphibius) since the Late Pleistocene. Mol. 6134–6147. Ecol. 24, 2507–2520. Moss C. J. 2001 The demography of an African elephant (Lox- Sukumar R. 2003 The living elephants: evolutionary ecology, odonta africana) population in Amboseli, Kenya. J. Zool. 255, behavior, and conservation. Oxford University Press, Oxford. 145–156. Thouless C. R., Dublin H. T., Blanc J. J., Skinner D. P., Daniel Nyakaana S., Arctander P. and Siegismund H. R. 2002 T. E., Taylor R. D. et al. 2016 African Elephant Status Report Population structure of the African savannah elephant inferred 2016: an update from the African Elephant Database. Occa- from mitochondrial control region sequences and nuclear sional paper series of the IUCN Species Survival Commission, microsatellite loci. Heredity 89, 90–98. No. 60. IUCN, Gland, Switzerland. Palkopoulou E., Lipson M., Mallick S., Nielsen S., Rohland N., Turkalo A. K., Wrege P.H. and Wittemyer G. 2017 Slow intrinsic Baleka S. et al. 2018 A comprehensive genomic history of growth rate in forest elephants indicates recovery from poach- extinct and living elephants. Proc. Natl. Acad. Sci. USA 115, ing will require decades. J. Appl. Ecol. 54, 153–159. E2566–E2574. Turkalo A. K., Wrege P.H. and Wittemyer G. 2018 Demography Petit R. J. and Excoffier L. 2009 Gene flow and species delimita- of a forest elephant population. PLoS One 13. tion. Trends Ecol. Evol. 24, 386–393. Wasser S. K., Shedlock A. M., Comstock K., Ostrander E. A., Poole J. H. 1989 Mate guarding, reproductive success and female Mutayoba B. and Stephens M. 2004 Assigning African ele- choice in African elephants. Ani. Behav. 37, 842–849. phant DNA to geographic region of origin: applications to the RasmussenH.B.,OkelloJ.B.A.,WittemyerG.,SiegismundH. ivory trade. Proc. Natl. Acad. Sci. USA 101, 14847–14852. R., Arctander P.,Vollrath F. et al. 2008 Age- and tactic-related White F. 1983 The Vegetation of Africa. UNESCO, Paris. paternity success in male African elephants. Behav. Ecol. 19, Yu X., Gimsa U., Wester-Rosenlof L., Kanitz E., Otten W.,Kunz 9–15. M. et al. 2009 Dissecting the effects of mtDNA variations on Roca A. L. 2015 Evolution: The Island of misfit mammoths. Curr. complex traits using mouse conplastic strains. Genome Res. 19, Biol. 25, R549–R551. 159–165.

Corresponding editor: T. N. C. Vidya