Conservation Genetics

, Volume 11, Issue 5, pp 1837–1846

Genetic signatures of population change in the British golden eagle (Aquila chrysaetos)

Authors

    • School of GeographyUniversity of Nottingham
    • Sheffield Molecular Genetics Facility, Department of Animal and Plant SciencesUniversity of Sheffield
  • Alain C. Frantz
    • Sheffield Molecular Genetics Facility, Department of Animal and Plant SciencesUniversity of Sheffield
  • Christopher P. Lavers
    • School of GeographyUniversity of Nottingham
  • Angus Davison
    • Institute of Genetics, School of BiologyUniversity of Nottingham
  • Deborah A. Dawson
    • Sheffield Molecular Genetics Facility, Department of Animal and Plant SciencesUniversity of Sheffield
  • Terry A. Burke
    • Sheffield Molecular Genetics Facility, Department of Animal and Plant SciencesUniversity of Sheffield
Research Article

DOI: 10.1007/s10592-010-0076-x

Cite this article as:
Bourke, B.P., Frantz, A.C., Lavers, C.P. et al. Conserv Genet (2010) 11: 1837. doi:10.1007/s10592-010-0076-x

Abstract

The golden eagle (Aquila chrysaetos) was once widely distributed in the uplands of the British Isles, but is now extinct in Ireland, and largely confined to the highlands and islands of Scotland. As the precise extent and severity of the reduction in population size are unclear, it is important to understand how the population was affected by the decline. We therefore genotyped 13 polymorphic microsatellite loci in 172 individuals from the contemporary British population and compared their genetic diversity to 70 British and 9 Irish museum specimens. Despite the recent population decline, there is only slight evidence for a concomitant loss of genetic variation. Instead, two likelihood-based Bayesian methods provided evidence for a severe ancient genetic bottleneck, possibly caused by the fragmentation of a large mainland European population and/or the founding effects of colonising the British Isles. As the population persisted despite this ancient bottleneck, our conclusion is that there is limited need for intervention to augment the present-day genetic diversity. The main short-term objective of conservation measures should be to increase population sizes by continuous safeguarding of individuals and habitat management. Finally, we also confirmed that, for management purposes, the species should be considered a single population unit and that the extinct Irish population was not differentiated from the British one.

Keywords

Golden eagleAquila chrysaetosMicrosatellitePopulation geneticsMSVAR

Introduction

The assessment of population genetic diversity plays an important role in conservation strategies of threatened and endangered species. It is believed that genetic variation is required for individuals to adapt to environmental changes and for populations to remain viable (Reed and Frankham 2003). Severe reductions in population size may lead to a loss of genetic diversity due to drift and to an increase in the frequency of rare deleterious recessive alleles, resulting in reduced fitness through inbreeding depression. The consequences of such events would be increased risk of population extinction (Frankham 1998; Brook et al. 2002).

While it is useful to identify potentially deleterious effects of reduced population size, genetic information can also be used to understand ecological patterns that can be of great importance to conservation. For example, historic (Slatkin 1987) or contemporary levels of dispersal and the location of barriers to gene flow (Pritchard et al. 2000; Corander et al. 2003) are now routinely investigated to help maintain demographic patterns and processes, to identify demographically distinct populations, and to determine whether populations require translocation or supportive breeding programs to enhance genetic variability and minimise extinction risk (Ryman and Laikre 1991; Moritz 1999; Janssens et al. 2008).

The golden eagle (Aquila chrysaetos) was once widely distributed in the uplands of the British Isles (Ussher and Warren 1900; Love 1982; Watson 1997), but is now extinct in Ireland and largely confined to the highlands and islands of Scotland (Watson 1997). While it is believed that this decline resulted from a combination of human population expansion in rural areas, habitat alterations and persecution, the precise extent and severity of the population crash are unclear, as is its effect on the genetic variability of the population. A study of breeding records over the past two centuries suggests that the greatest decline occurred in the late nineteenth and early twentieth century, with an estimated population nadir of perhaps 100–200 breeding pairs prior to legal protection in 1954 (Love 1982; Watson 1997; Watson and Whitfield 2002; Whitfield et al. 2008). Although some recovery has occurred in recent decades, breeding attempts outside Scotland remain ephemeral (Deane 1962; Katz 2004) and there has been no natural recolonisation of Ireland (O’Toole et al. 2002).

The British and historical Irish populations have long been considered continuous on the basis of similar ecology and geographic proximity, and there is some evidence of breeding territories overlapping with Scotland and Ireland (Deane 1962). Because golden eagles were extirpated from Ireland in the early twentieth century, a reintroduction project was initiated in 2001 that assumed birds from the British population represented the closest genetically related donor stock, thus fulfilling one of the key IUCN criteria for reintroduction (IUCN 1996; O’Toole et al. 2002). However, no genetic data existed to verify this assumption.

As a result of the ongoing threats to the golden eagle in the British Isles, it is important to understand how the population was affected by the reduction in population size that occurred before the species was legally protected. Specifically, we use data from microsatellite markers to investigate the severity of the loss of genetic diversity as a result of recent demographic history. We estimate levels of genetic diversity observed in the modern population and compare these with values obtained from museum specimens collected prior to the introduction of legal protection in 1954, including samples from historic British and extinct Irish populations (O’Toole et al. 2002). This approach has been successfully used either to confirm (e.g., Groombridge et al. 2000; Johnson et al. 2004) or reject (e.g., Paxinos et al. 2002) the negative influence of recent population declines on contemporary levels of genetic diversity. In order to make inferences about the severity of the recent population crash, we analyse the contemporary data for evidence of temporary excess in heterozygosity resulting from a recent severe reduction in effective population size (Cornuet and Luikart 1996) and investigate past population dynamics using methods that quantify the rate of change in population size (Beaumont 1999), and estimate the time since population size started to change and the effective population sizes at that time and in the contemporary sample (Storz and Beaumont 2002).

A further objective of this study was to assess genetic structure in both the historic and current populations of the golden eagle in the British Isles (O’Toole et al. 2002) to determine the suitability of the British birds as donor stock for reintroduction attempts in Ireland. Also, the presence of genetic partitioning over this relatively small geographical scale would have important consequences for the conservation management of the species.

Materials and methods

A total of 172 samples (30 blood samples and 142 feather samples) were collected from separate breeding territories across the current breeding population in the British Isles (Figs. 1, 2; Table 1). Blood was collected from nestlings and stored in 100% ethanol. Adult feathers were collected from the roosts and nests of known breeding territories. Genomic DNA was extracted from blood using an ammonium acetate salt extraction method (Nicholls et al. 2000) and from feathers using the QIAGEN DNeasy tissue kit, in accordance with the sampling technique of Horvath et al. (2005). Polymerase chain reactions (PCR) of 13 microsatellite loci (Table 2) were performed after Bourke and Dawson (2006). These loci were cross-utilised from Aquila adalberti, Aquila heliaca, and Haliaeetus Albicilla. Two loci were excluded from the current study: Hal-10 showed evidence of physical linkage and IEAAAG09 had poor amplification success. In addition, we obtained foot tissue samples from 79 museum specimens (70 from Britain and nine from Ireland). Because the timing of the population nadir remains uncertain we collected samples dating from the early 1790s to the late 1940s and prior to the introduction of legal protection under the Wildlife and Countryside Act, 1954. Approximately one-third of specimens (n = 28) are dated to the period 1790–1899, another third (n = 24) date from 1900 to 1930, and the remainder (n = 27) date from 1931 to 1949. Consensus genotypes were obtained from feather and museum samples as a precaution against genotyping errors from low quantity/quality DNA. These were scored based on three replications: homozygotes were accepted when all three replicates were homozygous; heterozygotes were accepted when two of three replicates were heterozygous.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-010-0076-x/MediaObjects/10592_2010_76_Fig1_HTML.gif
Fig. 1

Grey areas represent the current breeding distribution of the golden eagle in the British Isles (Royal Society for the Protection of Birds data available at http://www.rspb.org.uk/wildlife/birdguide). Dotted ellipses represent the former distribution of the extinct Irish population (Ussher and Warren 1900)

https://static-content.springer.com/image/art%3A10.1007%2Fs10592-010-0076-x/MediaObjects/10592_2010_76_Fig2_HTML.gif
Fig. 2

Regions of Great Britain taken from Eaton et al. (2007) used to summarize sample distribution and survey results (Table 1). A eastern Highlands; B northern moors and flows; C north-central Highlands; D south-central Highlands; E northwest Highlands; F west Highlands; G southwest uplands and north England; H Hebridean islands

Table 1

Distribution of samples from active territories in Great Britain

Region code

Region name

Sampled territories

Active territories 2003

A

Eastern Highlands

14

31

B

Northern moors and flows

17

21

C

North-central highlands

14

48

D

South-central Highlands

15

31

E

Northwest Highlands

16

45

F

West Highlands

27

77

G

Southwest uplands and north England

35

69

H

Hebridean islands

34

120

 

Total

172

442

Population data taken from the 2003 golden eagle survey (Eaton et al. 2007)

Table 2

Contemporary and historic golden eagle populations: number of individuals sampled (N), number of alleles per locus (A), allelic richness (Ar), mean observed (Ho) and expected (He) heterozygosity

Locus

Contemporary

Historic

N

A

Ar

He

Ho

N

A

Ar

He

Ho

Aa11

111

5

5.0

0.67

0.70

41

6

6.0

0.65

0.68

Aa12

157

5

3.7

0.58

0.57

72

4

3.9

0.61

0.56

Aa15

148

3

3.0

0.50

0.50

70

3

3.0

0.45

0.44

Aa26

120

5

4.1

0.64

0.77

61

5

4.8

0.63

0.62

Aa27

146

3

3.0

0.21

0.23

n/a

n/a

n/a

n/a

n/a

Aa36

146

6

6.0

0.60

0.64

n/a

n/a

n/a

n/a

n/a

Aa39

115

5

5.0

0.66

0.70

48

6

6.0

0.66

0.73

Aa43

133

3

2.8

0.17

0.15

68

4

3.4

0.10

0.10

Hal-13

145

3

3.0

0.55

0.61

67

3

3.0

0.56

0.43

IEAAAG04

128

5

4.8

0.61

0.66

58

7

6.3

0.57

0.62

IEAAAG13

144

3

2.5

0.39

0.42

69

3

2.8

0.42

0.45

IEAAAG14

145

3

2.3

0.15

0.15

65

3

2.6

0.22

0.22

IEAAAG15

158

4

4.0

0.51

0.55

72

4

3.8

0.54

0.60

Mean

 

4.1

3.8

0.48

0.51

 

4.4

4.2

0.49

0.50

We calculated mean number of alleles per locus (A), observed (Ho) and expected (He) heterozygosities (Nei 1978) for each locus, as well as the average expected heterozygosity, using GENETIX 4.05.2 (Belkhir 2004). A measure of allelic richness was included to control for differences in the number of alleles among populations that differ in sample size, calculated using the program FSTAT v2.9.3 (Goudet 2001). This analysis was performed on the 11 loci that could be scored in both the modern and the historic datasets. The data were tested for linkage equilibrium using an exact test based on a Markov chain method as implemented in genepop 3.4 (Raymond and Rousset 1995). The same program was used to perform the exact tests of Guo and Thompson (1992) for deviations from Hardy–Weinberg (HW) genotypic proportions at each locus. The sequential Bonferroni technique was used to eliminate significance by chance (Rice 1989). The Wilcoxon signed rank test (Wilcoxon 1945) was used to test for significant differences between historical and modern population parameters.

Program STRUCTURE v2.2 (Pritchard et al. 2000) was used to test the modern and historic dataset, as well as a combination of both, for genetic structure by estimating the number of subpopulations or clusters (K) present in the various datasets. Ten independent runs of K = 1–10 with 106 MCMC iterations and a burn-in period of 105 were performed, using the model with correlated allele frequencies and assuming admixture. For each value of K, the log-likelihood values were averaged and standard deviations calculated. The program Arlequin 3.1 (Schneider et al. 2000) was used to calculate FST values between historic and modern samples (Weir and Cockerham 1984).

We performed spatial autocorrelation analyses to assess the modern dataset for fine-scale genetic structure; i.e. we calculated the relationship between genetic relatedness between pairs of individuals and their corresponding pairwise geographical distance, using program SPAGEDI 1.2 (Hardy and Vekemans 2002). The slope of this relationship does not depend on an arbitrary choice of distance classes and therefore offers a convenient measure of the degree of spatial genetic structuring (Hardy and Vekemans 2002). As suggested by Vekemans and Hardy (2004), the kinship coefficient (Fij) presented in Loiselle et al. (1995) was chosen as a pairwise estimator of genetic relatedness, as it is a relatively unbiased estimator with low sampling variance. In order to be meaningful, the analysis should be limited to individuals of reproductive age. Inclusion of juveniles potentially biases results towards greater correlation between spatial and genetic distances (Coltman et al. 2003; Frantz et al. 2008). In addition to limiting the analysis to feather samples collected from adults, we omitted genotypes that had been scored at a maximum of five loci or for which no geographic coordinates were available.

To test the contemporary samples for evidence of a recent genetic bottleneck, allele frequency data were tested for ‘heterozygosity excess’ (HE) using the program BOTTLENECK 1.2.02 (Cornuet and Luikart 1996). We computed deviations from expected heterozygosity for every locus using two of the most appropriate mutation models for microsatellites, i.e. stepwise mutation model (SMM) and the two-phased model of mutation (Di Rienzo et al. 1994; Valdes et al. 1993). We used a variance of 30 for the TPM and assumed 30% of multi-step changes. Two statistical tests (sign test, and Wilcoxen signed-ranks test) were conducted to test for significant HE, which may indicate that a recent bottleneck has occurred. We performed the tests excluding the locus that deviated from Hardy–Weinberg proportions (Aa26, see Results). A third test available in BOTTLENECK 1.2.02 (standardised differences test) was not performed as it needs a minimum of 20 loci to produce reliable results.

We also tried to infer past population dynamics using two likelihood-based Bayesian methods that use information from the full allelic distribution in a coalescent–based framework under the SMM. The method by Beaumont (1999), implemented in program msvar 0.4, assumes that a stable population of size N1 started to either increase or decrease ta generations ago to a current population size N0. This change in population size can either be linear or exponential and mutations are assumed to occur with a rate of θ = 2 N0μ, where μ is the locus mutation rate. These assumptions can be used to estimate the posterior probability distribution of (1) the rate of population size change (r = N0/N1), (2) the time since the population started to expand or contract, scaled by the current population size (tf = ta/N0) and (3) θ = 2 N0μ. Sampling from the posterior distribution of parameter r using a Marlov Chain Monte Carlo (MCMC) method thus allowed us to quantify the change in population size. In this method, rectangular prior distributions are assumed for the log(r), log(tf) and log(θ).

The method by Storz and Beaumont (2002) is an extension of Beaumont’s method and allows quantification of the effective population sizes N0 and N1, μ, as well as the time T since the population started changing. The method, implemented in program msvar 1.3, assumes an exponential change in population size. Here, prior distributions for the parameters are assumed to be log-normal. The means and standard deviations of these prior log-normal distributions are themselves drawn from prior (or hyperprior) distributions. Hyperpriors for the means were specified by normal distributions with a mean of α and a standard deviation of σ. Hyperpriors for the standard deviations were assumed to be zero-truncated normal distributions with a mean of β and a standard deviation of τ. Generation time was calculated following Pianka (1978; T = (a + b)/2, where a = age at maturation and b = longevity). Age of maturation and longevity was estimated at 5 and 20 years, respectively (Watson 1997), giving a generation time of 12.5 years.

Two loci (Aa27 and Aa36) had repeat length variations that were not a consistent multiple of two or four and were therefore excluded from the MSVAR analyses. When analysing the data using msvar 0.4, we performed nine independent runs using different starting values (Table 3) and random seeds. Bounds for the prior distributions were set to 10−5 and 105 (on a log scale). In the case of msvar 1.3, 14 independent runs were performed using different random seeds, starting values, priors and run lengths (Table 4). Similar to Goossens et al. (2006), runs with the same priors (assuming that N1 and N0 were of the same size) were repeated five times as a test for convergence.
Table 3

Starting values for the MCMC runs of program msvar 0.4 (Beaumont 1999)

 

Starting values

θ

r

tf

Run 1

0.1

0.1

0.1

Run 2

1

0.1

1

Run 3

10

0.1

10

Run 4

0.1

1

0.1

Run 5

1

1

1

Run 6

10

1

10

Run 7

0.1

10

0.1

Run 8

1

10

1

Run 9

10

10

10

The same values were used for the linear and exponential models of change in population size

Table 4

Parameters for the MCMC runs of program msvar 1.3 (Storz and Beaumont 2002)

 

Starting values (mean, variance) for

Hyperpriors (α, σ, β, τ) for

Run lengths

Log(N0)

Log(N1)

Log(μ)

Log(T)

Log(N0)

Log(N1)

Log(μ)

Log(T)

Steps

Thinning

Iterations

Run 1

4, 1

4, 1

−3.5, 1

2, 1

3.4, 2, 0, 0.5

5, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

2, 3, 0, 0.5

105

104

109

Run 2

4, 1

4, 1

−3.5, 1

2, 1

3.4, 2, 0, 0.5

4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

3, 3, 0, 0.5

105

104

109

Run 3

4, 1

4, 1

−3.5, 1

2, 1

3.4, 2, 0, 0.5

3.4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

4, 3, 0, 0.5

105

104

109

Run 4

4, 1

4, 1

−3.5, 1

2, 1

3.4, 2, 0, 0.5

3.4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

2, 3, 0, 0.5

105

5 × 104

5 × 109

Run 5

4, 1

4, 1

−3.5, 1

2, 1

3.4, 2, 0, 0.5

3.4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

2, 3, 0, 0.5

105

5 × 104

5 × 109

Run 6

4, 1

4, 1

−3.5, 1

2, 1

3.4, 2, 0, 0.5

3.4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

2, 3, 0, 0.5

105

2 × 104

2 × 109

Run 7

4, 1

4, 1

−3.5, 1

2, 1

3.4, 2, 0, 0.5

3.4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

2, 3, 0, 0.5

105

5 × 104

5 × 109

Run 8

4, 1

4, 1

−3.5, 1

2, 1

5, 2, 0, 0.5

5, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

2, 3, 0, 0.5

105

104

109

Run 9

4, 1

4, 1

−3.5, 1

2, 1

5, 2, 0, 0.5

4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

5, 2, 0, 0.5

105

104

109

Run 10

4, 1

4, 1

−3.5, 1

2, 1

5, 2, 0, 0.5

3.4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

5, 2, 0, 0.5

105

104

109

Run 11

4, 1

4, 1

−3.5, 1

2, 1

3.4, 2, 0, 0.5

3.4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

5, 3, 0, 0.5

105

104

109

Run 12

4, 1

4, 1

−3.5, 1

3, 1

3.4, 2, 0, 0.5

3.4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

2, 2, 0, 0.5

105

104

109

Run 13

5, 1

5, 1

−3.5, 1

4, 1

3.4, 2, 0, 0.5

3.4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

5, 3, 0, 0.5

105

104

109

Run 14

6, 1

6, 1

−3.5, 1

5, 1

3.4, 2, 0, 0.5

3.4, 3, 0, 0.5

−3.5, 0.25, 0, 0.5

5, 3, 0, 0.5

105

104

109

For more information, see “Materials and methods” section

For all MSVAR-type analyses, we discarded the first 10% of the updates to avoid biases due to starting conditions. The outputs were analysed using a script for program R (Ihaka and Gentleman 1996) kindly provided by Vitor Sousa (Instituto Gulbenkian de Ciência). Convergence of individual chains was checked visually and convergence among chains was tested with the Gelman and Rubin (1992) statistic using the R package coda. A point estimate of <1.2 is an indicator of good convergence (Gelman and Hill 2007). When estimating parameter r using the linear model in msvar 0.4, a point estimate of 1.04 for the convergence criterion was obtained, with a 97.5% quantile of 1.15. For all other relevant parameters estimated using either method, we obtained point estimates ≤1.04 with 97.5% quantiles ≤1.08. Independent runs for the same parameter thus gave similar posterior distributions and were pooled into one dataset to produce larger samples of the posterior distribution.

Results

All of the 172 contemporary samples yielded amplifiable DNA, and the mean proportion of individuals typed was 0.802. All 79 historic samples yielded amplifiable DNA, and the mean proportion of individuals typed was 0.795. The 13 loci genotyped in the contemporary samples were all polymorphic, with between three and six alleles per locus and a mean of 4.1 (Table 2). Two loci (Aa27 and Aa36) displayed high levels of stuttering in the historic population and could not be scored unambiguously. These were therefore excluded from further analysis. The remaining 11 loci genotyped in the historic samples had between three and seven alleles per locus and a mean of 4.4.

The average He value for the modern samples of 0.48 was lower than the average Ho value of 0.51 but the difference was not statistically significant. Only locus Aa26 deviated significantly from Hardy–Weinberg proportions (P = 0.0013). This deviation was also significant after sequential Bonferroni correction (P < 0.0038). Fifteen pairs of loci were in linkage disequilibrium (P < 0.05) but these did not remain significant after sequential Bonferroni correction (P < 0.0006). In the historic samples, the average heterozygosity values were identical (He = 0.50; Ho = 0.50). Locus Hal-13 deviated significantly from Hardy–Weinberg proportions (P = 0.019). However, this deviation was not significant after sequential Bonferroni correction (P > 0.0045). Three pairs of loci were in linkage disequilibrium before sequential Bonferroni correction, but not after (P > 0.0009).

When analysing the contemporary data with the program STRUCTURE, the highest log-likelihood values were obtained for K = 1 (Fig. 3), suggesting the absence of population genetic structure in the modern British population of the golden eagle. Similarly, the program did not provide evidence for genetic structure in the historical samples. The historical and contemporary population samples were not significantly differentiated (FST = 0.003; P = 0.105). This was confirmed by STRUCTURE, which, when analysing both sets of data together, inferred the highest log-likelihood values for K = 1. The same conclusion was reached when analysing the modern samples with the historical Irish samples only. The spatial autocorrelation analysis, performed on 92 contemporary individuals, did not provide evidence for fine scale genetic structure as the slope of the correlation between pairwise spatial and genetic distances was not significantly different from zero (b = −0.0037; P = 0.108).
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-010-0076-x/MediaObjects/10592_2010_76_Fig3_HTML.gif
Fig. 3

Structure analysis (Pritchard et al. 2000) of the British population. Plot of mean estimated ln likelihood (10 replicates) against inferred number of genetic clusters (populations)

Levels of heterozygosity of microsatellite loci (Table 2) were not significantly different between historical and modern samples (Wilcoxon signed rank test: Ho: Z = −1.57, P = 0.875; He: Z = −1.10, P = 0.311). Although allelic diversity was greater in the historical population, the difference was not significant (Z = −1.13, P = 0.257). However, allelic richness, which accounts for differences in sample sizes and was based on a minimal sample size of 41, was significantly higher in the historical than the modern samples (Z = −2.55, P = 0.011). We did not find strong evidence for heterozygosity excess in the modern population (Table 5). No test was significant when using the SMM or TPM.
Table 5

Genetic analysis of population bottlenecking in the contemporary sample of A. chrysaetos

Model

N

HE

Sign test

Wilcoxona

TPM

12

8

P = 0.340

P = 0.424

SMM

12

4

P = 0.065

P = 0.064

Tests were performed excluding the locus that deviated from Hardy–Weinberg genotypic proportions. Mutation models: TPM two-phased model, SMM stepwise mutation model. N number of loci used, HE number of loci showing excess heterozygosity

aProbability is two-tailed

Using the method of Beaumont (1999), we found a clear signal of a population decline, with no support for a growing or even stable population (Fig. 4). Under a linear model of population size change, the median of the posterior distribution was estimated to be log −1.76 (N0/N1 = 0.0174), with a 95% credible interval of log −4.56 to log −0.92 (<0.0001–0.1200). Under a model of exponential change, we obtained a point estimate of log −1.51 (N0/N1 = 0.031), with a 95% credible interval of log −2.48 to log −0.86 (0.0033–0.1380). These estimates correspond approximately to a 60-fold and 30-fold decline in effective population size, respectively.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-010-0076-x/MediaObjects/10592_2010_76_Fig4_HTML.gif
Fig. 4

Quantification of population size change using the method by Beaumont (1999). N0/N1 represents the ratio of present (N0) to past (N1) population size. The solid curve corresponds to the posterior distribution under a model of exponential population size change. The dashed curve was obtained under a model of linear change. We performed nine independent runs for each model. Posterior distributions were similar and pooled into one dataset. The prior distribution is shown for comparison (flat dotted line)

We tried to estimate past and current effective population sizes, as well as the time since the decline started using the approach of Storz and Beaumont (2002) (Fig. 5). The posterior gave a point estimate of log 2.64 (0.95 CI: log 1.88–log 3.37), or 437 individuals (0.95 CI: 76–2,344), for current effective population size, and of log 3.81 (0.95 CI: log 3.01–log 4.76), or 6,457 individuals (0.95 CI: 1,023–57,543), for the effective population size before the decline. There is some overlap between the two posterior distributions, but these results also provide evidence for a decline in effective population size some time in the past.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-010-0076-x/MediaObjects/10592_2010_76_Fig5_HTML.gif
Fig. 5

Ancestral (N1) and present (N0) population sizes. The posterior distributions were estimated using the method by Storz and Beaumont (2002). The priors are shown for N0 (dotted line) and for N1 (dashed line)

The posterior distributions were very different from the priors (Fig. 5). This difference further supports the idea that the data contain information about a population decline. The posterior distribution for T, the time at which the population started to decrease, had a median value of log 4.34, corresponding to 21,877 years. Similar posterior distributions were obtained when, in the priors, the median, mean and mode were either 100 or 100,000 years ago (Fig. 6). Given the 95% credible interval of log 3.03–log 5.41, 97.5% of the posterior distribution provides evidence for a major decline before 1000 years ago, excluding the possibility of a recent dramatic decline in effective population size.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-010-0076-x/MediaObjects/10592_2010_76_Fig6_HTML.gif
Fig. 6

Time since the decline in population size. The posterior distributions were estimated using the method by Storz and Beaumont (2002). Different priors are shown as dashed lines

Discussion

A number of studies have shown that populations that experience genetic bottlenecks can be negatively affected by reduced levels of fitness, adaptive diversity and population viability (Keller et al. 1994; Keller 1998; Saccheri et al. 1998; Madsen et al. 1999). These populations may require some degree of management intervention. Given the continued threat to the golden eagle in the British Isles, the main objective of the present study was to investigate how the population responded to the decline it experienced in the late nineteenth and early twentieth century and identify any potential genetic problems that it may be experiencing.

A comparison of the current and historical levels of genetic diversity of the species showed the extant population to have lost relatively little in the way of genetic diversity since the nineteenth century. Although the direct temporal comparison of allelic richness showed that some loss in genetic diversity has occurred, levels of heterozygosity between historical and contemporary samples were comparable. It also appears that the reduction in the effective size of the population was not so severe as to create a temporary excess of heterozygosity detectable under tests in the program BOTTLENECK.

The Bayesian programs did provide evidence for a severe decline in the effective population size of the golden eagle. While Beaumont’s (1999) method detected at least a 30-fold decline, Storz and Beaumont’s (2002) method found a reduction in Ne from 6,457 to 437 individuals, with the difference between pre- and post bottleneck population sizes comparable to those observed between modern mainland European and British breeding populations (approximately 5,000–6,000 pairs and 442 pairs, respectively; Watson 1997 and Eaton et al. 2007). However, the genetic bottleneck detected was an ancient one. It was estimated to have occurred not less than 1,000 years BP, and most likely around 22,000 years BP. The timing of the bottleneck event coincides with the Last Glacial Maximum (LGM) in Europe, which has been dated to approximately 22,000–18,000 years BP, prior to the deglaciation of the British Isles approximately 15,000 years BP (Wilson et al. 2002). Thus the genetic bottleneck observed in the British Isles may be the result of population fragmentation in the face of extreme climatic and ecological change, or the signature of a founding event following the retreat of the glacial sheet from the British Isles. Lower levels of microsatellite and mitochondrial DNA variation in the British population compared to southern European populations and the dominance of rare mainland European haplotypes in the British Isles are consistent with the existence of southern glacial refugia and a founder event in the British population (Bourke, unpublished data).

Loss of genetic diversity as a consequence of a bottleneck varies depending on population size reduction and bottleneck duration measured in generation time (Allendorf 1986; Frankham et al. 2002). The severity of the recent population size reduction of the golden eagle in the British Isles is difficult to determine accurately owing to the dearth of information in the historical record. However, there is some evidence to suggest a population nadir of approximately 100–200 breeding pairs in the first half of the twentieth century (Love 1982; Watson 1997; Watson and Whitfield 2002; Whitfield et al. 2008). Furthermore, although the most significant distributional decline in the British Isles occurred over a period of approximately one century, the generational time of this potential bottleneck translates to only approximately eight generations.

Our genetic analysis suggests that the recent bottleneck was not severe enough to lead to a substantial reduction in the genetic diversity of the golden eagle in the British Isles. Furthermore, the longevity of the golden eagle probably further reduced the effect of genetic drift over this period and acted as a buffer against the rapid loss in genetic diversity. Hailer et al. (2006) drew similar conclusions in their analysis of Scandinavian white tailed sea eagle (Haliaeetus albicilla) populations with a history of demographic decline. Bearing in mind that losses in neutral genetic diversity may not always be correlated with losses in fitness and adaptive diversity (Hedrick 2001; McKay and Latta 2002; Moss et al. 2003), and despite suffering ancient loss in genetic diversity, the British population should retain sufficient fitness and adaptive diversity to ensure long-term persistence.

The British and historical Irish populations of golden eagles have long been considered continuous on the basis of similar ecology and geographic proximity. However, the breeding distribution of the species is fragmented by way of island occupation along the western seaboard. The Bayesian assignment method did not find any evidence for population genetic structure, which is indicative of one single and coherent population. Similar results were obtained when historical Irish individuals were included in the analysis, indicating no population structure was present across the British Isles.

The absence of genetic structure in the British Isles has management implications. A recent study by Whitfield et al. (2006) developed a population model which showed that, while breeding pairs were increasing in some regions (Natural Heritage Zones), most regions showed significant signs of decline. However, the study pointed out that these results were based on analysis of demographic data from each region and did not take account of dispersal between regions, which might compensate for local declines. Our study provides evidence for random mating across the British Isles population and suggests that population models which do not take account of such dispersal are likely to bias estimates of regional declines.

Previous field studies of natal dispersal in the species indicated that, while the golden eagle ranges over large areas during the sub-adult period, there is a tendency for individuals to return to the natal area on reaching breeding age (Steenhof et al. 1984; Haller 1994). However, the absence of any fine-scale genetic structure in this study is consistent with a population dominated by random mating and does not suggest that there is a strong propensity for individuals to return to their natal areas for breeding. Breeding territory fidelity has been reported to be high in this species. A study in the United States found that 12 out of 14 individuals relocated at least 400 km from their breeding territories returned to their territories within 1 year of release (Phillips et al. 1991). Cases of territorial fidelity for periods of >20 years have also been reported (Katz 2004). Given the high level of territorial fidelity observed, it may be that individuals are able to find and occupy vacancies in breeding territories during the early dispersive period and thus negate the need to contest a breeding space around the natal area.

Our results confirm that individuals from the British population are suitable for the Irish reintroduction efforts. However, we do not find evidence either for population genetic structure, or for fidelity to natal breeding areas. These observations raise the questions as to why the birds have, as yet, not recolonised Ireland naturally. It may thus not be sufficient to reintroduce individuals, without addressing the underlying ecological problems that contributed to the species’ extinction from Ireland in the first place. Factors limiting population expansion in the British Isles are over-grazing, afforestation (Whitfield et al. 2001), and most notably, despite the introduction of legal protection, persecution (Whitfield et al. 2003). The main short-term objective of conservation measures should be to increase population sizes by continuous safeguarding of individuals and habitat management.

Acknowledgements

We thank Scottish Natural Heritage and the Royal Society for the Protection of Birds for putting their support behind the contemporary sampling effort. We thank Stuart Benn, Roger Broad, Kenneth Canton, Ian Dillon, Desmond Dugan, Kenny Graham, JR Grant, Mike Gregory, Roger Hayward, Kenny Kortland, Ian MacPherson, Liz McDonald, Lorcan O’Toole, Robin Reid, Chris Rollie, Logan Steele, Bob Swann, Tom Talbot, Des Thompson, Dave Walker, Phil Whitfield, Andrew Wight, Richard Wood, the Centre for Ecology and Hydrology, the National Museums of Scotland and the Scottish Raptor Study Groups for providing contemporary DNA, blood and feather samples, and Ian Dillon and Mark Eaton for providing additional geographical information on samples. We thank Noble Henning, Barbara Kerr, the American Museum of Natural History, Cambridge Museum of Zoology, Glenveagh National Park, Hancock Museum, Kelvingrove Museum, Leicester Museum, Liverpool Museums, Museum of Comparative Zoology Harvard, National Museums of Scotland, Norfolk Museum, National Museum of Ireland, National Museum of Wales, Natural History Museum, Oxford University Museum of Natural History, Townley Museum and University College Cork for kindly providing historical samples. We thank Michele Clarke for acting as a research grant PI. This work was performed at the Sheffield Molecular Genetics Facility and funded by the Natural Environment Research Council, UK, and the University of Nottingham.

Copyright information

© Springer Science+Business Media B.V. 2010