SEARCH

SEARCH BY CITATION

Keywords:

  • BAPS;
  • Geneland;
  • GIS;
  • landscape genetics;
  • Lutra lutra;
  • otter;
  • population structure

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

1. After a major decline, the UK otter Lutra lutra population is now recovering in its known strongholds (northern England, Wales and Borders and southwest England) and also in central England where the population had become small, fragmented and was reinforced with captive bred individuals. Bayesian clustering and GIS are used here to identify the genetic structure of the UK otter population and to assess expansion from strongholds and the contribution of introduced otters. Large carnivores have recently started to recolonize landscapes where they were formerly absent, especially in developed countries and understanding the expansion of these populations is essential for informing conservation management, linking fragmented populations and re-establishing gene flow.

2. Three Bayesian clustering techniques were used (structure, geneland spatial and baps4 spatial) to estimate the number of otter populations (K). A novel progressive partitioning approach was tested to identify genetic substructuring at various hierarchical levels using successive partitions at K = 2.

3. Four regional populations were identified that reflect known population history. Isolated populations in southwest England and in Wales and its borders showed the lowest levels of genetic diversity. Higher diversity and private alleles in northern and central England reflect the proximity to genetically diverse Scottish populations and the positive effect of reintroductions.

4. Progressive partitioning was used to produce a more detailed analysis, by allowing comparison and combination of clusters identified by different techniques and by avoiding the subjective estimation and choice of K.

5.Synthesis and applications.. Although the otter population is increasing, our data show little sign of population expansion from the stronghold regions into central England, instead reflecting the success of population reinforcement in this area. Our progressive partitioning approach allows the identification of fine-scale substructure (11 groups) that enables the prioritization of management effort including identifying barriers to dispersal within and between populations and monitoring of introduced individuals.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Wild animal populations are under increasing anthropogenic pressure, leading to fragmentation, isolation and therefore a reduction in gene flow. Gene flow is usually considered necessary for the viability of small isolated populations (Mills & Allendorf 1996).

Conservation management aims to preserve evolutionary lineages and adaptive diversity across the geographic range of a species (Storfer 1996) by preserving the natural network of connections and therefore gene flow between populations (Crandall et al. 2000). It is therefore important to identify population units within a species range and the degree of gene flow between them. Fain, Straughan & Taylor (2010) examined genetic variation and connectivity in a recovering North American wolf population, supporting its assignment to the species of wolf Canis lycaon which supports its protected status and will influence future management decisions for the recovery of the species. Defining populations and their geographic boundaries can, however, be difficult, especially when substructuring occurs in continuously distributed populations where the grouping of individuals into geographically determined units reduces the ability to accurately describe biological reality.

To address this problem, recent methods aim to cluster individuals into groups defined on genetic criteria alone. Bayesian algorithms use individual multilocus genotypes derived from multiple genetic markers to assign individuals to clusters, on the assumption that markers are in Hardy Weinberg and linkage equilibrium within each randomly mating subpopulation (Manel, Gaggiotti & Waples 2005). However, authors such as Frantz et al. (2009) suggest caution when using these techniques as the clustering solution may be an artefact of incomplete sampling or isolation-by-distance, rather than an accurate reflection of population genetic structure.

In this study, we compare and combine several clustering programmes and although this approach is increasingly taken in such studies (Carmichael et al. 2007; Coulon et al. 2008), the number of populations (K) estimated frequently varies with technique, rendering comparison and combination of results problematic. In such scenarios, the results that best fit the known population history of the species are typically chosen.

Here, we also use a novel approach: progressive partitioning of clusters derived using Bayesian clustering. This method has been used in other areas of cluster analysis (Dubes & Jain 1976) and resembles a hierarchical approach used by Coulon et al. (2008). This method operates on the assumption that hierarchical clustering algorithms produce a nested series of partitions (Jain, Murty & Flynn 1999). Clusters identified in the first round of analysis are expected to be the most differentiated, with subsequent clusters having progressively lower levels of genetic differentiation. The progressive partitioning potentially allows for the identification of otherwise cryptic populations when looking for clusters at lower levels of genetic differentiation, although it has the disadvantage of explicitly ignoring the likelihood values for the different values of K.

Progressive partitions produced by Bayesian techniques can be displayed using Geographic Information System (GIS) and compared to identify consistent patterns of substructuring at different degrees of differentiation. Clusters that find agreement between techniques can be used to build a picture of population structure and combined to test further genetic summary statistics.

During the 20th century, the Eurasian otter Lutra lutra (Linnaeus, 1758) declined significantly throughout its range, leading to local extinctions in many areas (Conroy & Chanin 2000). The decline was related to habitat destruction, direct persecution and the bioaccumulation of pollutants (Conroy & Chanin 2000). In the UK, otter populations declined significantly during the 1950–60s (Coxon et al. 1999; Conroy & Chanin 2000), and by the mid 1970s, the mainland British population was largely confined to parts of Scotland, west Wales and the southwest of England (Jones & Jones 2004). Detailed monitoring has shown that since the late 1970s, there has been a gradual expansion of the otter populations in UK strongholds and from virtual extinction in most of England (Crawford 2010). Expansion may be the result of re-colonization from the west (southwest England and the Welsh borders) and from the north (Scotland), or the result of controlled releases made following recommendation by the Nature Conservancy Council (now Natural England) after the first otter survey of England (Lenton, Chanin and Jefferies 1980).

The main objectives of this study were to identify the population structure of the UK otter at differing degrees of genetic differentiation using Bayesian Clustering Techniques and a novel progressive partitioning approach. The goal was the detection of fine-scale population structure to allow the identification of gene flow between populations, of recolonization events from otter strongholds and to assess the success of otters reintroduced 20–30 years ago.

We expected to find three populations based around the otter strongholds (Wales and borders, southwest of England and Scottish borders), as reported by Dallas et al. (2002). However, translocation and reinforcement events in East Anglia and North Yorkshire could also have resulted in new genetically distinct populations becoming established. Based on population history, we therefore predicted that there would be five distinct otter populations in the UK, with the possibility of further demographic substructuring reflecting anthropogenic fragmentation.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Sample Set and Molecular Analysis

With an increasing otter population in Britain, mortality because of road traffic accidents (RTAs) has escalated during the last 15–20 years and has become one of the most important causes of death of otters in most European countries (Philcox, Grogan & Macdonald 1999). During this time, the UK Environment Agency and other regional organizations have collected otter RTA casualties throughout England and Wales and recorded their location. This provides an archive of geo-referenced samples from a wild population, which can be utilized in genetic analyses.

Postmortem examinations of otter road casualties were conducted by Cardiff University Otter Project (1992–2008), or by the Wildlife Veterinary Investigation Centre (southern England only, 1990–2007). Muscle samples were retained and stored in ethanol at −20 °C. DNA was extracted using the QIAGEN DNeasy tissue kit (QIAGEN, #65906) for 566 samples. The samples were genotyped for 15 microsatellite loci, using primers by Dallas & Piertney (1998) and Huang et al. (2005) and the methods described in Hobbs et al. (2006). PCR products were analysed using an abi prism®3100/3130 Genetic Analysers (Applied Biosystems, Foster City, CA, USA), and analysis was performed using Genescan v3.7, Genotyper v3.6 and GenemapperTM (Applied Biosystems).

Analysis

The multilocus genotypes for 566 individuals were analysed using the Bayesian clustering programmes structure (Pritchard, Stephens & Donnelly 2000; Falush, Stephens & Pritchard 2003; Pritchard & Wen 2003), baps4 spatial (Corander & Marttinen (2006) and geneland spatial (Guillot, Mortier & Estoup 2005). Bayesian clustering algorithms use individual multilocus genotypes derived from multiple microsatellite markers to assign individuals to clusters, on the assumption that markers are in Hardy Weinberg and linkage equilibrium within each randomly mating subpopulation (Pearse & Crandall 2004; Manel, Gaggiotti & Waples 2005). While these programmes identify population structure using the same principles, they can differ in their clustering results and even differ between runs of the same programme. Default parameters for each programme were used with specific parameters shown in Table 1.

Table 1.   Parameters used for the Bayesian clustering techniques
PackageAlgorithmIterationsModelKEstimation of K
structureDescribed by Pritchard, Stephens & Donnelly (2000); Falush, Stephens & Pritchard (2003); Pritchard and Wen (2003)10 00 000 iterations, using CONDOR (CONDOR is a specialized workload management system for compute-intensive jobs). Burn-in 100 000Admixture, model, assuming correlated allele frequenciesSet from 1 to 11, with 5 independent runs of eachTwo approaches used: (i) the highest estimated log probability of data Pr(X|K) estimates the most likely number of clusters (Pritchard, Stephens & Donnelly 2000). (ii) Evanno, Regnaut & Goudet (2005) use the second order rate of change of the likelihood function with respect to K
baps4 spatialDescribed in Corander & Marttinen (2006). Clustering at the individual level, using the spatial model Admixture analysis: Minimum number of individuals in a cluster = 1. Default values were used for admixture priors. 100 iterations to estimate the admixture coefficients, 200 reference individuals from each population and 20 iterations to estimate the admixture coefficients for the reference individualsVector of values for the maximum number of clusters (K) with five replicates of K = 5, 10 and 15After all the K values were processed, the stored results were merged based on the logML values with the best 10 partitions displayed The K value with the highest likelihood was chosen.
geneland spatialGuillot, Mortier & Estoup (2005)500 000 MCMC iterations to identify K 200 000 iterations once K was identifie Posterior probabilities of population membership for each individual were calculated using burn-in of 1000The Dirichlet (λ) distribution was used following Guillot, Mortier & Estoup (2005) using spatial data (Spatial D-model) The amount of uncertainty to spatial coordinates was set at 0·3 (author recommendation), maximum rate of Poisson process 566 (number of individuals); maximum number of nuclei in the Poisson-Voronoi tessellation 1698 (3 × the number of individuals).Priors on K-uniform between 1 and 11The most probable number of clusters (K) was found using five replicates and comparing the histograms

Two methods were used to estimate the most likely number of populations (hereafter K). First, the three Bayesian clustering techniques were used to estimate K, and assignments were mapped for comparison using ArcMap v9.2 (ESRI 2007) (shape files provided by the Environment Agency). Individuals were categorized within given populations if they had >50% probability of assignment to that population.

The Bayesian Clustering Algorithms can identify further substructure by re-analysing the identified clusters following an approach used in Pires et al. (2009) and Bray et al. (2009). Here, different estimates of K were found when using different Bayesian Clustering techniques. To enable comparison between these techniques, we standardized the K value from which further substructuring was investigated. We used a novel progressive partitioning approach restricting K to two, extrapolating two clusters at each subdivision of the populations for each of the clustering techniques. For each partition, individuals were assigned to a population if they had >50% assignment to that population, until the clusters no longer split (this occurs when either all individuals are assigned to one population or when all individuals showed c. 50% assignment to each of the two populations). For each partition, five replicate runs (all of K = 2) were performed and compared. Where at least three of the five runs were consistent, the assignment of one of these runs was used for the subsequent partitions.

Genetic Diversity

Genetic diversity and structure was assessed using standard summary statistics. We focused on two levels of population genetic structure: regions and subregions (identified by visually comparing maps of the partitions created from both optimal K and progressive partitioning methods). Individuals were included in regional analyses where assignment to a region was >0·9, with agreement between all three techniques using the progressive partitioning method. Individuals were included in subregion analysis where assignment was >0·5 to that subregion, with agreement between at least two of the three techniques using the progressive partitioning method. At the subregional level, a greater degree of admixture is expected, so the assignment threshold was reduced accordingly. We estimated observed (Ho) and expected (He) heterozygosity, mean number of alleles (A) and number of private alleles per population (Au) using gda (v1.1; Lewis & Zaykin 2001). Allelic richness (AR) was derived using fstat (v2.9.3; Goudet 1995) and adjusted for variation in subpopulation sample size. This programme was also used to estimate FIS in populations separately and overall. arlequin 3.1 was used to derive pairwise FST values among populations (following the method of Weir & Hill (2002)), and statistical significance was tested with 10 000 permutations as implemented in arlequin 3.1 (Excoffier, Estoup & Cornuet 2005) with Bonferroni correction. Hardy–Weinberg equilibrium (HWE) and linkage disequilibrium (LD) were tested using genepop 3.3 (Raymond & Rousett 1995) with all probability tests based on the Markov chain method (Guo & Thompson 1992) using 1000 de-memorization steps, 100 batches and 1000 iterations per batch.

Population History

The occurrence of recent demographic bottlenecks was assessed using bottleneck 1.2 (Cornuet & Luikart 1996) using the infinite allele model (IAM), stepwise mutation model (SMM) and two-phase model of mutation (TPM, with 95% SMMs); significance was assessed using the Wilcoxon test. The software 2mod (Ciofi et al. 1999) estimates the relative likelihood of a model of immigration-drift equilibrium, vs. drift since a certain time. The model of immigration-drift is assumed to be either an infinite island or continent-island model of gene flow, which both give rise to the same likelihood function (Rannala & Hartigan 1996). The calculation of the likelihoods for the pure drift case is as described by O’Ryan et al. (1998) and implemented in the program dlik1.1. The programme estimates the relative likelihoods of the two models using an MCMC procedure as described in Ciofi et al. (1999) simulating the extent of the interaction between drift and gene flow using the parameter F (the probability that two alleles share a common ancestor within a population) (Dhuyvetter, Gaublomme & Desender 2005).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Definition of Populations Based on Bayesian Clustering Techniques

The three clustering techniques differed in their estimate of the most likely ‘optimum’ number of partitions (K). structure showed optimum partitions at K = 9 (−19845·5) (for estimates of optimum K, see Appendix S1 in Supporting Information, Fig. S1a), but using an alternative method developed by Evanno, Regnaut & Goudet (2005) which uses the largest rate of change between partitions (ΔK) showed K = 4 (−36·4031), with less likely alternatives of K = 6 and K = 9 (see Appendix S1, Fig. S1b). geneland spatial gave an optimum at K = 6 (see Appendix S1, Fig. S2) with one admixed individual (from the Shetland Isles, see Fig. 1) being comprised of two private genetic groups and accounting for two of these populations. For baps4 spatial, the most likely number of clusters based on logML values varied between runs, from K = 8 to 10, with the highest at K = 9. The partitions identified by each of the Bayesian clustering algorithms were mapped for comparison using ArcMap v9.2 (ESRI 2007). Figure 1 displays the population assignments of individuals to the optimum number of populations derived from geneland spatial; population assignments from the other Bayesian clustering techniques are described below.

image

Figure 1.  Population assignment by geneland spatial, optimum partitions K = 6.

Download figure to PowerPoint

geneland spatial identified four main clusters, referred to hereafter as the ‘Wales and Borders’ (Fig. 1a), ‘North England’ (Fig. 1b), ‘Southwest England’ (Fig. 1c) and ‘Central England’ (Fig. 1d) clusters. baps4 spatial and geneland spatial identify a very similar Wales and Borders cluster; this same cluster was divided into two by structure at both K = 4 & K = 6. geneland spatial and structure (K = 4 and K = 6) identified a similar Southwest England cluster, but this was further divided by baps4 spatial into two. The remainder of the individuals were also assigned to North England and Central England clusters by baps4 spatial and structureK = 6; however, these clusters were grouped together by structureK = 4. The North England cluster was further divided by baps4 spatial, grouping individuals from Ireland and part of Yorkshire.

Outside the four main clusters identified by geneland spatial, there were other notable results such as the assignment of the individual from Shetland into its own cluster by baps4 spatial with geneland spatial assigning it 50 : 50 to 2 additional clusters (Fig. 1e,f). Many individuals were not assigned by structure (i.e. assignment <50%) at K = 6 and 4, and assignment of these individuals was inconsistent between structure runs.

Progressive Partitioning Using Bayesian Clustering

Progressive partitioning was conducted using each clustering technique, examining the results of K = 2 at each stage. Figures 2–4 show the resulting groupings.

image

Figure 2.  Clusters identified by the progressive partitioning approach (sequential steps of K = 2) using structure. Clusters with the thick black boxed outline represent the regions. Clusters with a thin black boxed outline are clusters chosen to represent subregions.

Download figure to PowerPoint

image

Figure 3.  Clusters identified by the progressive partitioning approach (sequential steps of K = 2) using geneland spatial. Clusters with the thick black boxed outline represent the regions. Clusters with a thin black boxed outline are clusters chosen to represent subregions.

Download figure to PowerPoint

image

Figure 4.  Clusters identified by the progressive partitioning approach (sequential steps of K = 2) using baps4 spatial. Clusters with the thick black boxed outline represent the regions. Clusters with a thin black boxed outline are clusters chosen to represent subregions. The dashed arrow (inline image) indicates steps not shown, leading to the 10 final partitions.

Download figure to PowerPoint

Comparison of Population Assignments for Progressive Partitioning Approach

Many clusters were consistent between techniques, and the use of progressive partitioning provided a method of identifying clusters at different degrees of genetic differentiation.

At a higher level of genetic differentiation, there were four main clusters identified by all techniques (thick boxed clusters Figures 2–4), referred to here as ‘regions’: Wales and Borders [1], Southwest of England [2], Northern England (including the Irish samples) [3] and Central England [4]. Subdivisions within regions varied somewhat between methods. Wales and Borders [1] was divided into four overlapping clusters by structure, four spatially distinct clusters by geneland spatial and three clusters by baps4 spatial. Southwest England [2] was divided into three clusters by structure and baps4 spatial and two clusters by geneland spatial. North England [3] was divided into four clusters by structure and into three clusters by geneland spatial and baps4 spatial. The most disparate result was for Central England [4], which was divided into two clusters by structure, five clusters by geneland spatial and 13 clusters by baps4 spatial.

Clustering solutions given by the different methods (optimal K and progressive partitioning) were compared. Subregions were accepted as robust clusters if they were inferred by more than one technique and especially if there was agreement between all techniques. Chosen robust clusters deriving from the four regions are shown in Fig. 5. In the Wales and Borders region [1], a Southwest Wales subregion (1a), a Northwest Wales subregion (1b) and a Mid-Eastern Wales subregion (1c) were identified by baps4 spatial and geneland spatial using the progressive partitioning method and by structure (K = 6). In Southwest England [2], a subregion on the tip of the Southwest Peninsula (2a) was overlapped by a larger subregion (2b) identified by all techniques using the progressive partitioning method.

image

Figure 5.  11 otter subregions identified by consensus between optimal K and progressive partitioning methods applied using three Bayesian clustering techniques.

Download figure to PowerPoint

For North England [3], two subregions were identified; a North England/Southern Scotland subregion (3a) and a subregion which clustered the Irish samples with samples in North Yorkshire (3b). These subregions were identified by baps4 spatial for both the optimal K method and progressive partitioning method. The structure progressive partitioning method identified these subregions but also found evidence for further substructure within 3b. geneland spatial progressive partitioning identified similar subregions but partitioned the Irish samples at an earlier stage; this was not categorized as a separate subregion because of lack of agreement between Bayesian techniques.

There was much complexity in the subdivisions for Central England [4]. There was strong agreement between all Bayesian techniques for an East Anglia subregion (4a), and an Oxfordshire subregion (4b), although the latter differed in its spatial limits between methods. An additional West Country subregion (4c) was identified by geneland spatial and situated at the western side of Central England [4] and adjacent to Southwest England [2]. Subregion (4c) was not identified by baps4 spatial; however, it was identified by the structure progressive partitioning method via the partitioning of Southwest England [2] rather than from Central England [4].

Other clusters either represented few (1–5) individuals or were found by only one method, so were not defined. One exception was made, for the individual from the Shetland Isles (5a) identified by both geneland spatial and baps4 spatial in both the progressive partitioning and optimal K methods. It is thought likely that this individual does represent a subpopulation, but insufficient sampling in this area precludes confirmation.

Summary Statistics

Summary statistics were estimated for the regions and subregions identified above. Selection criteria for regional analysis (individual assignment >0·9, agreement between all techniques, using the progressive partitioning method) were met by 454 individuals (out of 566, 80·2%). In total, 266 individuals were assigned to the Wales and Borders region, 50 individuals to the Southwest England region, 67 to North England region and 71 individuals to the Central England region. Selection criteria for subregion analysis (assignment >0·5, agreement by at least two of the techniques using the progressive partitioning method) were met by 332 individuals from 566 genotypes (58·7%) (Fig. 5) and used for further analysis.

Population Genetic Diversity: Regions

Allelic diversity ranged from 3·73 alleles per locus for Southwest England to 5·6 in North England (Table 2). Each region showed private alleles ranging from one in Southwest England to 17 in Central England (Table 2). Values of observed (HO) and expected (HE) heterozygosity were in the range of 0·46–0·65 and 0·49–0·70, respectively (Table 2). Otters from Southwest England and Wales and Borders showed the lowest levels of genetic diversity, while individuals from North England and Central England showed the highest.

Table 2.   Average summary genetic statistics for each region, over 15 loci
LocusNAHeHoFISArAu2mod F values
  1. Population has a significant deviation from Hardy–Weinberg equilibrium (*< 0·05, **< 0·01,***< 0·001).

  2. N, sample size; A, mean number of alleles per locus; Ho, observed heterozygosity; He, expected heterozygosity; FIS, inbreeding coefficient Ar, allelelic richness; Au, number of private alleles.

[1] Wales and Borders2664·40·520·490·06***3·9020·35
[2] Southwest England503·730·490·460·08**3·7310·38
[3] North England675·60·700·650·07**5·4490·01
[4] Central England715·530·680·640·05*5·37170·16

All regions displayed significant departure from Hardy–Weinberg expectations (FIS = 0·053–0·075; Table 2), and the fact that allele frequencies featured greater homozygosity than expected suggests a Wahlund effect. Further, all regions were found to be highly significantly differentiated from each other based on FST values (Table 3), with North England/Central England comparison having the lowest (FST 0·10) and Wales and Borders/Southwest England comparison showing the greatest differentiation (FST 0·28).

Table 3.   Pairwise (FST) values between regions identified by Bayesian clustering
 [1] Wales and Borders[2] Southwest England[3] North England[4] Central England
  1. *FST values highly significant, < 0·001.

[1] Wales and Borders0   
[2] Southwest England0·28*0  
[3] North England0·19*0·20*0 
[4] Central England0·22*0·23*0·10*0

No consistent evidence was found for a genetic bottleneck in any regional population with the results dependent on the mutation model used. As is commonly observed, the IAM found significant (< 0·05) heterozygote excess for all populations; however, no significant results were found for any population under either the TPM or SMM (given the inconsistent nature of these results, no further exploration of demographic bottlenecks using Bayesian Coalescent simulation was carried out). For the 2mod analysis, North England [3] and Central England [4] yielded low F value modes (0·09, 0·16), which indicate that the probability of genes being identical by descent is low and was more likely to be influenced by gene flow. In contrast, Wales and Borders [1] and Southwest England [2] yielded much higher F values (0·35, 0·38, respectively) indicating that these populations have been predominantly influenced by drift.

Population Genetic Diversity: Subregions

Allelic diversity ranged from 2·93 alleles per locus for subregion (2a) to 5·73 in subregion (3b) (Table 4). Some subregions possessed no private alleles, while subregion (3b) showed five private alleles, representing a third of all detected. Subregions (4a,b) derived from the Central England region [4] possessed a further five private alleles (Table 4). Values of observed heterozygosity (HO) and expected Heterozygosity (HE) ranged between 0·45–0·7 and 0·44–0·72, respectively. Levels of genetic diversity in the subregions reflected the levels of genetic diversity in the regions from which they were derived. Surprisingly, subregions (3b and 4a) displayed significant heterozygote deficiencies as compared to Hardy–Weinberg expectations. This departure from Hardy–Weinberg expectations is evident despite the fact that these subregions show the highest levels of genetic diversity, suggesting further substructuring within these two subregions. All subregions were significantly differentiated from each other (Table 5), and as expected, FST values between subregions derived from the same region were lower than between subregions derived from different regions.

Table 4.   Average summary genetic statistics for all 11 subregions, over 15 loci
PopulationNAHeHoArFisAu2mod F values
  1. Population has a significant deviation from Hardy–Weinberg equilibrium (*< 0·05, **< 0·01,***< 0·001).

  2. N, sample size; A, mean number of alleles per locus; Ho, observed heterozygosity; He, expected heterozygosity; FIS, inbreeding coefficient; Ar, allelelic richness; Au, number of private alleles.

(1a) Southwest Wales624·070·510·503·190·02310·23
(1b) Northwest Wales83·070·510·533·07−0·02700·32
(1c) Mid-Eastern Wales973·80·470·472·840·00100·36
(2a) Southwest Peninsula132·930·440·452·71−0·02210·39
(2b) Southwest England323·330·460·452·630·02810·36
(3a) North of England/Scottish Borders264·80·680·654·010·04520·09
(3b) Irish & North Yorkshire345·730·720·74·610·022*50·06
(4a) East Anglia394·530·610·593·670·034*20·30
(4b)Oxfordshire215·330·670·74·39−0·03830·08
(4c) West Country93·730·580·583·64−0·00200·16
(5a) Shetland11·270·270·271·2700
Mean (*not including results from 5a)34·14·1320·5650·5623·4760·0011·5 
Table 5.   Pairwise multi-loci (FST) values for 11 subregions identified by Bayesian clustering
  1. *Significant P-value <0·05, **significant P-value <0·05, ***highly significant P-value<0·001.

Subregion1a1b1c2a2b3a3b4a4b4c5a
 1a0          
 1b0·09***0         
 1c0·12***0·10***0        
 2a0·32***0·36***0·36***0       
 2b0·30***0·32***0·32***0·17***0      
 3a0·20***0·25***0·16***0·26***0·23***0     
 3b0·17***0·22***0·14***0·20***0·21***0·06***0    
 4a0·24***0·30***0·22***0·30***0·29***0·12***0·14**0   
 4b0·22***0·29***0·20***0·26***0·26***0·09***0·11**0·12***0  
 4c0·26***0·29***0·24***0·18***0·05***0·13***0·14**0·20***0·16***0 
 5a0·430·480·410·480·440·210·260·32*0·21*0·310

Bottleneck tests for subregions gave equivocal results and depended on the mutation models used. The IAM model found significant heterozygote excess in seven of 10 subregions (subregions 1a, 1c, 2b, 3a, 3b, 4a & 4b). Under the TPM model, significant heterozygote excess was found for subregion 3a only. No significant results were found for the SMM mutation model.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Bayesian clustering indicated that the sampled otters could be divided into four regional populations, recapitulating Dallas et al. (2002). These populations reflect known demographic history, with strongholds of the North England, Southwest England and Wales and Borders (Crawford et al. 1979; Lenton, Chanin and Jefferies 1980; Jones & Jones 2004), as well as Central England (Jessop & Cheyne 1992) where Otter Trust (OT) releases were made. Regional populations showed significant departure from HWE and were further subdivided into ten subregions. Two of these subregions showed significant departure from HWE, both in areas where populations had been reinforced with rehabilitated (3b) or captive bred (4a) otters.

Regions were significantly differentiated with low genetic similarity (measured by FST), as found by Dallas et al. (2002) and by Stanton et al. (2009) who found substantial differences in mitochondrial DNA frequencies among areas represented by the regions within the UK.

Southwest England [2] showed the greatest evidence of isolation, followed by Wales and Borders [1]. Despite being geographically adjacent, both FST and 2mod results imply no gene flow between them. The inference that these populations are demographically isolated is supported by the lowest levels of genetic diversity, and despite the lack of convincing evidence for a demographic bottleneck, they showed evidence of being affected by genetic drift (2mod). While appearing continuous, substructure was found in both regions, including a southwest peninsula population (subregion 2a) also identified by Dallas et al. (2002). Substructure in an apparently continuous population was also identified in the North American river otter Lutra canadensis by Latch et al. (2008). Further analysis should be undertaken to identify the origin of this substructuring, for example, exploring correlations with landscape features and identifying historical and contemporary factors affecting gene flow.

North England [3] and Central England [4] regions possessed higher genetic diversity, with the results from 2mod indicating that both populations are more likely to be under the influence of gene flow rather than drift. It is possible that there has been, and continues to be, immigration into North England from populations in Scotland, which are reported to have greater genetic diversity than southern UK populations (Dallas et al. 2002). North England also contains areas where rehabilitated otters were released by the Vincent Wildlife Trust (VWT). It is notable that one of the subregions identified within North England, groups samples from Ireland with those from North Yorkshire (3b). This could be evidence of the success of VWTs rehabilitation programme, which released otters into North Yorkshire from a number of source locations including several from Northern Ireland (R. Green, VWT, Pers. comm). Central England [4] also showed high genetic diversity, and 2mod results imply substantial immigration into this population; this may be attributable to the reinforcement campaign by the OT.

The Central England population [4] was concentrated in East Anglia during the 1960s and 1970s and was thought by many to be small, fragmented and inviable (Crawford et al. 1979; Lenton, Chanin and Jefferies 1980), but subsequent surveys (Strachan & Jefferies 1996; Crawford 2003) have shown that this population has expanded considerably. Because radio-tracking studies have identified large home range sizes (38·8 ± 23·4 km Green, Green & Jefferies 1984), there was an expectation that otters would disperse into Central England from strongholds in the west (southwest England and the Welsh borders) and from the north (Scotland) (Coxon et al. 1999; Conroy & Chanin 2000). However, the genetic data do not support this, with no evidence of contributions from adjacent populations it seems probable that the increase in numbers in Central England results primarily from reinforcement with captive bred otters or natural expansion of the small indigenous otter population.

Central England is probably the most anthropogenically influenced population, having suffered the effects of persecution and pollution, and by the early 1980s, its survival seemed doubtful (Strachan & Jefferies 1996). The OT released 117 individuals between 1983 and 1999, from captive bred stock. Unfortunately, details of some of the source populations remain undisclosed. Given the demographic history, a population bottleneck might be expected but is not supported by the genetic data, which shows high levels of genetic diversity compared with otter strongholds. The 2mod results indicate that it is likely that there has been gene flow into the population. The high genetic diversity and number of private alleles (17 private alleles compared to nine in North England, two in Wales and Borders and one in the Southwest England regions) suggest that the founding stock was supported by individuals not sourced from surrounding populations. The phenomenon of increased genetic variation from a small population that has received introductions reflects a study by Kolbe et al. (2004) who found that an area that had received additional introductions of Cuban Lizard Anolis sagrei from multiple source locations showed much higher levels of genetic diversity than an area that received introductions from a single source. Analysis of the breeding stock and captive breeding lineages of the reintroduced otters would add greatly to the understanding of this population.

Progressive partitioning used in combination with the most likely estimate of K and visual comparison using GIS potentially allows for a better understanding of population clusters identified. Using the assumption that at lower values of K, the genetic differentiation between population groups would be greatest, the approach identified population structure at differing levels of genetic differentiation. It also has the advantage that the models were able to produce clustering solutions that were similar enough that they could be compared and combined, allowing identification of regional and subregional clusters not possible to deduce from traditional estimates of K alone.

There were, however, some differences in the clusters identified and their assignment of individuals, for example structure could not identify individuals from un-sourced populations, and baps4 spatial appeared to further subdivide some clusters based on their private allele frequencies. At lower levels of genetic differentiation, there was less agreement between the models with (58·7% of 566 individuals with agreed assignment to the subregions compared to 80·6% for the regional clusters).

Conservation Implications for the UK Otter Population

This study supports the conclusion of Dallas et al. (2002) that there is limited or no gene flow between regional populations. The expectation that natural dispersal from strongholds has contributed to the population expansion in central England as concluded in a recent otter survey of England (Crawford 2010) is therefore unlikely. This finding emphasizes the importance of identifying cryptic genetic structure when monitoring populations.

Within regions, evidence suggests dispersal is limited despite the highly mobile nature of this species. Dispersal may be limited by a number of factors, which may be extrinsic (e.g. environmental barriers, unsuitable habitat) or intrinsic (e.g. reluctance to disperse into areas with an unfamiliar prey base).

Management of the otter population should consider regions as management units and focus on ensuring gene flow between subregions by understanding what barriers (landscape, anthropogenic features) are creating the population structure within each of the regions. Ideally, gene flow should be re-established between the regions; however, with doubt over the origins of some reintroductions, genotypes from additional Scottish and continental European populations should first be analysed to help identify their source. If introductions are found to have been from multiple non-native origins into the UK Central England otter population, then the unique opportunity arises that allows the comparison of the spread, adaptive change and interaction between populations that have received reintroductions from non-native sources, with those that have received reintroductions from surrounding populations, and those that have not received introductions. This would provide valuable information on the progression and consequences of invasive species introduction; a growing global problem.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The authors would like to thank all agencies involved in collecting otter carcasses, Vic Simpson for otter samples from southwest of England, Rosemary Green for information provided on VWT reintroductions and the Environment Agency for their funding of the Cardiff University Otter Project. The editorial team and two anonymous referees for suggesting improvements.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Bray, T.C., Chikhi, L., Sheppy, A.J. & Bruford, M.W. (2009) The population genetic effects of ancestry and admixture in a subdivided cattle breed. Animal Genetics, 40, 393400.
  • Carmichael, L.E., Krizan, J., Nagy, J.A., Fuglei, E., Dumond, M., Johnson, D., Veitch, A., Berteaux, D. & Strobeck, C. (2007) Historical and ecological determinants of genetic structure in arctic canids. Molecular Ecology, 16, 34663483.
  • Ciofi, C., Beaumont, M.A., Swingland, I.R. & Bruford, M.W. (1999) Genetic divergence and units for conservation in the Komodo dragon Varanus komodoensis. Proceedings of the Royal Society B: Biological Sciences, 266, 22692274.
  • Conroy, J.W.H. & Chanin, P.R.F. (2000) The status of the Eurasian otter (Lutra lutra) in Europe. A review. (eds J.W.H. Conroy, P. Yoxon & A.C. Gutleb), Proceedings of the First Otter Toxicology Conference. Journal of the International Otter Survival Fund, 1, 728.
  • Corander, J. & Marttinen, P. (2006) Bayesian identification of admixture events using multilocus molecular markers. Molecular Ecology, 15, 28332843.
  • Cornuet, J.M. & Luikart, G. (1996) Description and power analysis of two tests for detecting recent population bottlenecks from allele frequency data. Genetics, 144, 20012014.
  • Coulon, A., Fitzpatrick, J.W., Bowman, R., Stith, B.M., Makarewich, C.A., Stenzler, L.M. & Lovette, I.J. (2008) Congruent population structure inferred from dispersal behaviour and intensive genetic surveys of the threatened Florida scrub-jay (Aphelocoma coerulescens). Molecular Ecology, 17, 16851701.
  • Coxon, K.E., Chanin, P.R.F., Dallas, J.F. & Sykes, T. (1999) The use of DNA fingerprinting to study the population dynamics of otters (Lutra lutra) in Southern Britain: feasibility study. The Environment Agency, R&D report W202 .
  • Crandall, K.A., Bininda-Emonds, O.R.P., Mace, G.M. & Wayne, R.K. (2000) Considering evolutionary processes in conservation biology. Trends in Ecology and Evolution, 15, 290295.
  • Crawford, A. (2003) Fourth Otter Survey of England 2000–2002. Environment Agency, Bristol.
  • Crawford, A. (2010) Fifth Otter Survey of England 2009–2010. Environment Agency, Bristol.
  • Crawford, A., Evans, D., Jones, A. & McNulty, J. (1979) Otter Survey of Wales 1977–78. Society for the Promotion of Nature Conservation, Lincoln.
  • Dallas, J.F. & Piertney, S.B. (1998) Microsatellite primers for the Eurasian otter. Molecular Ecology, 7, 12481251.
  • Dallas, J.F., Marshall, F., Piertney, S.B., Bacon, P.J. & Racey, P.A. (2002) Spatially restricted gene flow and reduced microsatellite polymorphism in the Eurasian otter Lutra lutra in Britain. Conservation Genetics, 3, 1529.
  • Dhuyvetter, H., Gaublomme, E. & Desender, K. (2005) Bottlenecks, drift and differentiation: the fragmented population structure of the saltmarsh beetle Pogonus chalceus. Genetica, 124, 167177.
  • Dubes, R. & Jain, A.K. (1976) Clustering techniques: the user’s dilemma. Pattern Recognition, 8, 247260.
  • Environmental Systems Research Institute (ESRI) (2007) ArcMap 9.2 Geographic Information System. Redlands, California.
  • Evanno, G., Regnaut, S. & Goudet, J. (2005) Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Molecular Ecology, 14, 26112620.
  • Excoffier, L., Estoup, A. & Cornuet, J. (2005) Bayesian analysis of an admixture model with mutations and arbitrarily linked markers. Genetics, 169, 17271738.
  • Fain, S.R., Straughan, D.J. & Taylor, B.F. (2010) Genetic outcomes of wolf recovery in the western Great Lakes states. Conservation Genetics, 11, 17471765.
  • Falush, D., Stephens, M. & Pritchard, J.K. (2003) Inference of population structure using multilocus genotype data: linked loci and correlated allele frequencies. Genetics, 164, 15671587.
  • Frantz, A.C., Cellina, S., Krier, A., Schley, L. & Burke, T. (2009) Using spatial Bayesian methods to determine the genetic structure of a continuously distributed population: clusters or isolation by distance? Journal of Applied Ecology, 46, 493505.
  • Goudet, J. (1995) FSTAT (Version 1.2): a computer program to calculate F-Statistics. Journal of Heredity, 86, 485486.
  • Green, J., Green, R. & Jefferies, D.J. (1984) A radio-tracking survey of otters Lutra lutra on a Perthshire river system. Lutra, 27, 85145.
  • Guillot, G., Mortier, F. & Estoup, A. (2005) GENELAND: a computer package for landscape genetics. Molecular Ecology Notes, 5, 712715.
  • Guo, S.W. & Thompson, E.A. (1992) Performing the exact test of Hardy-Weinberg proportion for multiple alleles. Biometrics, 48, 361372.
  • Hobbs, G.I., Chadwick, E.A., Slater, F.M. & Bruford, M.W. (2006) Landscape genetics applied to a recovering otter (Lutra lutra) population in the UK: preliminary results and potential methodologies. Hystrix Italian Journal of Mammalogy, 17, 4763.
  • Huang, C.-C., Hsu, Y.-C., Lee, L.-L. & Li, S.-H. (2005) Isolation and characterization of tetramicrosatellite DNA markers in the Eurasian otter (Lutra lutra). Molecular Ecology Notes, 5, 314316.
  • Jain, A.K., Murty, M.N. & Flynn, P.J. (1999) Data clustering: a review. ACM Computing Surveys, 31, 264323.
  • Jessop, R.M. & Cheyne, D.L. (1992) The re-introduction of the European otter Lutra l. lutra into lowland England carried out by the otter trust 1983/92: a progress report. Otters, Journal of the Otter Trust, ????, 1116.
  • Jones, T. & Jones, D. (2004) Otter Survey of Wales 2002. Environment Agency, Wales.
  • Kolbe, J.J., Glor, R.E., Rodriquez Schettino, L., Chamizo Lara, A., Larson, A. & Losos, J.B. (2004) Genetic variation increases during biological invasion by a Cuban lizard. Nature, 431, 177181.
  • Latch, E.K., Scognamillo, D.G., Fike, J.A., Chamberlain, M.J. & Rhodes, O.E. Jr (2008) Deciphering ecological barriers to north American river otter (Lontra canadensis) gene flow in the louisiana landscape. Journal of Heredity, 99, 265274.
  • Lenton, E.J., Chanin, P.R.F. & Jefferies, D.J. (1980) Otter Survey of England 1977–79. Nature Conservancy Council, London.
  • Lewis, P.O. & Zaykin, D. (2001) Genetic data analysis: computer program for the analysis of Allelic data, Version 1.1. Available from: URL http://hydrodictyon.eeb.uconn.edu/people/plewis/software.php.
  • Manel, S., Gaggiotti, O.E. & Waples, R.S. (2005) Assignment methods: matching biological questions techniques with appropriate techniques. Trends in Ecology & Evolution, 20, 136142.
  • Mills, L. & Allendorf, F.W. (1996) The one-migrant-per-generation rule in conservation and management. Conservation Biology, 10, 15091518.
  • O’Ryan, C., Bruford, M., Beaumont, M., Wayne, R., Cherry, M. & Harle, E. (1998) Genetics of fragmented populations of African buffalo (Syncerus caffer) in South Africa. Animal Conservation, 1, 8594.
  • Pearse, D.E. & Crandall, K.A. (2004) Beyond FST: analysis of population genetic data for conservation. Conservation Genetics, 5, 585602.
  • Philcox, C.K., Grogan, A.L. & Macdonald, D.W. (1999) Patterns of otter Lutra lutra road mortality in Britain. Journal of Applied Ecology, 36, 748762.
  • Pires, A.E., Amorim, I.R., Ginja, C., Gomes, M., Godinho, I., Simões, F., Oom, M., Petrucci-Fonseca, F., Matos, J., & Bruford, M.W. (2009) Molecular structure in peripheral dog breeds: portuguese native breeds as a case study. Animal Genetics, 40, 383392.
  • Pritchard, J.K., Stephens, M. & Donnelly, P. (2000) Inference of population structure using multilocus genotype data. Genetics, 155, 945959.
  • Pritchard, J.K. & Wen, W. (2003) Documentation for STRUCTURE software: Version 2. Available from: http://pritch.bsd.uchicago.edu.
  • QIAGEN, #65906, DNeasy tissue handbook, protocol: purification of total DNA from animal tissues, 1820.
  • Rannala, B. & Hartigan, J.A. (1996) Estimating gene flow in island populations. Genetics Research, 67, 147158.
  • Raymond, M. & Rousett, F. (1995) Genepop (version 1.2) population genetics software for exact tests and ecumenicalism. Journal of Heredity, 86, 248249.
  • Stanton, D.W.G., Hobbs, G.I., Chadwick, E.A., Slater, F.M. & Bruford, M.W. (2009) Mitochondrial genetic diversity and structure of the European otter (Lutra lutra) in Britain. Conservation Genetics, 10, 733737.
  • Storfer, A. (1996) Quantitative genetics: a promising approach for the assessment of genetic variation in endangered species. Trends in Ecology & Evolution, 11, 343348.
  • Strachan, R. & Jefferies, D.J. (1996) Otter Survey of England 1991–94. The Vincent Wildlife Trust, London.
  • Weir, B.S. & Hill, W.G. (2002) Estimating F-statistics. Annual Review of Genetics, 36, 721750.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Appendix S1. Results for the estimates of the most likely ‘optimum’ number of partitions (K) for the Bayesian Techniques.

Fig. S1. Estimation of K using structure.

Fig. S2. Estimation of K using geneland.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JPE_2028_sm_AppS1-FigsS1-S2.doc27KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.