Journal of Heredity Advance Access originally published online on January 11, 2006
Journal of Heredity 2006 97(2):119-132; doi:10.1093/jhered/esj012
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Regional Genetic Structuring and Evolutionary History of the Impala Aepyceros melampus
From the Department of Evolutionary Biology, Institute of Biology, University of Copenhagen, Universitetsparken 15, DK-2100 Copenhagen Ø, Denmark
Address correspondence to E. D. Lorenzen at the address above, or e-mail: edlorenzen{at}bi.ku.dk.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionConclusionsAppendix ASupplementary DataReferences |
|---|
Samples of 162 impala antelope (Aepyceros melampus) from throughout its distribution range in sub-Saharan Africa were surveyed using eight polymorphic microsatellite loci. Furthermore, 155 previously published mitochondrial DNA (mtDNA) sequences from the same localities were reanalyzed. Two subspecies of impala are presently recognized—the isolated black-faced impala (Aepyceros melampus petersi) in southwest Africa and the common impala (Aepyceros melampus melampus) abundant in southern and east Africa. All tests performed indicated significant genetic differentiation at the subspecific level. Furthermore, individual-based analyses split the common impala subspecies into two distinct genetic groups, conforming with regional geographic affiliation to southern or east Africa. This was supported by assignment tests, genetic distance measures, pairwise
values, and analysis of molecular variance. We suggest that the presence of such previously unknown regional structuring within the subspecies reflects a pattern of colonization from a formerly large panmictic population in southern Africa toward east Africa. This scenario was supported by a progressive decline in population diversity indices toward east Africa and a significant increase in the quantity /(1 – ). Both microsatellite and mtDNA data indicated a genetic distinctiveness of the Samburu population in Kenya.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAppendix ASupplementary DataReferences |
|---|
Defining the distribution of wild animal populations is essential for determining the appropriate scale of their conservation and management (Beaumont and Bruford 1999; Moritz 1994; Paetkau 1999). A population's distribution is a product of ecological and evolutionary processes and has important long-term consequences (Moritz 1999). However, it is difficult to determine whether topographic features such as rivers or mountains, political jurisdictions such as country boundaries, or differences in morphology actually represent natural boundaries in terms of population genetics. It can therefore be useful to confirm whether the subjective classifications are consistent with genetic information (Beaumont and Bruford 1999; Pritchard et al. 2000; Rosenberg et al. 2001).
Distributed throughout southern and east Africa, the impala is one of the most abundant antelope species on the continent, numbering more than 1.5 million individuals (East 1998). It is a medium-sized antelope (Jarman and Jarman 1973), and its distinctiveness presently warrants it the sole representative of the Subfamily Aepycerotinae (Meester and Setzer 1977; Skinner and Smithers 1990), although disagreements as to the taxonomic affinity persist (Skinner and Smithers 1990). It has previously been placed with the Antilopinae and the Alcelaphinae and has been considered close to the Reduncinae (reviewed in Kingdon 1982; Meester and Setzer 1977). Presently, two subspecies of impala are recognized (Haltenorth and Diller 1977; Kingdon 1997; Nowak 1999). The geographically isolated and morphologically distinct black-faced impala (Aepyceros melampus petersi, Bocage 1879) is found only in the extreme semiarid zone in northern Namibia and southwest Angola (East 1998; Meester and Setzer 1977; Skinner and Smithers 1990; Swart 1967; Figure 1). It is listed as vulnerable by the International Union for the Conservation of Nature and Natural Resources (East 1998), numbering only some 2,000 individuals in the wild (Green and Rothstein 1998). The common impala (Aepyceros melampus melampus, Lichtenstein 1812) is abundant throughout its range and is widely distributed from Kenya and southern Uganda to Northern KwaZulu-Natal in South Africa, extending westward in the more southerly parts of its range (Figure 1). Favored for farming and hunting, the subspecies has been widely introduced to privately owned land and game reserves in Zimbabwe, South Africa, and Namibia (East 1998). The intraspecific taxonomy of the common impala has been subject to disagreements. Up to five subspecies have previously been listed (Meester and Setzer 1977), but with overlapping and ill-defined boundaries. Presently one subspecies is recognized (Kingdon 1997).
|
Previous analyses of the genetic structure of African bovids have revealed a significant regional distinction of populations into southern and eastern Africa. For example, hartebeest (Alcelaphus buselaphus), topi (Damaliscus lunatus), kudu (Tragelaphus strepsiceros), sable antelope (Hippotragus niger), and roan antelope (Hippotragus equinus) exhibited varying degrees of regional differentiation (Arctander et al. 1999; Mathee and Robinson 1999; Nersting and Arctander 2001; Van Hooft et al. 2002). It has been suggested that these previously pan-African species became extinct during the Pleistocene except in refuges in the south and east. The violent climatic fluctuations characteristic of this period have therefore influenced the appearance, geographical expansion, and subsequent evolution of these species (Arctander et al. 1999). Interestingly, the wildebeest (Connochaetes taurinus) survived only in the south, with subsequent recolonization toward the east leading to its current distribution range (Arctander et al. 1999). This scenario of refugia is supported by the fact that a similar pattern of spatial variation has been observed in carnivores (Girman et al. 1993) and rodents (Mathee and Robinson 1997).
The objectives of this study were to analyze DNA polymorphism in the impala (A. melampus) in order to (1) determine the spatial genetic structuring of the species across its distribution based on nuclear microsatellite loci, particularly within the common impala subspecies (A. m. melampus) in southern and east Africa, and (2) investigate the evolutionary history of the impala using both microsatellites and mitochondrial DNA (mtDNA) loci.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAppendix ASupplementary DataReferences |
|---|
Samples
We used eight microsatellite markers to genotype a total of 162 impala, collected from 11 localities in seven different countries across the species' distribution range in southwest, south, and east Africa (Figure 1). Eighty-three samples of the black-faced impala subspecies (A. m. petersi) were collected from three sites in and around Etosha National Park, Namibia: Ombika (OM, n = 16), Olifantsbad (OL, n = 33), and Halali (HA, n = 34). The 79 common impala (A. m. melampus) samples were sampled between 1992 and 1997 from Chobe National Park, Botswana (CH, n = 10); Shangani in Zimbabwe (SH, n = 11); Kafue National Park, Zambia (KAF, n = 5); South Luangwa National Park, Zambia (LU, n = 6); Selous Game Reserve, Tanzania (SE, n = 12); Burko Forest Reserve, Tanzania (BK, n = 13); Lake Mburu National Park, Uganda (LM, n = 12); and Samburu National Reserve, Kenya (SA, n = 10) (Figure 1).
All samples were collected by remote skin biopsy sampling (Karesh et al. 1987) from dry skin or ear nicks. In the field, samples were immediately stored in a solution of 25% dimethyl sulfoxide saturated with NaCl (Amos and Hoelzel 1991). In the laboratory, samples were stored at –80°C.
Laboratory Methods
DNA Extraction and Amplification
Total genomic DNA was extracted from tissue using the Qiagen DNeasy Tissue Kit. The double-stranded DNA was purified, desalted, and concentrated through QiaQuick PCR purification kit columns and amplified using the polymerase chain reaction (PCR). We analyzed eight microsatellite loci originally isolated in domestic cattle, goat, and sheep genomes (one bovine, two caprine, and five ovine), which amplified robustly and repeatedly and were found highly polymorphic (Table 1 with associated reaction conditions).
|
PCR amplifications were modified from the original protocols as follows: 10 min at 95°C, 10 cycles of touchdown with each cycle stepped down by 1°C followed by 25–40 cycles consisting of 95°C for 30 s, annealing fixed at the lowest temperature of the touchdown cycles for 30 s and extension at 72°C for 30 s, and final extension for 1 h. Cycles were carried out in a 10-µl reaction volume containing 100–200 ng of genomic DNA, 1x PCR buffer (15 mM Tris-HCl, pH 8.3, 50 mM KCl), 5 mM MgCl2, 0.2 mM of each deoxynucleoside triphosphate, 0.5 µM of each primer (forward primer labeled with FAM or JOE; Applied Biosystems, Foster City, CA), 0.5 units AmpliTaq Gold DNA polymerase, 0.1 mg/ml of purified bovine serum albumin and deionized water to 10 µl.
Microsatellite Genotyping
The PCR products were separated on a 5% denaturing, polyacrylamide gel with 1x TBE (0.089 M Tris, 0.089 M borate and 0.0020 M EDTA) buffer in an automated DNA sequencer (ABI 377, Perkin Elmer Inc., Norwalk, CT). Microsatellite allele sizes were scanned using the ABI software GENESCAN 3.1 (Applied Biosystems), and individual genotypes were determined using GENOTYPER 2.1. When ambiguous banding patterns occurred, samples were genotyped again to confirm allele sizes. If the PCR failed or no bands could be scored after multiple trials, the individual sample was re-extracted, and where it failed again, it was abandoned.
Statistical Analyses
Impala population structure was inferred from the spatial distribution of microsatellite genotypes across all the eight loci. In order to cope with the fairly continuous impala distributions we applied two different statistical approaches. First, we ignored any a priori notation of groupings and rather based the analyses on a spatial scale using a model-based Bayesian procedure (Pritchard et al. 2000). Next, we characterized the amount and distribution of genetic variation within and among groups or localities and performed the analyses on sample allele or genotype frequencies. These localities were predefined according to the geographical proximity of individuals and borders between countries (Figure 1).
Individual-Based Analyses
Spatial Structure.
Population structure based on the genotypes of individuals was assessed using a model-based Bayesian procedure, implemented in the program STRUCTURE version 2.1 (Pritchard et al. 2000). This model involves placing individuals into K (unknown) populations of origin and simultaneously assigns the individuals to the populations with explicit estimates of their 90% confidence intervals. These populations of origin are genetic clusters with distinctive allele frequencies, and individuals are probabilistically assigned to one cluster (the population of origin) or more than one cluster (the parental populations) if they are genetically admixed. K is chosen in advance but can be varied across independent runs of the algorithm. Individuals can have membership in multiple clusters, with the membership coefficients summing to 1 across clusters.
We analyzed spatial structure based on the "admixture model" and correlated allele frequencies; runs were based on 106 iterations and a burn-in period of 50,000 iterations. Ten individual repetitions of each estimate of K (2–11) were run in order to check for consistency in the results.
Frequency-Based Analyses
Genetic Diversity and Genotypic Linkage Disequilibrium.
Allele frequencies, observed heterozygosities (HO), and expected heterozygosities (HE) were estimated using the software GENEPOP version 3.1c updated from Raymond and Rousset (1995), as implemented for online access by E. Morgan (http://wbiomed.curtin.edu.au/genepop/). We used the Microsatellite Toolkit ver 3.1 (Park 2001) add-in utility for Microsoft Excel to compute the mean number of alleles (MNA) per locus for the 11 populations.
Deviations of genotypic frequencies from Hardy-Weinberg (HW) proportions were expressed as f, the Weir and Cockerham (1984) estimator of FIS, using randomization tests. Probabilities of significance were estimated using the Markov chain method (dememorization 1,000, batches 1,000, iterations per batch 1,000) (Guo and Thomson 1992) as implemented in GENEPOP. The null hypothesis of no genotypic linkage disequilibrium was tested in all samples, and significance values were computed for each locus pair by unbiased estimates of Fisher's exact tests using the Markov chain method available in GENEPOP. To determine the statistical significance of these tests, the probability values were adjusted for multiple simultaneous table-wide tests using the sequential Bonferroni adjustments (Rice 1989).
Severe reductions in effective population sizes of the samples (i.e., bottlenecks) were tested using the program BOTTLENECK version 1.2.02 (Piry et al. 1999). The tests are based on the fact that populations that have experienced a recent reduction in effective population size exhibit a more rapid reduction in allelic diversity than heterozygosity (i.e., gene diversity HE) at polymorphic loci. Hence, in a recently bottlenecked population, HE (as expected under HW) exceeds the heterozygosity expected in a population at mutation-drift equilibrium (HEQ) (Cornuet and Luikart 1996; Luikart and Cornuet 1998). The two-phase microsatellite mutation model (TPM) is intermediate to the stepwise mutation model (SMM) and the infinite allele model (IAM), and most microsatellite data sets better fit the TPM than the SMM or IAM (Di Rienzo et al. 1994). The TPM recommended by the authors of BOTTLENECK for microsatellites consists of mostly one-step mutations, and a small percentage (5%–10%) of multistep changes (Luikart et al. 1998). We ran the program using a 95% one-step mutations and 5% multistep mutations (with an exponential distribution of step numbers having an average of four steps) (Luikart et al. 1998; Piry et al. 1999). To see if results differed with the mutation model assumed, the program was run also assuming an SMM and IAM. Cornuet and Luikart (1996) and Luikart et al. (1998) proposed three tests, which are implemented in the BOTTLENECK software for the detection of HE excess. Of these, we applied the "sign test" and "Wilcoxon sign-rank test" (the latter recommended by Maudet et al. 2002 and Piry et al. 1999). The "standardized differences test" requires at least 20 polymorphic loci (Cornuet and Luikart 1996) and was therefore not applied to this data set.
Population Structure.
Population structure was quantified using analysis of molecular variance (AMOVA; Excoffier et al. 1992) as implemented by the program ARLEQUIN version 2.000 (Schneider et al. 2000). This uses F statistics introduced by Wright (1951) and modified by Weir and Cockerham (1984) to partition the total genetic variation into specified subdivisions and tests for overall levels of differentiation. The proportion of genetic variation was determined for the following components: between individuals within samples (populations grouped by sites of collection) (FIS), among samples pooled within broad regional scales (southwest, south, and east) (FSR), and among these regional scales (FRT). The significance of each variance component was tested with permutation tests (Excoffier et al. 1992).
The Weir and Cockerham (1984) FST analogue , a measure of among-population variance in allele frequencies, was calculated as pairwise differences between all samples using the ARLEQUIN software and its significance tested by permuting individuals among samples. We did not use Slatkin's (1995) RST value, developed as an analogue to FST, as this takes explicit account of the SMM (Balloux and Lugon-Moulin 2002). Estoup et al. (1995) have shown that population data from compound microsatellites fit the IAM better than the SMM, and hence, the latter is not appropriate for our data, where six out of eight loci are compound or imperfect (Table 1). The linear regression of /(1 – ) against the geographical distance between pairs of common impala sampling sites was calculated, as suggested by Rousset (1997), and tested for significance by a Mantel test (Mantel 1967).
Following Hansen et al. (2001), we used the program GENECLASS2 (Piry et al., 2004, available at http://www.montpellier.inra.fr/CBGP/softwares/index.html) for individual assignment tests as defined by Paetkau et al. (1995). These were used to determine the likelihood of each individual's genotype being found at the location from which it was sampled as described by Cornuet et al. (1999). The individual was excluded from a given candidate population if the individual probability of belonging to the particular population was lower than 1%. To avoid possible bias of self-assignment of individuals to their population of origin, the "leave one out" procedure was followed excluding the tested individual when calculating the allele frequency distribution of their own population. In order to avoid problems potentially caused by nonsampled microsatellite alleles (when a population has a zero frequency for a given allele, individuals having that allele will be assigned a zero likelihood of originating from that population), we used a frequency-based approach (Paetkau et al. 1995) assuming an allele frequency of 0.01 in the event of zero frequency. Probabilities of individuals belonging to populations were calculated using a resampling algorithm developed by Paetkau et al. (2004), with 10,000 simulated individuals, as recommended by the authors of the program.
The extent of divergence among the 11 populations was also quantified by Nei's standard genetic distance DS (Nei 1972). Pairwise distances were used to estimate a population phylogram using the neighbor-joining (NJ) algorithm (Saitou and Nei 1987) available in PHYLIP version 3.5c (Felsenstein 1993). Percent confidence values on tree topology were estimated by 1,000 bootstraps performed resampling loci, and the tree was visualized using TREEVIEW (Page 1996).
Sequence Data
One hundred and fifty-five sequences of the mitochondrial control region (
400 bp) (GenBank accession numbers AF 301740–AF 301894; Nersting and Arctander 2001) from individuals throughout the impala range were reanalyzed. To verify that the sequences were not of nuclear origin, Nersting and Arctander (2001) sequenced the control region fragment, cyt b, from selected samples. Sixty-four of the control region sequences were from the populations analyzed in the present study, and were therefore genotyped using microsatellites. An AMOVA was performed to determine the genetic variations: individuals within samples, among samples within regions (southwest, south, and east) (
SR), and among regions (RT). The tests and the significance of each variance component were determined with permutation tests (Excoffier et al. 1992) in the program ARLEQUIN (Schneider et al. 2000). An NJ tree was estimated from the matrix of Nei's net (DA) number of nucleotide differences between populations (Nei and Li 1979) using PHYLIP 3.5c (Felsenstein 1993) and visualized using TREEVIEW (Page 1996).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAppendix ASupplementary DataReferences |
|---|
HW and Genotypic Linkage Disequilibrium
After sequential Bonferroni correction for multiple tests, no site-by-locus comparison in the common impala was significantly different from HW proportions. There was one deviation in the black-faced sample for locus OarFCB 48 (P < .05), which showed an excess of homozygotes, which could be due to null alleles. Analyzing the seven remaining loci without OarFCB 48 did not result in any significant departures from HW proportions in the populations. However, pooling the black-faced samples into one large population resulted in heterozygote deficiencies at two loci (OarFCB 48 and OarFCB 304), leading to significant departures from HW (P < .05). The deficiency at OarFCB 304 was interpreted as the Wahlund effect, indicating the differentiation between the black-faced impala populations, which were therefore analyzed separately. Only one allele was observed in KAF at OarFCB 48, and no HW test was performed on this site-by-locus comparison.
None of the individual tests for genotypic linkage disequilibrium were significant after Bonferroni correction for multiple testing (data not shown). The variation at the eight loci was therefore seen as statistically independent.
Individual-Based Analyses
Spatial Structure.
Population structure was assessed using the entire sample set (n = 162). At K = 2, the clusters were anchored by subspecific origin (Figure 2a). Increasing the value of K to 3 resulted in the common impala group splitting into two groups, conforming to the geographical proximity of the samples (Figure 2b). CH, SH, KAF, and LU in the south clustered together, as did LM and SA in the east. SE was assigned with almost equal probability to both common impala clusters (49% to south, 51% to east). BK showed an intermediate clustering probability (26% to south, 73% to east) between values observed for SE and those observed for LM and SA, the latter clustering together with very high probability (94.2% and 98.5%, respectively). The two clusters representing the southern and eastern common impala samples remained the same for K = 4. The proportion of black-faced individuals assigned to each population was roughly symmetric for K = 4 and K = 5 (Figure 2c,d), although the latter split the common cluster obtained from the previous value. SA was the only population that retained a proportion of membership >0.80 to one of the clusters (q = 0.90). LM, which also had a highly significant value in the pairwise comparisons (see subsequently), had a q value of 0.67 for belonging to another of the clusters.
|
Frequency-Based Analyses
Genetic Diversity.
In total, 82 alleles were observed across the eight loci surveyed. The number of alleles per locus ranged from 4 (OarFCB 48) to 15 (OarFCB 304) (Appendix, Table A1), with an average of 10 alleles per locus.
The common impala subspecies had a higher number of unique alleles (35% of the alleles genotyped) with a total cumulative frequency (Cfreq) of 20%. The black-faced impala subspecies showed only two unique alleles (Cfreq = 0.7%). Within the common impala subspecies, the southern samples had 15 alleles not found in the eastern samples (Cfreq = 8%). The eastern samples had 16 alleles not found in the southern samples (Cfreq = 7%).
The amount of genetic variability in the samples was estimated as the number of alleles per locus and as observed (HO) and expected (HE) heterozygosity (the latter is shown in the Appendix, Table A1). The measures varied substantially among loci and samples. Average values of observed and expected heterozygosity were HO = 0.61 (min = 0.10, max = 1) and HE = 0.67 (min = 0.19, max = 0.98). Within the common impala, both the average number of alleles and the observed heterozygosity tended to be slightly higher in the southern samples than in the east. The highest average heterozygosity was observed in the southern common samples CH and SH (78% and 70%, respectively). These also had relatively high MNA values (5.8 and 5.3, respectively) exceeded only by SE (MNA = 6.4). The lowest average heterozygosity across all samples was observed in SA (48%), which was also the least variable (MNA = 3.75).
Recent bottlenecks in the history of the impala were not statistically supported. None of the studied samples exhibited a statistically significant excess of HE over HEQ across the eight microsatellites under any of the microsatellite mutation models assumed, regardless of the method applied.
Quantitative estimates of hierarchical gene diversity indicated significant genetic population structure at every level where 19.3% of the genetic variation occurred between regions, 5.1% was due to differentiation among samples within each region, and 75.6% was due to variation in samples (Table 2). The largest percentage of within-region variation among samples was observed in east Africa (SR = 0.14), attributed in part to the genetic distinctiveness of SA (with this sample removed, SR = 0.09), when compared to the southwest and south (with SR = 0.05 and 0.03, respectively).
|
Population Structure.
In the pairwise
comparisons, all but three comparisons within the common impala showed significant differentiation between samples at the 5% level (Table 3). This was within the southern common impala, where KAF was not significantly differentiated from CH, SH, or LU. Within the east African samples, all pairwise comparisons were highly significant (P < .001). At the 1% level, the pairwise comparisons between KAF and SE and between CH and SH no longer retained significance, and lowering the P value to the .01% level resulted in the comparison between KAF and BK no longer being significant.
|
Figure 3 shows the regression of
/(1 – ) against geographical distance. A Mantel test showed that the positive association between genetic differentiation and geographical distances separating samples was significant (P < .01).
|
Results of the assignment tests are depicted in Table 4. The black-faced impala individuals not assigned to sample of origin were all assigned to another black-faced impala sample. Regarding the common impala, all individuals within the southern group were assigned to region of origin; no individuals were assigned to samples in the east African region. Within east Africa, the geographically intermediate samples from SE and BK had individuals assigned to samples in southern Africa and individuals from LM and SA not assigned to their population of origin were assigned to other east African samples.
|
The genetic distances of the microsatellites between impala samples is illustrated in Figure 4 as an NJ tree. There is a marked topology in the tree, the genetically distinct and geographically isolated black-faced impala forming a sister group to the southern common impala samples with a bootstrap value of 99. Distances between black-faced and southern impala ranged between 0.55 and 1.53 (mean DS = 0.93) (Table 3). Furthermore, a clear grouping within the common samples along a south to east gradient was indicated. CH, SH, KAF, and LU in the south clustered separately from SE, BK, LM, and SA in the east with a bootstrap value of 86. SA was separated from the remaining east African samples by a strikingly long branch (DS = 0.40–0.57; mean = 0.50). Distances between SA and the southern common impala samples (CH, SH, KAF, and LU; DS = 0.95–1.34; mean = 1.06) exceeded those separating the black-faced impala from the southern commons.
|
Sequence Data
The AMOVA performed on the sequence data revealed a high degree of genetic substructure among regional groups (
RT = 0.34; Table 2). Within the east African region, the SR value was appreciably higher than those observed in the southern region (0.29 and 0.06, respectively). In the NJ tree based on DA between populations, there was a clear segregation of the black-faced impala populations and the common impala populations into two separate clades. No regional structure within the common impala samples was revealed using control region data. Interestingly, the sample from SA was separated from the rest of the common impala by a marginally longer branch (data not shown).
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAppendix ASupplementary DataReferences |
|---|
Spatial Population Structure
The use of highly polymorphic microsatellite loci and assignment tests has greatly increased the potential for understanding population structure across a landscape. The Bayesian approach as implemented in STRUCTURE has proved to be a powerful instrument in many instances where genetic differentiation is high (Maudet et al. 2002; Pritchard et al. 2000; Randi and Lucchini 2002; Randi et al. 2001; Rosenberg et al. 2001) and can also be used when population structure is more complex (Rosenberg et al. 2002). Correspondence between regional affiliation and genetic ancestry has previously been reported (Bowcock et al. 1994; Manel et al. 2002; Pritchard et al. 2000; Rosenberg et al. 2002), and some regional genetic structuring within the impala was detected without using prior geographical information.
With K = 2 (Figure 2a), clusters were anchored by the two presently recognized subspecies of impala: A. m. petersi and A. m. melampus (Kingdon 1997; Shortridge 1934; Skinner and Smithers 1990; Swart 1967). This was supported by the NJ tree, where the geographically isolated black-faced impala constituted a conspicuous and well-supported sister group to the common impala using both microsatellite and mtDNA data (Figure 4). With K = 3 (Figure 2b), the common impala was split into two clusters across its spatial scale, representing southern and eastern regions. Selous National Reserve—located in southeast Tanzania midway between south and east—clustered in both with equal probability (Figure 2b). The genetic affiliation of eastern samples to the southern cluster decreased with geographic distance; Burko Forest Reserve in northern Tanzania had a correspondingly lower probability of clustering in the south, revealing a level of genetic admixture between the two regions. A certain degree of regional affiliation within the common impala was also revealed using prior geographic information in the assignment tests (Table 4). Individuals from Selous and Burko were assigned (with decreasing fractions) to southern samples. Recent gene flow between previously allopatric regional groups harboring separate evolutionary histories or a degree of isolation by distance of one large population could explain the mixed clustering of the intermediate samples. However, some caution should be expressed in the interpretation of the results as sample sizes for the intermediate KAF and LU were small (n < 10). The regional split of the common impala into a south and an east African group was well supported in the microsatellite distance tree (Figure 4) but was not obvious in the haplotypic tree (Figure 3; Nersting and Arctander 2001). Reanalyzing the mtDNA data using populations as operational taxonomic units did not uncover any regional structuring within the subspecies.
Pairwise comparisons (Table 3) revealed a degree of regional affiliation between southern common impalas and gene flow between south and east. Within the eastern region, all pairwise comparisons were highly significant, reflecting a high degree of differentiation between samples. Within the southern region, however, only two pairwise comparisons were highly significant, suggesting a higher level of gene flow or less genetic differentiation among samples in the south. Furthermore, in concordance with the assignment tests, the genetic affinity between samples from the two regions increased with decreasing geographical proximity. Comparisons between geographically intermediate Kafue National Park in Zambia (southern region) and Selous and Burko in Tanzania (eastern region) revealed low genetic differentiation.
In the individual-based analyses, increasing the value of K to 4 (Figure 2c) resulted in the splitting of the black-faced impala into two groups. As significant genetic structuring is detected only if the underlying genetic differentiation is high, the pattern observed suggested a lack of genetic structuring within the black-faced subspecies, a higher degree of genetic differentiation being found within the common impala. This was supported by the assignment tests, where no correlation was found between the fraction of black-faced individuals assigned to a location other than their origin and the proximity of that location to the sampling site. This was expected; the black-faced impala populations in Etosha National Park are derived from 280 individuals captured close to the Angolan border and released in the park around 1970 (Green and Rothstein 1998; Lorenzen and Siegismund 2004).
Higher genetic diversity values (MNA and HE; Appendix, Table A1) were observed in the southern African common impala than in the southwest or east, Chobe National Park (Botswana) and Shangani (Zimbabwe) being the most diverse. Selous (southeast Tanzania) had the highest observed MNA, likely an effect of being an intermediate compilation of the disparate genetic clusters in south and east (Figure 2b). Samburu (Kenya) had the lowest HE and MNA observed across all samples. This was also seen in the mtDNA analyses (Nersting and Arctander 2001), where Samburu had a markedly lower within-population nucleotide diversity value (0.75%). Highest values were again observed in the south (3.66% in Chobe, 2.64% in Shangani), Burko and Lake Mburu National Park in Uganda showing intermediate values (1.11% and 1.14%, respectively).
Microsatellites attributed a greater component of the total variance to the variance within samples than did the AMOVA on control region sequences (Table 2). This was expected, as microsatellites mutate at a higher rate than the mitochondrial control region. Therefore, a higher degree of variation is found in the former, which is apparent at the individual level (Estoup and Angers 1998). The east African region showed a higher degree of among-sample variation and lower within-sample diversity (Appendix, Table A1) in both analyses. This was not only caused by the marked genetic divergence of the Samburu impala supported by microsatellite and mtDNA data, individual-based analyses, a markedly lower MNA and HE, and highly significant values in all pairwise comparisons but also caused by greater fragmentation within this region. The control region analysis partitioned a greater percentage of the overall variance to differences among the three geographical regions, which was not surprising given the genetic distinctiveness of the black-faced haplotypes (see Nersting and Arctander 2001 for further details). Even though lower on an absolute scale, the differentiation among regions at the microsatellites approaches about half the maximum possible. As Hedrick (1999) showed, the upper limit for FST [we used the Weir and Cockerham (1984) analogue ] in highly variable markers is less than the expected average homozygosity in the samples, which is 0.38 in our case.
Population History
The south/east contrast seen in the spatial genetic structuring of the common impala could be a reflection of distinct regional evolutionary histories and is supported by the individual-based analyses and the distance tree. Subsequent gene flow after secondary contact could explain these results coupled with the diversity indices, pairwise comparisons, assignment tests, and the regression analysis. Given that previous studies of several other species of African bovids have shown the same pattern of vicariance, sable antelope (H. niger), roan antelope (H. equinus), hartebeest (A. buselaphus), topi (D. lunatus), wildebeest (C. taurinus), kudu (T. strepsiceros), and buffalo (Syncerus caffer) (Arctander et al. 1999; Mathee and Robinson 1999; Nersting and Arctander 2001; Van Hooft et al. 2002), argues for such an interpretation.
Conversely, the considerable gene flow between south and east—the degree of genetic affiliation increasing with geographic proximity—could be the result of isolation by distance, the distinctiveness of the east African localities being the result of discontinuous sampling of discrete populations. The higher genetic diversity values and lacking differentiation between localities in the south coupled with the genetic distance tree could suggest a colonization scenario toward the east from the large, long-standing, panmictic, and more genetically diverse southern population. A similar pattern has previously been proposed for the wildebeest (Connochaetes taurinus), whose present distribution range in east Africa is believed to be the result of a more recent expansion from southern Africa (Arctander et al. 1999). However, it is difficult to distinguish history from recurrent processes within the common impala subspecies from our results.
The large amount of alleles unique to the common impala subspecies, their high cumulative frequency, the larger haplotypic diversity, and the greater genetic variability suggest that the black-faced subspecies resulted from a migration westward. The microsatellite distance tree supports a migration from the southern region; however, this could reflect frequency similarity rather than population history. Subsequent prolonged isolation of this founding population in the southwest, perhaps during the climatic fluctuations of the Pleistocene (deMenocal 1995; Partridge et al. 1995), has resulted in the genetically and morphologically distinct subspecies. A Pleistocene refuge for arid-adapted species in southwest Africa has previously been proposed for the roan antelope (Mathee and Robinson 1999), the topi (Arctander et al. 1999), and the kudu (Nersting and Arctander 2001). However, if the impala was previously distributed throughout southern Africa, subsequent allopatry between the southwestern and southern populations could have resulted in the present subspecific distributions. With respect to the Etosha populations analyzed, as there has been no gene flow between locations since they were established 30 years ago, the level of genetic differentiation between populations can be seen as a measure of the degree of founder effect and genetic drift within the park since the populations were established.
Individuals from Samburu National Reserve in Kenya, the most northern locality within the impala distribution range, continuously grouped within one cluster with very high probabilities in the individual-based analyses (Figure 2e), emphasizing the genetic distinctiveness of this population. This coupled with microsatellite (Figure 4) and mtDNA distance data, a markedly lower MNA and HE (Appendix, Table A1) and highly significant values in all pairwise comparisons (Table 3), supported the marked genetic divergence of the Samburu population. This could be explained by the reported significant decrease in impala numbers in the Samburu National Reserve in the 1970s and the 1980s (East 1997). During this period, the deterioration of Kenya's wildlife sector reached crisis proportions. An unprecedented wave of poaching reduced the country's elephant population by >85% and the black rhino population by >98%. Parks were affected by inadequate equipment for antipoaching patrols and road maintenance, illegal encroachment of cattle, and rampant corruption at entry points where vital revenue was being lost (Leakey 1988). This affected wildlife populations across species across the entire country. The impala population in Samburu was seriously affected but has since stabilized and now numbers some 800 individuals (East 1997). Although not statistically supported, this circumstantial evidence could indicate a bottleneck in the recent history of this population. Another reason for the genetic distinctiveness of the Samburu impala could perhaps be found in the subspecific distributions of the kudu (T. strepsiceros). The kudu is an antelope with the same wide, congruent distribution range as the impala, with equivalent habitat requirements. Three subspecies are recognized by Kingdon (1997), with the subspecies Tragelaphus strepsiceros chora being confined to north-east Africa. The southernmost range of T. s. chora is Samburu. If a subspecies with a similar distribution range has existed in the history of the impala, and since perished in the north, the impala in Samburu National Reserve could be the last remnant of a genetically distinct impala subspecies in north-east Africa.
|
Conclusions |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAppendix ASupplementary DataReferences |
|---|
In conclusion, individual-based analyses using microsatellites uncovered regional structuring between southern and east African common impala (A. m. melampus), not previously observed using mitochondrial control region data. We propose that the genetic differences within this subspecies reflect isolation by distance after a recent colonization from a long-standing population in southern Africa toward the east and north. The isolation has not been for a duration long enough to be evident with mtDNA. This was supported by high levels of genetic admixture only within and between southern impala and geographically intermediate samples from the east, a progressive decline in genetic diversity toward the east, and the lacking morphological diagnostics of impala from the two regions (Kingdon 1997; Nowak 1999). Both microsatellite and mtDNA data indicated a genetic distinctiveness of the Samburu population in Kenya.
|
Appendix A |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionsAppendix ASupplementary DataReferences |
|---|
Table A1.. Measures of genetic variation across eight microsatellite loci in black-faced and common impala by collection site
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||