The genetic structure of phenologically differentiated Large Blue ( Maculinea arion ) populations ( Lepidoptera : Lycaenidae ) in the Carpathian Basin

The infraspecific taxonomy of the European populations of the Large Blue (Maculinea arion) is confusing. Several subspecies have been described mostly based on external morphological features. In the Carpathian Basin two subspecies have been distinguished. Maculinea arion arion flies from mid-May to mid-June and Maculinea arion ligurica is on the wing from the end of June to mid-August. The two forms show some differentiation in habitat use, but occasionally can also share habitats with two peaks in the appearance of butterflies. Our aim was to study the level and structure of genetic variation in a set of populations of the two phenologically different M. arion. Imagos were collected from 8 localities between 2000 and 2006. Enzyme polymorphism was analysed at 13 enzyme loci using polyacrylamide gel electrophoresis. In the analysis of the data, we estimated the parameters of polymorphism. To study the pattern of genetic differentiation F-statistics, hierarchical F-statistics and AMOVA were computed. GeneClass and Structure were both applied to analyse the differentiation between the two phenologically different sets of populations. Cavalli-Sforza and Edwards’ arc distances were calculated and a UPGMA dendrogram was constructed on the basis of the distance matrix. PCA analysis was also carried out using the allele frequencies of the individuals. The level of polymorphism was relatively high in M. arion. The results of all analyses indicated that the differences between the two sets of phenologically different populations accounted for a low percentage of the total differentiation. In addition, a sizeable amount of variation could be attributed to the differences among the samples collected from the same population in consecutive years. Thus, we concluded that the “spring” and “summer arion” could not be considered as separate ESUs, although we could attribute conservation value to both forms on the basis of their phenological differentiation and habitat use. 519 * Here we follow the recommendation of Balletto et al. (2010) concerning the generic names Maculinea Van Eecke, 1915 vs. Phengaris Doherty, 1891. the Carpathian Basin M. arion has considerable morphological variation and several subspecific taxa have been named (Szabó, 1956; Gozmány, 1968). Following Bálint (1996) generally two subspecies (or even species) have been distinguished. Maculinea arion arion (Linnaeus, 1758) (see below as “spring type arion”) flies from mid-May to mid-June and prefers short-grass dry swards with cushions of early flowering Thymus species (Th. serpyllum L., Th. pannonicus All. and related species), which serve as initial food plants. Maculinea arion ligurica (Wagner, 1904) (see below as “summer type arion”) is on wing from the end of June to mid-August and mostly occurs at xerothermic oak forest fringes, on woodland clearings and in fen-like habitats in hilly areas. Females oviposit among flower-buds of late-flowering Thymus species (mostly Th. pulegioides L.) and/or Origanum vulgare L. Hitherto no evidence has been found concerning the genetic or reproductive isolation between the two forms (Munguira & Martin, 1997). Although the two forms locally share habitats (Varga, 2003; Tóth & Bereczki, pers. obs.) their syntopic populations are morphologically and phenologically differentiated. Our knowledge on the host ant use of the two subspecies is insufficient. Myrmica sabuleti Meinert, 1861 is known to be the main host ant of M. arion in Western Europe and it was also found as a putative host ant in some M. arion habitats in Hungary (Tartally, 2008). Thus, the taxonomic status of these two subspecies is fairly uncertain, in a similar way to the case of the M. alcon species group (Als et al., 2004; Pech et al., 2004; Bereczki et al., 2005; Fric et al., 2007; Pecsenye et al., 2007). Despite the conservational significance of the two phenologically and ecologically different forms, former genetic and morphometric studies (Als et al., 2004; Pech et al., 2004; Fric et al., 2007) did not concern them. Therefore, our goal was to study genetic variation in phenologically different M. arion populations. In addition, we wanted to know whether these taxa represented separate Evolutionary Significant Units (ESUs). Numerous definitions of ESU exist (Ryder, 1986; Dizon et al., 1992; Vogler et al., 1993; Moritz, 1994; Waples, 1998; Paetkau, 1999). The majority of these definitions suggest that an ESU should be geographically discrete and show divergence of both molecular and nonmolecular traits (Ryder, 1986; Dizon et al., 1992; Vogler et al., 1993). In fact, the ESU concept has different components, namely reproductive and historical isolation as well as adaptive distinctiveness (Crandall et al., 2000). Therefore, the aims of our study were: (i) to estimate the level of genetic diversity of M. arion; (ii) to reveal the structure of genetic variation, particularly the differentiation between the “spring” and “summer type” of M. arion; (iii) to determine whether these taxa represent separate Evolutionary Significant Units (ESUs). The story of Large Blues in the UK demonstrated that the detailed ecological, demographic and genetic knowledge on the species in question is the pre-requisite of a successful conservation strategy. Thus, we feel that our results will contribute to the successful conservation of M. arion in the Carpathian Basin. MATERIAL AND METHODS


INTRODUCTION
Maculinea Van Eecke, 1915* (Lepidoptera: Lycaenidae) species are the most intensively studied group of Palearctic butterflies due to their very special social parasitic life cycle and also owing to the severe decrease and fragmentation of their habitats.Numerous Western conservation biologists regard them as "flagship" species (Anonymus, 1993).Simultaneously they are also umbrella species (Shrader-Frechette & McCoy, 1993;Lambeck, 1997;New, 1997;Fleishman et al., 2005;Spitzer et al., 2009) representing the needs of other species through their special and complex habitat requirements.During their development they need particular food plants and host ant species.Their early larval instars feed on the seeds of their initial food plant(s) as mono-or oligophagous feeders.However, they gain only a little weight in this period (Elmes et al., 1991a, b;Thomas & Wardlaw, 1992).After 2-3 weeks larvae drop to the ground and wait for foraging ant workers (mostly Myrmica Latreille, 1804), who find and drag them to their nests (Pueefoy, 1953;Thomas, 1976).In the next stage of development they follow either a "predatory" or "cuckoo" strategy in the nests for 10-11 months (Thomas et al., 2000), gaining approximately 98-99% of their final body weight (Elmes et al., 1991a).After pupation imagos emerge from the nest.These special habitat requirements result in their vulnerability (Settele & Kuhn, 2009).
In the last few decades all Maculinea species have experienced severe declines over most of their ranges (Thomas, 1995;Wynhoff, 1998).One of the most spectacular declines was shown by Large Blues -Maculinea arion (Linnaeus, 1758).The species became extinct in the Netherlands in 1964 (Tax, 1989), in the UK in 1979 (Thomas, 1995) and in Belgium in 1996 (although later the species was reintroduced into the UK and it also recolonized Belgium) (Goffart, 1997;Thomas et al., 2009).It shows serious regression all over Europe, especially on the northern border of the distribution area of the species (Wynhoff, 1998).Large Blues are endangered on a European scale.The species is listed in the IUCN Red List of Threatened Species as "near threatened" and it is included in Annex IV of the European Habitats' Directive (Munguira & Martin, 1997;Van Swaay et al., 1998).
The taxonomic status of Large Blue populations in Central and Western Europe is confusing.Several subspecies have been described mostly on the basis of external morphological features such as light silvery blue vs. dark violet-blue colouration or black maculation of wings.In the Carpathian Basin M. arion has considerable morphological variation and several subspecific taxa have been named (Szabó, 1956;Gozmány, 1968).Following Bálint (1996) generally two subspecies (or even species) have been distinguished.Maculinea arion arion (Linnaeus, 1758) (see below as "spring type arion") flies from mid-May to mid-June and prefers short-grass dry swards with cushions of early flowering Thymus species (Th.serpyllum L., Th. pannonicus All. and related species), which serve as initial food plants.Maculinea arion ligurica (Wagner, 1904) (see below as "summer type arion") is on wing from the end of June to mid-August and mostly occurs at xerothermic oak forest fringes, on woodland clearings and in fen-like habitats in hilly areas.Females oviposit among flower-buds of late-flowering Thymus species (mostly Th. pulegioides L.) and/or Origanum vulgare L. Hitherto no evidence has been found concerning the genetic or reproductive isolation between the two forms (Munguira & Martin, 1997).Although the two forms locally share habitats (Varga, 2003;Tóth & Bereczki, pers. obs.) their syntopic populations are morphologically and phenologically differentiated.Our knowledge on the host ant use of the two subspecies is insufficient.Myrmica sabuleti Meinert, 1861 is known to be the main host ant of M. arion in Western Europe and it was also found as a putative host ant in some M. arion habitats in Hungary (Tartally, 2008).Thus, the taxonomic status of these two subspecies is fairly uncertain, in a similar way to the case of the M. alcon species group (Als et al., 2004;Pech et al., 2004;Bereczki et al., 2005;Fric et al., 2007;Pecsenye et al., 2007).Despite the conservational significance of the two phenologically and ecologically different forms, former genetic and morphometric studies (Als et al., 2004;Pech et al., 2004;Fric et al., 2007) did not concern them.Therefore, our goal was to study genetic variation in phenologically different M. arion populations.
Therefore, the aims of our study were: (i) to estimate the level of genetic diversity of M. arion; (ii) to reveal the structure of genetic variation, particularly the differentiation between the "spring" and "summer type" of M. arion; (iii) to determine whether these taxa represent separate Evolutionary Significant Units (ESUs).The story of Large Blues in the UK demonstrated that the detailed ecological, demographic and genetic knowledge on the species in question is the pre-requisite of a successful conservation strategy.Thus, we feel that our results will contribute to the successful conservation of M. arion in the Carpathian Basin.
stored at -80°C until electrophoresis.Sample sizes varied between 11 and 59.

Molecular analyses
Allozyme polymorphism was studied at 13 loci (see Table 2) by vertical polyacrylamide gel electrophoresis.Thoraxes homogenized in 300 µl of extraction buffer were used to study Gpdh, G6pgdh, Hk, Idh, Mdh, Me, Pgi, Pgm and Sod.Abdomens homogenized in 200 µl of extraction buffer were used to analyse Acon, Acph, Aox and Est.The extraction buffer, the electrophoresis buffer systems and running conditions, together with the staining solutions were used according to Bereczki et al. (2005).Genotypes of the individuals were scored according to their enzyme pattern.

Statistical analyses
Parameters of polymorphism (average number of alleles per locus, proportion of polymorphic loci, average observed heterozygosity, average number of multi-locus genotypes, total number of alleles and distribution of alleles) were calculated for each population.Genetic differentiation among the populations was first analysed by Wright's F-statistics (Wright, 1978;Weir, 1996).In this analysis, the distribution of genetic variation is characterised by three indices (FIT -total variation, FIS -within population variation and FST -between population variation).The analyses were computed by using Fstat ver.1.2 (Goudet, 1995).An exact test for population differentiation (Raymond & Rousset, 1995a) was also conducted to test for independence of the allelic composition of the populations.Genepop ver.1.0 (Raymond & Rousset, 1995b) was used to perform this test.Indices of differentiation (FST, average Cavalli-Sforza and Edwards' arc distances, frequency of private alleles and standard deviation of allele frequencies) were calculated for "spring" and "summer type", respectively.
The distribution of the total genetic variation at various levels of the hierarchy was also studied by AMOVA (Excoffier et al., 1992;Weir, 1996).In this analysis, the total genetic variation was partitioned into the following components: among groups (i.e.putative subspecies), among populations within a group, among samples within populations and among individuals within samples.AMOVA was carried out by Arlequin ver 2.000 (Schneider et al., 2000).
The genetic structure of the populations was also analysed by the Bayesian-clustering method.Here, we estimated the most probable number of genetically differentiated groups (K) in our samples and assigned the individuals to these groups.Structure 2.3.2 was run to carry out these analyses with an initial burn in 30,000 and a running length of 150,000 (Pritchard et al., 2000).
In the last part of the study, a principal component analysis (PCA) was conducted using the allele frequency data of the individuals to reveal the relationship between the previously sorted groups (putative subspecies).PCA analyses were performed using Past ver.1.56 (Hammer et al., 2001).
Finally, we carried out an assignment test.In this test log likelihood values were calculated for each individual multilocus genotype for all the samples on the basis of the expected distribution of genotype frequencies of the respective sample (Paetkau et al., 1995).Using these values we assigned each individual to the most probable sample.The computation of the assignment test was performed by GeneClass2 (Piry et al., 2004)

RESULTS
Alternative alleles occurred at all loci in at least one sample.M. arion exhibited the highest level of polymorphism of all sampled Maculinea species: the portion of polymorphic loci was 45.8% and the average observed frequency of heterozygotes was 0.155.The parameters of genetic variation indicated a slightly higher level of polymorphism in the "spring type" than in the "summer type" populations (Table 1).A similar tendency was observed in the distribution of alleles (Fig. 2).The proportion of rare alleles was slightly higher, while that of the fixed alleles was smaller in the "spring arion" samples than in the "summer type" ones.The total number of alleles was much higher in the "summer arion" than in the "spring type" (Fig. 2) but this could have been the consequence of larger sample sizes (Appendix 1).
The results of F-statistics indicated that a relatively high portion of variation was explained by the differences among the samples (Table 2: FIT).Nevertheless, only 5 loci of the total 13 proved to be significantly differentiating (Table 2: FST).At the same time, the results of the exact test showed that the samples were significantly differentiated at seven loci (Table 2: exact test).The within population variation was also sizeable.Although the average FIS value suggested significant heterozygote deficiency this was only attributable to 3 loci (Table 2: FIS).
The fixation index (Table 3: FST) showed significant differentiation both among the "spring" and "summer arion" samples.FST values and Cavalli-Sforza and Edwards' arc distances suggested a slightly higher level of differentiation among the "summer arion" than among the "spring type" populations (Table 3: FST and D).However, it is important to note that we had more "summer arion" than "spring type" samples.Moreover, the "summer arion" originated from several geographical regions.The other two measures of differentiation (frequency of private alleles and standard deviation of allele frequencies) were similar in the two types (Table 3: q and sp).
The results of AMOVA showed the distribution of the total sample variation between the spatial and temporal components (Fig. 3).The spatial component could be further divided into two parts: variation between the putative subspecies (Fig. 3: BSS) and variation among the populations within the subspecies (Fig. 3: AP).The differences between the two subspecies accounted for a lower portion of the total level of differentiation than that among populations.The temporal component of variation indicated a sizeable level of differentiation among the samples/generations collected from the same population (Fig. 3: AG).Most of the total genetic variation was attributable to the variation among the individuals (Fig. 3: AI).We also carried out AMOVA in the case of the "spring" and "summer arion", respectively (Fig. 3).Both the spatial and the temporal component of variation (Fig. 3: AP and AG) proved larger in the "summer arion" samples than in the "spring type" ones.Probably it arose from partly larger sample sizes of the "summer arion" and partly the fact that the "summer arion" samples originated from four geographical regions located at larger distances from each other.In contrast, fewer "spring type" samples originated from the same region (Fig. 1).
We also analysed the differentiation between the putative subspecies using Structure (Fig. 4).The most probable number of genetically differentiated groups (K) in our samples was 3 (LnP(D) = -3797.5,Var[LnP(D)] = 454.4,a1 = 0.0695).The "spring" and the "summer arion"  3. Indices of differentiation.FST -variation among populations; D -Cavalli-Sforza and Edwards' arc distances; qfrequency of private alleles; sp -standard deviation of allele frequencies.* 0.00077 < p < 0.0038 (after Bonferroni's correction based on 13 loci).did not separate clearly from one another.The "spring arion" samples showed a more homogenous pattern than those of the "summer arion", in the case of which periodically recurring patterns could be observed.Probably all this originated from the fact that the sample sites of the "spring arion" were located in the same geographic area (Fig. 1).On the contrary, the "summer arion" samples were collected in four geographic regions a long distance apart.Therefore, the combination of the spatial and tem-poral variation may have generated larger differences among the "summer arion" samples similarly to the results of AMOVA (see previously).

522
Cavalli-Sforza and Edwards' genetic distances were calculated and an UPGMA dendrogram was constructed on the basis of the distance matrix, which also included all the examined Maculinea species (Fig. 5).The samples of the two different putative subspecies were randomly scattered on the dendrogram.The scale of differentiation between the two types was similar to that of two types of M. alcon ([Denis & Schiffermüller], 1775).Earlier the latter two forms were generally treated as two distinct species based on the different initial food plants and host ant species (Elmes & Thomas, 1987a, b;Thomas et al., 1989;Elmes et al., 1991aElmes et al., , b, 1994;;Munguira & Martin, 1997).However, recent studies suggested that those represented just local ecological adaptations (Als et al., 2004;Pech et al., 2004;Bereczki et al., 2005;Fric et al., 2007;Pecsenye et al., 2007).At the same time, it is remarkable that the samples originating from the same population but collected in different years (generations) were not always clustered together.These findings are in good agreement with the results of Structure and AMOVA (see previously).The results of the PCA analyses fully confirmed our results discussed so far (Fig. 6).While M. arion samples exhibited overlapping clouds of points the other Maculinea species were separated evidently in the reduced space of variables.The two axes explained 60.5% of the total variation.The first axis was predominantly determined by the Mdh, Gpdh and Hk loci.The second axis was mostly affected by the allele frequency distribution at the Mdh, Hk and Gpdh locus.
In the assignment test (Fig. 7) there were individuals in both types approximately at the same percentage (6.21and 5.24%, respectively) that could not be classified.The program grouped the "spring arion" individuals correctly to a higher extent (71.19%) than the "summer type" ones (43.95%).It was the consequence of the fact that the genetic composition of the "spring arion" samples was relatively homogenous, while there were considerable differences among the "summer arion" samples.
The level of polymorphism was the highest in M. arion compared to the other sampled Maculinea species.This is probably due to its higher dispersal ability.Maculinea butterflies are generally regarded as sedentary species, although their mobility seems to be stronger than previously thought (Thomas, 1995;Munguira & Martin, 1997).The dispersal ability of M. arion is the highest of all Maculinea species (Pajari, 1992;Geissler & Settele, 1990;Settele et al., 1996).Our field experience also supports this observation (Árnyas, 2007; Árnyas et al., unpubl. data).We have also observed a regular movement of M. arion butterflies between steppic grassland patches separated by sparsely forested areas in the Aggtelek karst area (field observations of Z. Varga).As a consequence of their higher dispersal ability, their reproductive opportunities can become wider.Therefore, they probably have a more intense gene flow, which can partly balance the effect of genetic drift.Additionally, their larvae are able to survive starvation up to three weeks after the desertation of ant workers, until the nest is recolonized by a neighbouring Myrmica colony (Thomas & Lewington, 2010).It is possible that the high mobility of M. arion is associated with this kind of larval survival strategy.A sufficient quantity of ant brood is necessary for the sustenance of larvae thus imagos fly over large areas and spread across habitat patches.Therefore, they are not present at high density in any particular patch.Additionally, in contrast to other Maculinea species, it is only exceptional that more than one M. arion larva can develop in a single host ant nest (Thomas & Lewington, 2010).All this can counterbalance the stochasticity of larval adoption.
As for the level of polymorphism in M. arion it appeared slightly higher in the "spring arion" than in the "summer type" populations, but there were no significant differences between the two types concerning their genetic diversity.

The genetic structure
The differences between the putative subspecies of M. arion accounted for a lower percentage of the total level of differentiation.The scale of separation between the "spring" and the "summer arion" was similar to the level of differentiation between the "pneumonanthe" and the "cruciata type" of M. alcon.Molecular data indicate that the latter two are not separated at the species level and they represent local ecological adaptations (Als et al., 2004;Pech et al., 2004;Bereczki et al., 2005;Fric et al., 2007;Pecsenye et al., 2007).
The level of differentiation among populations (the spatial component of variation) was lower than that among the samples from the same population (the temporal component of variation).That means a considerable fluctuation in allele frequencies occurred among generations.These findings are consistent with our former results (Bereczki et al., 2005) and imply a strong effect of genetic drift, which is expected considering the life cycle of Maculinea species.It is also supported by field experience (Korosi et al., 2005) which revealed that the population size of M. arion fluctuates considerably despite the fact that the habitat and abundance of food plant appears stable.

ESUs
Neither of the ESU criteria is realized in the case of the two phenologically differentiated sets of populations of M. arion; therefore, the "spring" and "summer arion" cannot be considered as ESUs.The two sets of populations were not separated evidently on the basis of allozyme studies.Additionally, our data only weakly support the ecological isolation of the two forms, since there is a certain extent of overlapping in the initial food plant usage of the "spring" and "summer arion" due to the relatively long flowering period of different Thymus species, and also because the "summer arion" can simultaneously use Thymus spp.and Origanum vulgare.Nevertheless, differences in food plant use also imply some difference in habitat preference.The "spring arion" mostly feeds on xerophilous "steppic" Thymus species (e.g.Thymus pannonicus), while the "summer arion" occupying heath-like acidic habitats also utilize Thymus pulegoides (oviposition observed by Z. Varga).Although the two forms generally have different habitat preferences locally they often co-occur (Tóth & Bereczki, pers. obs.).These sympatric populations are phenologically more or less isolated.Nevertheless, there are non-evolutionary reasons for attributing conservation value to different sets of populations (Moritz, 1994).In the case of the two forms of M. arion this could be the phenological separation and the differences in habitat preference which can imply different types of habitat conservation (i.e.short-grass steppic vs. heath-like habitats).Although we do not have sufficient information regarding the status of the two forms of M. arion, we have to strive to preserve their genetic variability as widely as possible.Recently genetic variation has been ascribed great importance in terms of the climate change, as larger variation may facilitate the adaptation to changes in climate.Nevertheless, it would be important to carry out further studies to enhance our knowledge of the ecological characteristics of the two types of M. arion better and be able to work out more effective conservation strategies.

Fig. 2 .
Fig.2.Total number and the distribution of alleles in populations and in the case of M. arion "spring type" (sp) and "summer type" (sum), respectively.Categories: rare: p < 0.05; common: 0.05 < p < 1; fixed: p = 1.For the abbreviations of the populations see Appendix 1.

Fig. 3 .
Fig. 3.The distribution of variation at different levels of the hierarchy (AMOVA).Levels of hierarchy: BSS: between the putative subspecies; AP -among populations; AG -among generations originated from the same population; AI -among individuals; WI -within individuals. 523

Fig. 5 .
Fig. 5. Dendrogram (unweighted pair group method with arithmetic mean) constructed on the basis of Cavalli-Sforza and Edwards' arc distances.Outgroups were marked with bold style.Samples of M. arion "spring type" are underlined.For the abbreviations of the populations see Appendix 1. Numbers before abbreviations indicate the year of sampling.

Fig. 6 .
Fig. 6.The results of PCA.Each symbol represents one sample in the reduced space by variances.The first two axes explain 60.5 percent of the total variance.Mdh, Gpdh and Hk loci have the highest loading in Axis 1. Mdh, Hk and Gpdh loci have the highest loading in Axis 2.

Fig. 7 .
Fig. 7.The results of the assignment test.

TABLE 1 .
. Parameters of polymorphism.N -sample size; nA -average number of alleles per locus; P -proportion of polymorphic loci; H -average observed frequency of heterozygotes; G -average number of multilocus genotypes.

TABLE 2 .
Results of F-statistics and exact test of population differentiation computed for all 21 samples regardless of the putative subspecies they belong.FIT -total genetic variation; FST -variation among populations; FIS -variation within populations; ns -not significant; * 0.00077 < p < 0.0038; *** p < 0.000077 (after Bonferroni's correction based on 13 loci).
Sample sites (number of individuals / year).