A total of 114 captive elephants (6 Asian; 108 African) from 43 private institutions or North American zoos accredited by the Association of Zoos and Aquariums were sampled and evaluated to investigate genetic status. Because previous analyses of the captive collection indicated potential cytonuclear dissociation between mitochondrial DNA (mtDNA) sequence and microsatellite nuclear DNA genotype data, we investigated this phenomenon within the captive collection with 2 X-linked genes (BGN and PHKA2) and 1 Y-linked gene (AMELY). These data reveal that individuals with forest-derived elephant mtDNA lineages carried only savannah elephant nuclear gene haplotypes. These results are concordant with a previous study of wild populations sampled across Africa, indicating that cytonuclear genomic dissociation was captured in the founders of the North American African elephant collection. These results are important for resolving questions that can potentially impact future management and breeding programs related to the collection.
In a previous study, Lei et al. (2008) evaluated blood samples from 114 captive elephants (6 Asian; 108 African) in order to assess the captive collection's genetic diversity. Utilizing a region of the mitochondrial DNA (mtDNA) genome spanning the nicotinamide adenine dinucleotide hydrogenase subunit 4 (ND4) through the cytochrome subunit b (Cyt-b) genes, 3 haplogroups were revealed. These analyses are concordant with previous studies (Groves and Grubb 2000; Grubb et al. 2000; Roca et al. 2001, 2005, 2007; Comstock et al. 2002) that defined the split between the forest elephant (Loxodonta cyclotis) and African savannah (Loxodonta africana) species, including a division between eastern and southern savannah elephant mtDNA patterns (Georgiadis et al. 1994; Nyakaana et al. 2002). Although Barriel et al. (1999) and Eggert et al. (2002) suggest that forest elephants in West Africa comprise a divergent clade, no samples representing this region were available for this study.
Roca et al. (2005) found that in some savannah locales distant from present-day forest habitats, many individuals with savannah-specific nuclear genotypes carried forest elephant mtDNA. This phenomenon was defined as “cytonuclear genomic dissociation” (Roca et al. 2005, 2007). This indicates that ancient hybridizations occurred between forest females and savannah males, which are larger and reproductively dominant to forest or hybrid males. Recurrent backcrossing of female hybrids to savannah bulls subsequently replaced the forest nuclear genome (Roca et al. 2005).
Analyses using 19 microsatellite nuclear loci failed to segregate the North American captive African elephant collection into discrete genetic clusters. Additionally, the incongruence between mtDNA sequence and nuclear microsatellite analyses (i.e., cytonuclear genomic dissociation) should be resolved before important management decisions are defined for the collection (Lei et al. 2008). Without comparative information from nuclear gene sequences, however, we could not completely verify or rule out savannah/forest hybridization in the captive collection. Roca et al. (2005) used gene sequences from 2 X-linked genes (BGN and PHKA2) and 1 Y-linked gene (AMELY) to verify this dissociation in wild populations. Therefore, we have evaluated nuclear gene sequence data to determine whether or not cytonuclear genomic dissociation was captured in the founder individuals. This information may prove useful to the Association of Zoos and Aquariums (AZA) Elephant Taxonomic Advisory Group/Species Survival Plan (SSP) for future management and breeding assignments.
Materials and Methods
Sampling and DNA Extraction
Blood samples from 114 captive elephants (6 Elephas maximus; 108 L. africana) at 43 private institutions or North American zoos accredited by the AZA were collected (Supplementary Data, Table S1). The Asian elephant samples were utilized as out-groups in subsequent genetic analyses. Between 5 and 10 ml of whole blood was collected in 10-ml ethylenediamine tetraacetic acid tubes. The samples were refrigerated immediately upon collection and shipped overnight with ice packs to the laboratory where the white blood cells were isolated. We extracted total genomic DNA using either a standard PCI (phenol:chloroform:isoamyl alcohol) protocol or QIAGEN QIAamp DNA Blood Mini Kit (Valencia, CA) following manufacturer's protocol.
Nuclear Gene Amplification and Sequencing
We amplified intronic regions of the X-linked genes (BGN and PHKA2) and the Y-linked gene (AMELY) using published primer sequences (Roca et al. 2001, 2005) under the following thermocycler profile conditions: 95 °C for 4 min; 40 cycles at 95 °C for 15 s, 60 °C (cycles 1 and 2), 58 °C (cycles 3–8), 56 °C (cycles 9–14), 54 °C (cycles 15–20), and 52 °C (cycles 21–40) for 30 s; 72 °C for 60 s; and 72 °C for 5 min. Polymerase chain reaction (PCR) amplifications were carried out in 25 μl reaction volumes containing 2–5 ng total genomic DNA, 12.5 μM each primer, 200 μM dNTPs, 10 mM Tris–HCl (pH 8.0), 1.5 mM MgCl2, 100 mM KCl, and 0.5 units of Taq polymerase (BIOLASE; Bioline, Inc., Taunton, MA).
We electrophoresed the amplified samples in a 1.2% agarose gel to verify the PCR products and purified them using QIAquick PCR purification kit (QIAGEN, Valencia, CA). We sequenced the purified products with BigDye Terminator system (Applied Biosystems, Foster City, CA). These products were then purified using Sephadex G-50 (Sigma-Aldrich, St. Louis, MO) and electrophoresed on an ABI 3100 DNA Analyzer (Applied Biosystems). We deposited all sequences in GenBank under the following accession numbers: AY768831-AY768938 and AY914325-AY914559. Lei et al. (2008) compared all generated sequence fragments with GenBank accession sequences to verify the forest- and savannah-derived mtDNA. The ND5 sequences from Lei et al. (2008) and Roca et al. (2005) were used to phylogenetically compare our mtDNA data set with the results of Roca et al. (2005). We used accessioned GenBank BGN, PHKA2, and AMELY nuclear gene sequences (Roca et al. 2001, 2005) for African forest, African savannah, and Asian elephants to establish the reference baselines.
Sequence Data Analysis
We aligned sequence fragments with Sequencher (Gene Codes Corporation, Ann Arbor, MI) and executed the overall alignment of the consensus sequence files in ClustalW (Thompson et al. 1994). Initial sequence comparisons and measures of variability were performed using MEGA version 4 (Tamura et al. 2007). We performed phylogenetic analyses of the data sets using maximum parsimony (MP; all sequence fragments), neighbor joining (NJ; AMELY, ND5), minimum evolution (BGN, PHKA2), and maximum likelihood (ML; all sequence fragments) methods in PAUP* 4.0b10 (Swofford 1998). We employed heuristic searches with 10 replicates of random taxon addition and tree bisection and reconnection branch swapping for MP and ML but performed exhaustive searches for AMELY because fewer unique haplotypes exist for this fragment. We used MODELTEST3.7 (Posada and Crandall 1998) to determine the best suited model of sequence evolution and, afterward, implemented the best suited model for each sequence fragment in PAUP* 4.0b10 (Supplementary Data, Table S2). Hierarchical likelihood ratio tests indicated that the HKY+I+G model of substitution and gamma distribution (Hasegawa et al. 1985) was the best fit for PHKA2 and AMELY sequence fragments. The HKY model of substitution (Hasegawa et al. 1985) was the best fit for the ND5 sequence fragment, whereas the JC+G model of substitution and gamma distribution (Jukes and Cantor 1969) was the best fit for BGN sequence fragment. We evaluated the robustness of topologies through bootstrap replicates excluding uninformative characters as follows: 1000 replicates for NJ and MP and 100 replicates for ML. We report only nodes with more than 50% support. Because Debruyne (2005) defined the 2 major mtDNA clades of African elephant as S Clade and F Clade, we use identical terminology here.
Among North American captive African elephants analyzed in this study, mtDNA ND5 sequence fragment presented 20 variable sites and 18 parsimonious sites, defining 6 haplotypes (Supplementary Data, Table S3). Based on the accessioned mtDNA ND5 sequence fragments, 11 fixed nucleotide differences were found between the forest elephant–derived and the savannah elephant haplogroups. With both forest-derived and savannah mtDNA lineages represented in the captive collection, the mtDNA ND5 sequences revealed 2 distinct African elephant mtDNA lineages that were supported by high bootstrap values (Figure 1).
We also evaluated the 108 captive individuals for the 2 X-linked genes, BGN (647 bp) and PHKA2 (1001 bp). Regardless of whether an individual possessed an mtDNA haplotype representative of a forest or savannah lineage (Figure 1), all individuals carried typical savannah elephant nuclear gene sequences (Figures 2 and 3; Supplementary Data, Tables S4 and S5). Among the 108 captive African elephants, BGN had 1 variable site representing 3 haplotypes and PHKA2 had 2 variable sites representing 3 haplotypes. We evaluated all captive male individuals (n = 15) using the intronic region of the Y chromosome gene AMELY (1551 bp), and all sequences aligned to the savannah haplotype AMELY clade I (Figure 4; Supplementary Data, Table S6; Roca et al. 2005).
The average divergence between the 2 mtDNA lineages was 6.0% and ranged from 4.9% to 7.4% (standard deviation = 1.3%; Supplementary Data, Table S7). The genetic distance estimated between the published forest elephant mtDNA lineages and the forest-derived mtDNA lineages in captive elephants (0.005 ± 0.003), as well as the genetic distance between published and captive savannah elephant mtDNA lineages (0.028 ± 0.006), was negligible. The estimated genetic distance between S Clade and F Clade mtDNA carried by captive savannah elephants (0.055 ± 0.013) agrees with the genetic distance estimated between that of the wild population (0.048 ± 0.009; Roca et al. 2005).
By comparing ND5 mtDNA sequences, 21 out of the 108 (19%) North American captive African elephants sampled clustered among the published wild forest elephant haplotypes (Figure 1). This proportion is comparable to the proportion of savannah elephants carrying F Clade mtDNA reported by Debruyne (2005) (5 of 28 or 18%) and by Roca et al. (2005) (47 of 229 or 21%). The genetic distance between the captive forest-derived mtDNA lineage and the extant forest elephant mtDNA sequence fragments was comparatively low (0.005 ± 0.003; Supplementary Data, Table S7). The genetic distance between captive forest-derived mtDNA lineages and captive savannah elephant mtDNA lineages was similar to the genetic distance of wild forest and savannah elephant estimates (Figure 1; Supplementary Data, Table S7).
The capture and exportation records of individual African elephants carrying forest mtDNA haplotypes indicated Botswana, South Africa, and Zimbabwe as countries of origin according to the African Elephant Studbook (Olson 2003). These data geographically disagree with the recognized distributions of the African forest elephant. Roca et al. (2005) and Debruyne (2005) also found similar incongruence in the wild sampled populations of Botswana, South Africa, and Zimbabwe, where only savannah elephants should be found. It was established that forest and savannah elephants fall into 2 morphologically distinct groups (Groves and Grubb 2000; Grubb et al. 2000). However, there are no morphological differences between captive individuals with forest-derived and savannah elephant mtDNA (Olson 2003).
Multilocus genotype analyses from Lei et al. (2008) failed to separate individuals by species. Nei's genetic distance between haplogroups I/II and haplogroup III was 0.047 (Lei et al. 2008). However, Comstock et al. (2002) found that Nei's genetic distance between forest elephant and savannah elephant was more than 0.659. The subsequent analyses presented here using the X-linked and the Y-linked genes were utilized to determine the nuclear genetic affinities of the elephants carrying forest-derived mtDNA. According to Roca et al. (2005), there are 4 BGN and 3 PHKA2 fixed nucleotide differences between forest and savannah elephants, but we detected no forest elephant nuclear gene haplotypes among the captive collection (Supplementary Data, Tables S4 and S5). Roca et al. (2005) also identified 7 fixed nucleotide differences in wild forest and savannah elephant species in the AMELY gene sequence fragment, but no forest elephant nuclear gene haplotypes were identified among the captive collection (Supplementary Data, Table S6). After comparing with Roca et al. (2005), we found that all individuals with forest-derived mtDNA have savannah nuclear gene haplotypes (Figures 2–4; Supplementary Data, Tables S3–S6). Because no nuclear evidence was found with multilocus genotype or X- and Y-linked nuclear gene sequence fragment data, we conclude that the forest-derived mtDNA lineage detected in the captive collection is a relic of historical hybridizations. Even though an mtDNA forest-derived elephant signature is present in some captive savannah elephants, there is no evidence of forest elephant introgression or admixture in the nuclear genome as the cytonuclear dissociation phenomena exist in both the captive collection and in many in situ African elephant populations. Based on the results of this study, we presently recommend that no changes be implemented in the captive management strategies for the North American African elephant SSP collection.
We want to foremost thank Don Harms for collecting all blood samples and for his initial laboratory work on this project. We extend our deepest appreciation to Janine L. Brown, Elizabeth W. Freeman, Deborah J. Olson, and the 43 AZA-accredited North American zoos or private institutions who contributed blood samples. We also thank Steig E. Johnson and the staff of Henry Doorly Zoo's Genetics Laboratory for their helpful comments and edits of this manuscript.