Conservation Genetics

, Volume 13, Issue 1, pp 247–255

Landscape genetics of a recent population extirpation in a burnet moth species

Authors

    • Invertebrate BiologyNatural History Museum Luxembourg
  • Jan O. Engler
    • Department of BiogeographyTrier University
    • Zoological Research Museum Alexander Koenig
  • Dennis Rödder
    • Zoological Research Museum Alexander Koenig
  • Thomas Schmitt
    • Department of BiogeographyTrier University
Research Article

DOI: 10.1007/s10592-011-0280-3

Cite this article as:
Habel, J.C., Engler, J.O., Rödder, D. et al. Conserv Genet (2012) 13: 247. doi:10.1007/s10592-011-0280-3
  • 243 Views

Abstract

The intensification of agricultural land use over wide parts of Europe has led to the decline of semi-natural habitats, such as extensively used meadows, with those that remain often being small and isolated. These rapid changes in land use during recent decades have strongly affected populations inhabiting these ecosystems. Increasing habitat deterioration and declining permeability of the surrounding landscape matrix disrupt the gene flow within metapopulations. The burnet moth species Zygaena loti has suffered strongly from recent habitat fragmentation, as reflected by its declining abundance. We have studied its population genetic structure and found a high level of genetic diversity in some of the populations analysed, while others display low genetic diversity and a lack of heterozygosity. Zygaena loti was formerly highly abundant in meadows and along the skirts of forests. However, the species is currently restricted to isolated habitat remnants, which is reflected by the high genetic divergence among populations (FST: 0.136). Species distribution modelling as well as the spatial examination of panmictic clusters within the study area strongly support a scattered population structure for this species. We suggest that populations with a high level of genetic diversity still represent the former genetic structure of interconnected populations, while populations with low numbers of alleles, high FIS values, and a lack of heterozygosity display the negative effects of reduced interconnectivity. A continuous exchange of individuals is necessary to maintain high genetic variability. Based on these results, we draw the general conclusion that more common taxa with originally large population networks and high genetic diversity suffer stronger from sudden habitat fragmentation than highly specialised species with lower genetic diversity which have persisted in isolated patches for long periods of time.

Keywords

Habitat fragmentationPopulation bottleneckLand-use changeAllozyme electrophoresisSpecies Distribution ModellingZygaena loti

Introduction

The intensification of agricultural land use and, consequently, the destruction of semi-natural habitats and the resulting fragmentation of formerly interconnected habitats is a major threat for the existence of many species. As such, this aspect of conservation biology is a strongly debated topic (Allendorf and Luikart 2007; Ladle and Whittaker 2010). In particular, species with specific habitat demands often exist in isolated populations (Collinge 2000; Hanski and Gaggiotti 2004). The intensification of agricultural land use is generally accompanied by a decrease in landscape permeability, which leads to reduced rates of gene flow (e.g. Goldberg and Waits 2010) as the remaining populations are often limited in size and isolated from each other. As a result, negative edge effects make these populations even more fragile for fluctuations and stochastic processes than interconnected populations (Melbourne and Hastings 2008). From a population genetic point of view, this situation may lead to repeated bottleneck events with their associated negative impacts on the genetic structure (e.g. the retention of deleterious alleles and/or a general loss of genetic information) and is often accompanied by a reduction of fitness (Reed and Frankham 2003; Frankham et al. 2008).

In order to interpret the population genetic structures of species, biologists must consider their ecology: while species with specific habitat demands can exist in isolation over long periods of time, they often display a reduced level of genetic diversity (Debinski 1994; Schmitt and Seitz 2004, Bereczki et al. 2005; Habel et al. 2009a, 2010). In contrast, common, highly abundant species generally show high(er) levels of genetic variability (Schmitt et al. 2003, 2005; Louy et al. 2007), although they mostly maintain their high intraspecific variability through the continuous exchange of individuals and hence continuous gene-flow.

To examine the impact of rapid habitat changes and resulting reduced abundance on the population genetic structure of organisms, we selected the burnet moth species Zygaena loti as our model organism. This lepidoteran is highly sedentary, similar to almost all Zygaena moths (Schmidt-Koehl 1977; Föhst 1992; Bourn 1995; Binzenhöfer et al. 2005), and is restricted to xerothermic meadows, clearings and subalpine slopes (Hofmann 1994). Based on earlier observations, this species is not endangered and is a common moth. The larvae feed on a variety of common herbal food plants (e.g. Hippocrepis comosa, Lotus corniculatus, Securigera varia and other Faboideae). Our study region in southwest Germany with adjoining areas in France and Luxembourg is located in the core area of its distribution range, which extends from Spain to Siberia. In the past, the burnet moth species Z.loti was widely distributed and commonly found on extensively used meadows and the skirts of forests, resulting in a large network of interconnected populations (Hofmann 1994). However, the large-scale agricultural intensification that has occurred in many parts of Europe in recent decades, and the resulting decrease in extensive land-use practices, such as sheep pasturing, have led to a severe reduction in the number and size of these habitats, with negative effects on the flora and fauna specific to them (Butaye et al. 2005; Wenzel et al. 2006). In addition, most of the remaining semi-natural grasslands are highly fragmented. As a result, the abundance of our study species has severely declined in the study area over the last few decades (Wenzel et al. 2006; S. Caspari and M. Weitzel, personal communication).

In order to study the genetic effects of this negative trend we analysed 16 allozyme loci of 271 individuals from seven populations of Z. loti. We estimated the number of panmictic clusters in space using a Bayesian approach and incorporating the spatial location of the data. We further applied species distribution modelling (SDM) based on high-resolution environmental data. By incorporating these two analytical approaches, we were able to highlight the permeability of the landscape structure and determine the genetic impact of the collapse of a population network.

Material and methods

Sampling

A total of 271 individuals of Z. loti were sampled at seven locations in June 2003–2005. The sampled populations were chosen to include populations of different sizes and degrees of connectivity. The sampling sites are located in parts of western Germany with adjoining areas in France and Luxembourg (see Fig. 1). The moths were netted in the field and immediately stored in liquid nitrogen until analysis by allozyme electrophoresis. Details on all sampled individuals, including dates of sampling, are given in Table 1.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-011-0280-3/MediaObjects/10592_2011_280_Fig1_HTML.gif
Fig. 1

Location of the study area and sampled populations (red circles): 1 Lissendorf (Germany, G), 2 Schönecken (G), 3 Ingendorf (G), 4 Niederanven (Luxembourg), 5 Freudenburg (G), 6 Montenach (France, F), 7 Clouange (F)

Table 1

Study sites with exact GPS coordinates, number of samples and date(s) of sampling(s) and means of the six parameters of genetic diversity analysed

Localitya

GPS coordinates

N

A

AR

He (%)

Ho (%)

Ho/He

P95 (%)

Ptot (%)

Occurrence of alleles (%)

Date(s) of sampling(s)

Lissendorf (G)

50.32; 6.60

40

2.33

2.33

22.8

9.4

0.41

60.0

60.0

80.0

24 June 2005

Schönecken (G)

50.14; 6.46

40

2.27

2.25

26.9

10.1

0.38

66.7

66.7

77.7

20 August 2005

Ingendorf (G)

49.94; 6.46

40

1.80

1.78

15.5

12.4

0.80

40.0

66.7

62.2

16/17 June 2004

Niederanven (L)

49.66; 6.25

40

1.73

1.70

14.0

12.2

0.87

33.3

60.0

60.0

13/16 June 2003

Freudenburg (G)

49.55; 6.54

36

2.33

2.32

25.5

6.6

0.26

66.7

73.3

80.0

26 June 2003 and 11 June 2004

Montenach (F)

49.43; 6.39

36

1.93

1.91

16.5

9.8

0.59

40.0

53.3

66.6

21/23 June 2005

Clouange (F)

49.27; 6.09

40

1.60

1.58

9.5

7.7

0.81

26.7

40.0

55.0

16 June 2003 and 08 June 2004

Total number/means (± SD)

 

276

2.00 (± 0.31)

1.98 (± 0.31)

18.7 (± 6.5)

9.7 (± 2.1)

0.52 (± 2.14)

47.6 (± 16.5)

60.0 (± 10.9)

68.57 (± 10.33)

 

A Mean number of alleles. AR allelic richness. He, Ho percentage of expected and observed heterozygosity, respectively. Ho/He ratio between Ho and He; P95 percentage of loci with the most common allele not exceeding 95%, Ptot total percentage of polymorphic loci

aG Germany, F France, L Luxembourg

Half of the abdomen of each adult was homogenised utrasonically in Pgm-buffer (Harris and Hopkinson 1978), and the homogenate was centrifuged at 17,000 g for 5 min. Electrophoresis was performed on cellulose acetate plates (Hebert and Beaton 1993) under the running conditions described in Table 2. A total of 13 enzyme systems representing 16 loci were analysed.
Table 2

Electrophoretic conditions for Zygaena loti

Locus

EC nr.

Number of loci

Buffera

Applications

Running time at 200 V (min)

G6pdh

1.1.1.49

1

TC

3

40

6Pgdh

1.1.1.44

1

TC

3

40

Idh

1.1.1.42

2

TC

2

40

Aat

2.6.1.1

2

TM

3

40

Mdh

1.1.1.37

2

TC

2

40

Pgi

5.3.1.9

1

TG

1

40

Gpdh

1.1.1.8

1

TM

3

40

Fum

4.2.1.2

1

TG

3

45

Me

1.1.1.40

1

TM

3

30

Gapdh

1.2.1.12

1

TC

3

30

Acon

4.2.1.3

1

TC

2

30

Hbdh

1.1.1.30

1

TM

3

45

Pgm

5.4.2.2

1

TC

2

30

aTC Tris-citrate, pH 8.2 (Richardson et al. 1986), TG Tris–glycine, pH 8.5 (Hebert and Beaton 1993), TM Tris-maleic acid, pH 7.0 (adjusted from TM pH 7.8 (Richardson et al. 1986)

Molecular genetics

Allele frequencies, genetic diversity indices [mean numbers of alleles per locus (A), percentage of expected heterozygosity (He), percentage of observed heterozygosity (Ho), percentage of polymorphic loci (Ptot) and percentage of polymorphic loci with the most common allele not exceeding 95% (P95)], and Nei’s genetic distance (Nei 1978) were calculated with Arlequin ver. 2.0 software (Schneider et al. 2000). Allelic richness (AR) based on the smallest sampling number (N = 36) was calculated using FStat software (Goudet 1995). Year-by-year analyses were performed using exclusively individuals sampled during a single year (i.e. generation). The Hardy–Weinberg equilibrium (Louis and Dempster 1987), genetic disequilibrium (Weir 1991), locus by locus F-statistics and analysis of molecular variance (AMOVA) were calculated using Arlequin ver. 2.0 software. FIS represents the genetic variance component among individuals within populations, and FST is the genetic variance component among populations. FCT was calculated for testing differences among years. Isolation by distance was tested using a mantel test as implemented in Arlequin ver. 2.0 software.

The software Geneland ver. 3.2.4 (Guillot et al. 2008; Guillot and Santos 2009) was used to estimate the number of panmictic groups throughout a given set of populations within the study area as well as their spatial boundaries. Geneland clusters individuals from different populations in such a way that the Hardy–Weinberg equilibrium and linkage equilibrium of each population are maximised. Spatial coordinates of the populations were incorporated in these analyses as in contrast to other widely used Bayesian clustering approaches like Structure (Pritchard et al. 2000), Geneland accounts for spatial correlation between populations. In a recent comparative study, Geneland has been shown to frequently outperform other spatial clustering approaches used for analysing genetic data (Safner et al. 2011).

In a first step, the number of populations (K) was estimated using a Markov Chain Monte Carlo (MCMC) algorithm for subsequent parameter setting. A total of 100,000 MCMC iterations were performed, with one sample saved per 100th iteration. The maximum rate of Poisson’s process was kept at its standard setting of 100. The MCMC was run ten times at these settings, allowing for an inference of K of between 1 and 7 (i.e. the maximum refers to the number of populations analysed in this study). The MCMC was then run another ten times with a fixed K, depending on the inferred value in the initial runs. From this, the iteration with the highest log posterior probability of population membership after a burn-in of 100 (i.e. deleting the first 100 values out of 1,000 saved from the thinned chain) was chosen for subsequent analyses.

To test for recent reductions in effective population sizes, we used the Bottleneck ver. 1.2.02 software program (Cornuet and Luikart 1996). This analysis is based on the expectation that a recent reduction in the effective size of a population leads to a reduction in both allele numbers and heterozygosity. As allele number decreases faster than heterozygosity in a recently bottlenecked population, the observed values are always higher than the expected equilibrium heterozygosity, which is computed from the observed number of alleles under the assumption of an equilibrium population (Luikart and Cornuet 1998).

Species distribution modelling

The potential distribution of Z. loti was modelled using the software Maxent 3.3.3e (Phillips et al. 2006, Phillips and Dudík 2008). Maxent has recently been developed to model species distributions from environmental niche estimates and has a wide applicability in the biological sciences (Elith et al. 2011), frequently outperforming competitive approaches in this field (Elith et al. 2006). For model building, several high-resolution (approximately 90 × 90 m in the study area) climatic and topographic layers were used together with a layer describing land-use characteristics as environmental predictors that putatively best describe the species’ environmental niche (Table 3). Information on species occurrence within the study region was obtained through several field surveys and supplemented from data from the GBIF database (www.gbif.org), resulting in 27 exact occurrence locations. Models were developed using the default Maxent settings: 100 models were developed each time, splitting the species records into 70/30%, with 70% used for model training and 30%, which were omitted from training, used for testing the predictive ability of the models via the area under the receiver operating characteristic (ROC) curve (Swets 1988). For further processing, the average prediction per grid cell across all 100 models was computed.
Table 3

List and sources of environmental layers used for the species distribution model as well as single variable contributions to the final model

Variable

Contribution to SDM (%)

Climate (obtained from www.worldclim.org; Hijmans et al. 2005)

  Mean temperature of warmest quarter (bio10)

0.33

  Mean temperature of coldest quarter (bio11)

2.56

  Annual precipitation (bio12)

1.84

  Precipitation seasonality (bio15)

3.70

  Precipitation of warmest quarter (bio18)

0.60

  Isothermality (bio3)

6.96

  Temperature annual range (bio7)

2.35

  Mean temperature of wettest quarter (bio8)

1.44

  Mean temperature of driest quarter (bio9)

1.10

Topography (obtained from the SRTM Shuttle mission; available at:www.earthexplorer.usgs.gov)

  Aspect

3.32

  Slope

19.77

Land use (obtained from the European Environmental Agency; available at: www.eea.europa.eu)

  CORINE land use

56.03

AUC of the SDM: 0.902 (SD 0.044)

 

SDM Species distribution model, AUC area under the receiver operating characteristic (ROC) curve, SD standard deviation

Results

Genetic diversity

Six parameters of genetic diversity (A, AR, He, Ho, Ptot, P95) were calculated for all populations (Table 2). The mean values and standard deviations were A, 2.00 (±0.31); AR, 1.98 (±0.31); He, 18.7 (±6.5); Ho, 9.7 (±2.1); Ptot, 60.0 (±10.9); P95, 47.6 (±16.5). On average, only 65.2% of all alleles detected were found in any one population. No significant differences in gene diversity and genetic structure (FCT) between generations (sampling years) were detected. Hence, all data were treated as a single dataset. The Bottleneck software program identified an excess of homozygotes for the Ingendorf, Niederanven and Clouange populations.

Genetic differentiation

Of the 16 loci examined, 12 were polymorphic (allele frequencies are given in Appendix). Forty-eight cases of significant deviation from Hardy–Weinberg equilibrium of single loci in discrete samples were detected. This is much higher than the number expected by chance (12.2). In general, no linkage disequilibrium was identified, thereby allowing the use of standard algorithms of population genetic analyses. The total genetic variance (as estimated by AMOVA) was 1.606 and was partitioned as follows: among populations, 0.218; among individuals within populations, 0.652; within individuals, 0.736. The respective F-statistics values among populations were FST = 0.136 and FIS = 0.470. The mean genetic distance (Nei 1978) among populations was 0.0525 (±0.0217 SD). Neither cluster analyses nor isolation by distance tests reflected the spatial distribution of populations.

All ten initial MCMC runs showed the highest average log posterior probability values for K = 4 (Fig. 2). The spatial distribution of these clusters was scattered across the study area, with the exception of one cluster connecting the populations Ingendorf, Niederanven and Clouange (Fig. 3a). A second panmictic cluster merged the Lissendorf and Montenach populations, but with a lack of any spatial connection. The remaining populations appeared to be isolated and the landscape interpolation for the most likely group classification was very patchy (Fig. 3a).
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-011-0280-3/MediaObjects/10592_2011_280_Fig2_HTML.gif
Fig. 2

Frequency of the number of panmictic clusters (K) along the Markov Chain after the burn-in

Species distribution model

The species distribution model suggests a very scattered potential distribution of Z. loti across the entire study area (Fig. 3b), wherein the predictive performance of the model is ‘excellent’ according to previous proposed classifications (AUC = 0.902; SD = 0.044; Swets 1988). The explanatory power of the variable remarkably differed between the a priori selected predictors, wherein CORINE land cover had the highest overall contribution (56%), followed by slope (19.7%). The remaining predictors, including all climatic factors, had just minor contributions to the final model best describing the environmental niche of this species, at least on this fine spatial scale (Table 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-011-0280-3/MediaObjects/10592_2011_280_Fig3_HTML.gif
Fig. 3

Spatial distribution of all four estimated panmictic clusters of Zygaena loti obtained using the Geneland software. a Each colour represents one group. b Potential distribution derived from the species distribution model (SDM) obtained using Maxent software. In the SDM, warmer colours indicate higher occurrence probabilities. The populations investigated are marked with red circles and numbered according to Fig. 1

Discussion

The genetic diversity estimates were found to vary strongly among the seven Z. loti populations analysed. Populations with high levels of genetic diversity show above-average diversity compared to other Zygaenid species (Z. anthylidis, Dieker et al. unpublished data; Z. exulans, Schmitt and Hewitt 2004; Z. infausta, Schmitt and Seitz 2004) and other sedentary butterflies (Debinski 1994; Britten et al. 1994; Bereczki et al. 2005; Louy et al. 2007). Low levels of genetic diversity detected in some Z. loti populations correspond with that found for other lepidopterans with specific habitat requirements, such as Melitaea aurelia (Habel et al. 2009b) and Maculinea species (Gadeberg and Boomsma 1997; Bereczki et al. 2005). Populations of Z. loti with low genetic diversity are additionally characterised by strong signals of heterozygosity deficit, most probably caused by population bottlenecks. However, the level of genetic diversity cannot be explained by habitat characteristics alone: for example, population “Ingendorf” is small, but adjacent to other populations, and yet it harbours low genetic diversity; individuals at “Freudenburg” exist in a large, however, strongly isolated habitat, but they display high genetic diversity; finally, the population from “Schönecken” is found in a 600 ha-conservation area, which is subdivided into several sub-habitats, and represents a genetically highly diverse population.

The results from the Geneland analyses provide further insights. While some populations based on their genetic structure still appear to be well connected (i.e. populations Clouange, Niederanven, Ingendorf), others are clearly isolated from each other (i.e. Schönecken and Freudenburg) or from other panmictic similar populations (i.e. Lissendorf and Montenach). This leads us to conclude that local effects, such as fluctuations, combined with the lapse of former interconnected population networks have triggered this discontinuous genetic structure.

Population studies in general have shown that some species display specific adaptations for long-term survival in small and isolated populations [e.g. the Red Apollo Parnassius apollo in the Mosel valley (Habel et al. 2009a) or Euphydryas galettii in the Rocky Mountains (Debinski 1994)]. Such species do not have compensative gene flow between populations and are strongly affected by repeated bottlenecks. However, these conditions dot not necessarily have negative impacts on individual fitness; rather, they usually lead to very simple genetic structures with low numbers of alleles per locus. Such genetic impoverishment is often in line with strong ecological adaptation, but there is no reduction in individual fitness if the habitat requirements are fulfilled. This coherence between narrow ecological niche width and low genetic diversity becomes evident in other Zygaenids, such as Z. carniolica (unpublished data), Z. anthylidis (Dieker et al. unpublished data), Z. exulans (Schmitt and Hewitt 2004) and Z. infausta (Schmitt and Seitz 2004). All of these species have specific habitat requirements (specific host plants or habitat structures, Hofmann 1994) and thus occur mostly in isolated populations, consequently leading to lowered genetic diversity over a longer time period. However, in our study species, Z.loti, many of the alleles studied showed high frequence diversity, with some of these alleles being rare. Species with rather panmictic distributions often have such a complex genetic structure, as has also been shown for the generalistic butterflies Polyomatus icarus (Schmitt et al. 2003), Maniola jurtina (Habel et al. 2009c), and Melanargia galathea (Habel et al. 2008). Species with interconnected populations generally do not show responses to isolation described above, but maintain their high genetic diversity by continuous individual exchange—as formerly was the case in Z. loti.

Our data highlight both high genetic diversity similar what is found in common species and low genetic diversity, which is typical for specialists. Conservation problems may arise when such genetically diverse populations become isolated and go through population bottlenecks within short time frames. This appears to have happened to Z. loti. A similar example is provided by the hermit butterfly, Chazara briseis, in the Czech Republic. Being widespread over the entire country in former times, the species has now vanished from the entire country except for one metapopulation system. An intensive mark–release–recapture experiment together with population genetic analyses revealed surprisingly high genetic diversity (Kadlec et al. 2010), clearly not showing the preconditions necessary to survive in isolation. This might be one of the reasons for many unoccupied habitat patches and the observed rapid population decline.

From a conservationists’ point of view, our data indicate that only few genetically viable populations are still surviving. These populations have to be the focus of conservation efforts. The single solution to re-establish the former high genetic diversity of Z. loti is to re-connect fragmented habitats to allow for migration and gene flow between populations (cf. Wallis DeVries et al. 2002; Wood and Pullin 2002; Vandewoestijne et al. 2005; Honnay et al. 2006). The scattered potential distribution within the study area highlights the patchy matrix of suitable habitats. However, the SDM also shows that the patches with high suitability are not isolated but connected through corridors of lower but still acceptable suitability scores. Consequently, habitat restoration at specific locations connecting viable populations to provide stepping stones seems to be a possible and practicable approach for conserving this and other species associated with similar habitats. In particular populations, which have retained high levels of genetic diversity, deserve protection and can be used for future resettlement of unoccupied suitable habitat patches. If no conservation efforts are applied, Z. loti might become another enigmatic case of extinction—not due to the complete loss of habitat structures, but through genetic distortion triggered by too fast habitat change. Our hypothesis, namely, that species with high genetic diversity being maintained by a population network react even more sensitively to land-use changes than specialist species with low genetic diversity, is confirmed by our data.

Acknowledgements

We acknowledge a grant from the German Science Foundation DFG (grant nr. SCHM 1659/3-1 and 3-2), from the German Academic Exchange Service (DAAD) and a scholarship “Arten- und Biotopschutz” from the Ministry of Rhineland-Palatinate to JCH. We are grateful to the governments of Rhineland-Palatinate and Luxembourg for the sampling permits.

Copyright information

© Springer Science+Business Media B.V. 2011