Host ant specificity of large blue butterflies Phengaris ( Maculinea ) (Lepidoptera: Lycaenidae) inhabiting humid grasslands in East-central Europe

. Butterflies of the genus Phengaris have a highly specialised life cycle involving an obligatory relationship with Myrmica ants. A knowledge of the host ant specificity is essential for understanding the relationship between a particular Phengaris species and its hosts and also important for the conservation of these butterflies. Data on host ant specificity were collected in Poland, the Czech Republic, Slovakia and Ukraine. Five different Myrmica species were used by P. teleius as hosts ( M. scabrinodis , M. rubra , M. ruginodis , M. rugulosa and M. gallienii ) and at most localities it was not possible to distinguish a primary host – i.e. several Myrmica species were parasitized to similar extents. Three populations of P. nausithous were found in Poland and Ukraine. In every case, M. rubra was its primary host, although in the Kraków region (Poland) two nests of M. scabrinodis and two of M . ruginodis were infested by this butterfly species. P. alcon in the four populations investigated in Poland and Ukraine invariably only used M. scabrinodis as a host despite the presence of other Myrmica species. These results obtained suggest lack of host specificity in P. teleius and high host specificity in P. nausithous , which mainly uses M. rubra as its host across Europe. Moreover, the three populations of P. alcon investigated seem to be highly specific and use M. scabrinodis as a host, which confirms the high local specialisation of these populations.


INTRODUCTION
Most of the relationships between ants and lycaenid butterflies seem to be mutualistic (Pierce, 1987;Fiedler, 1991Fiedler, , 2001 but in some cases, larvae of certain butterfly species can exploit ant nests as a food resource and shelter and behave as well specialized social parasites (Cottrell, 1984;Maschwitz et al., 1984;Fiedler, 2001). Probably, the most studied parasitic myrmecophilous relationship is the one between Phengaris Doherty, 1891 (a junior synonym -Maculinea van Eecke, 1915, see Fric et al., 2007 butterflies and Myrmica Latreille, 1804 ants (Thomas & Settele, 2004). Females of Phengaris lay eggs on a specific foodplant and then after about three weeks, young larvae hatch from eggs and feed on the seeds or flowers of the plant. On reaching the fourth instar, the larvae drop to the ground and wait for foraging Myrmica ants, which take them to their nests. Phengaris caterpillars are parasites of Myrmica nests and have evolved different strategies for exploiting Myrmica host nests (Thomas & Elmes, 1998). Caterpillars of Phengaris teleius Bergsträsser, 1779 and P. arion Linnaeus, 1758 prey on ant brood and are called "predatory" species (Thomas et al., 1991;Thomas & Wardlaw, 1992), whereas those of P. alcon Denis &Schiffermüller, 1775 andP. rebeli Hirschke, 1905 are termed "cuckoo" species, as they mimic ant larvae and are fed directly by workers (Elmes et al., 1991a, b). There is no precise information on the feeding behaviour of P. nausithous Bergsträsser, 1779 larvae inside Myrmica nests, but this species may have an intermediate strategy (Thomas & Settele, 2004). The cuckoo species are more advanced in their behaviour and chemical mimicry of their host ants compared to the predatory species (Thomas & Elmes 1998;Als et al., 2004). Moreover, the cuckoo-feeding manner is more efficient and as up to 6-7 times more imagoes are produced per ant nest compared to the predatory species of Phengaris (Thomas & Wardlaw, 1992;Thomas et al., 1993).
Earlier work suggests that each Phengaris species has evolved to parasitize a single and different Myrmica species (Thomas et al., 1989) with each species adapted to a single "primary host" ant species and only occasionally found with other Myrmica, which are regarded as secondary hosts. Thus, M. scabrinodis Nylander, 1846 was recorded as the main host ant of P. teleius, while M. rubra Linnaeus, 1758of P. nausithous, M. ruginodis Nylander, 1846of P. alcon, M. schencki Emery, 1984of P. rebeli and M. sabuleti Meinert, 1860of P. arion respectively (Thomas et al., 1989. In contrast, recent studies conducted across Europe raise doubts about high host ants specialization of Phengaris (Elmes et al., 1994Als et al., 2002;Stankiewicz & Sielezniew, 2002;Tartally & Cs sz, 2004;Witek et al., 2006). Based on the above reports on Phengaris host ant specificity, Pech et al. (2007) conducted a review testing two hypotheses, (a) the species specificity and (b) local specialization hypotheses. They concluded that host ant specificity is poorly supported at present and there is no clear evidence for species-specificity in P. teleius, P. alcon and P. rebeli. Also, local-specialization only occurs in some populations and similarly in both cuckoo and predatory species. In this paper, extensive new data on host ant specificity are presented for P. teleius, P. nausithous, P. alcon from four East-central European countries: Poland, the Czech Republic, Slovakia and Ukraine. Most of this information is based on comparatively large samples and sheds further light on the discussion of both the host ant species-specificity of Phengaris butterflies in Central Europe and the local specialization of particular butterfly populations. This also has important conservation implications.

Study sites
The data on host ant specificity of P. teleius, P. nausithous and P. alcon were collected in four Central European countries: Poland, the Czech Republic, Slovakia and Ukraine. Information on particular field studies is presented in Fig. 1 and Appendix 1. The largest number of Myrmica nests was investigated at three sites in Poland (2520 nests). In the Czech Republic, Slovakia and Ukraine 257, 221 and 564, Myrmica nests were found. All of the localities investigated were wet grasslands where Sanguisorba officinalis Linnaeus was abundant. The data were collected from 2002 to 2004 in the case of Polish localities, and in 2005 in the Czech Republic, Slovakia and Ukraine. A total of 3562 Myrmica nests were investigated at 13 localities ( Fig. 1).

Methods
Field sampling was conducted from mid June to the beginning of July, i.e. in the period shortly before butterflies start to emerge. Myrmica nests were searched for within a distance of up to 2 m from Phengaris food plants, which is roughly the foraging distance of Myrmica ants , thus all nests examined potentially could have been parasitised. However, this was not always possible for Gentiana pneumonanthe Linnaeus as in early summer it is difficult to detect every plant of this species. In such cases Myrmica nests in parts of meadows where there were cluster of G. pneumonanthe were searched. All located Myrmica nests were opened to check for the presence of Phengaris larvae or pupae without the full excavation of nests. All discovered larvae and pupae were counted and in addition from each Myrmica nest 10-20 worker ants were collected and preserved in alcohol. Identification keys by Czechowski et al. (2002) andliwinska et al. (2006) were used for species identification of Myrmica and Phengaris, respectively.
To quantify the strength of host specificity the F index (proportion of adopted larvae that developed in nests of the primary host) of Thomas & Elmes (1998) was calculated: where: a is the proportion of the Myrmica nests of the primary host species within 2 m from Phengaris host plants and b is the proportion of fully grown Phengaris larvae or pupae found in those ant nests. Theoretically, F can vary between 0 (no specificity) to (complete specificity).

RESULTS
P. teleius was the most abundant species found in 292 nests (at 10 localities) belonging to five Myrmica species (Table 1, Appendix 1). The percentage of nests infested by P. teleius ranged from 1.58% at P elou (the Czech Republic) to 10.75% in the Kraków region (Poland). Larvae and/or pupae of P. teleius were found mostly in M. scabrinodis and M. rubra nests. M. scabrinodis was present at all localities used as a host by P. teleius at seven of them, whereas M. rubra was present at six localities and used as a host at each of them (Table 1). Other species of Myrmica such as: M. ruginodis, M. gallienii Bondroit, 1919 and M. rugulosa Nylander, 1846 served as hosts locally, and in some cases a relatively high proportion of their nests was parasitized (Table 1). Interestingly, there were three cases of different host ants at close by localities (20-40 km from each other): Novobarovo and Kireshi (Ukraine), Rudniki and Lviv -Rudno (Ukraine) and Stefanova and Drgonova Dolina (Slovakia). Fisher exact tests or Chi square test were calculated separately for each site in order to compare frequencies of occupied and unoccupied nests for all the ant species at these sites. The only significant result was obtained for the Kraków region ( 2 = 12.19, d.f. = 3, P = 0.006). In this intensively sampled region M. scabrinodis was the most abundant Myrmica species but the percentage of parasitism was the lowest compared to the other ant species (Table 1). Additionally, F indexes calculated for 6 populations of P. teleius showed that there was a mixed host used in the Kraków region (Poland), Sliwa (Poland) and Rudniki (Ukraine) (Table1).
P. nausithous was mostly associated with M. rubra (23 infested nests) and only in the Kraków region were its larvae occassionaly found in colonies of M. scabrinodis (2 nests) and M. ruginodis (2 nests). Larvae and pupae of P. alcon were found at four localities and invariably only M. scabrinodis (39 nests) was used as a host, even if other Myrmica species were abundant (Table 3). For the three populations of P. alcon investigated, F values show that M. scabrinodis is the species' primary host.
Depending on the locality, the number of larvae/pupae per infested Myrmica nest ranged (mean and SE in parentheses) from: 1-8 (2.2 ± 0.281), 1-11 (1.9 ± 0.41), 1-11 (1.61 ± 0.08) for P. alcon, P. nausithous and P. teleius, respectively and differed significantly between species (Kruskal-Wallis, 2 = 9.06, d.f. = 2, P = 0.01). Post hoc pairwise comparisons among groups (using Bonferonni correction, P = 0.016) revealed a significant difference between the number of larvae/pupae of P. alcon and P. teleius per infested Myrmica nest at all the localities (Mann-Whitney U Test, P = 0.001). Thomas et al. (2005) pointed out that the host specificity of social parasites of ants can increase as the results of better penetration of ant society and by interacting with ants in early stages of the life cycle. Thus, Phengaris should be species-specific in term of their host ants. However, the most recent data on host ant specificity in Phengaris butterflies contradicts this and in a few populations local specialization occurs (Als et al., 2002;Stankiewicz & Sielezniew, 2002;Stankiewicz et al., 2005;Tartally & Varga, 2005;Pech et al., 2007).

DISCUSSION AND CONCLUSIONS
Data presented in this paper concern 10 populations of P. teleius from different geographical regions of Central Europe. Of the six potential host species present there P. teleius larvae used five of them, the exception being M. sabuleti. In Kraków, Sliwa and Rudniki, mixed host populations are used. Interestingly, the F value (evaluates the strength of host specificity), is lower for M. scabrinodis than M. rubra and also lower than 1, which indicates lack of specificity for M. scabrinodis. These results are surprising, because the previous studies showed that the survival of P. teleius larvae reared in M. scabrinodis nests is five times greater than in the nests of other Myrmica species (Thomas et al., 1989 1992). Results presented here show that in 70% of the localities investigated, where M. scabrinodis and P. teleius co-occurred, the former species was used as the host. However, it is important to note that M. scabrinodis is the most frequent Myrmica species found at P. teleius localities and this may account for why it is so frequently used as a host. Moreover, similar data from Asia, where several host ant species were recorded, including M. angulinodis Ruzsky, 1905M. forcipata Karavaiev, 1931M. kamtschatica Kupyanskaya, 1986M. kurokii Forel, 1907M. lobicornis Nylander, 1846, and M. ruginodis suggests multiple host use by P. teleius (for a review see Als et al., 2004. Furthermore, Aphaenogaster japonica Forel, 1911 was found to be parasitized by P. teleius in Japan (Yamaguchi, 1988), proving that Myrmica is not the only host genus. To sum up, the findings of earlier studies and those presented in this paper confirm the lack of specificity of P. teleius and shows that larvae of this butterfly can use almost every Myrmica species that co-occur in its habitat.
Larvae and pupae of P. nausithous were found in three different populations in Poland, Slovakia and Ukraine, in which M. rubra was predominantly used as the host. This confirms the findings of earlier studies (Thomas et al., 1989;Figurny & Tomaszewicz, 1997;Stankiewicz & Sie-lezniew, 2002;Tartally & Varga, 2005) that P. nausithous almost exclusively uses M. rubra as its host across Europe. This high specificity is consistent with the finding that M. rubra has similar hydrocarbon profile over a wide geographical scale (from West Russia -near Moscow to West Scotland -Hebrides), which indicates that its colonies are chemically very similar across Europe (Elmes et al., 2002). Interestingly, in the Kraków region, we found two nests of M. scabrinodis and two M. ruginodis with larvae of P. nausithous. In this context, it is worth noting, that hydrocarbon profiles of M. ruginodis and M. rubra are chemically the most similar among the Myrmica species investigated by Elmes et al. (2002). The use of M. scabrinodis is also previously reported from Spain (Munguira & Martín, 1997) and recently from Romania (Tartally et al., 2008). However, one cannot exclude the possibility that all these exceptions are cases in which P. nausithous larvae were initially adopted by a M. rubra colony, the nest site of which was subsequently abandoned and taken over by a colony of another species. It is known that switching nest sites is a typical behaviour of Myrmica ants .
The F values calculated for Polish populations of P. alcon revealed that M. scabrinodis is its primary host ant there. This is not surprising since at each locality all larvae were found exclusively in nests of this Myrmica    Sielezniew & Stankiewicz (2001). Admittedly, the same authors also discovered two neighbouring populations of P. alcon in SE-Poland that simultaneously used M. scabrinodis and M. vandeli Bondroit, 1920 as host ants, but suspected that M. vandeli is a temporary social parasite of M. scabrinodis which is almost identical chemically (Radchenko et al., 2003;Sielezniew & Stankiewicz, 2004;Stankiewicz et al., 2005). Pech et al. (2007) pointed out that firm evidence for local specialization of P. alcon exists only for a few Danish populations (Als et al., 2002;Nash et al., 2008) while other observations do not reveal any local specificity (Elmes et al., 2002;Schlick-Steiner et al., 2004). In contrast, our data clearly show that in all the Polish populations of P. alcon investigated high localspecialization occurs.
It is suggested that the cuckoo is more efficient than the predatory strategy (Elmes et al., 1991b) resulting in higher numbers of larvae/pupa produced per Myrmica nest. For a few populations of P. alcon, P. nausithous and P. teleius recorded by Thomas & Elmes (1998) the mean number of larvae/pupae per infested nest was 6.0, 2.5 and 1.2, respectively. In this study the corresponding figures are 2.2 for P. alcon, 1.9 for P. nausithous and 1.6 for P. teleius. Hence, even though the general pattern recorded in both studies is very similar, the main difference is in the number of P. alcon larvae/pupae per Myrmica nest. The sample size is almost the same in both studies (39 vs. 37 nests) but Thomas & Elmes (1998) found quite a few nests with relatively high numbers (> 10) of P. alcon larvae. A possible reason could simply be a stochastic difference in the spatial overlap in the distribution of G. pneumonathe and ant nests which resulted in more Phengaris larvae entering Myrmica nests at the sites investigated by Thomas & Elmes (1998). Another possible explanation is that Thomas & Elmes (1998) sampled populations of P. alcon that used not only M. scabrinodis (as in the present study) but also M. rubra and M. ruginodis. It is known that M. rubra and M. ruginodis nests are usually larger than those of M. scabrinodis (Wardlaw & Elmes 1996;Radchenko et al., 1997;Skórka et al., 2006) and thus likely support more Phengaris larvae.
To summarise, this study confirms the high specificity of P. nausithous for M. rubra throughout its distribution range and recorded a new host (M. ruginodis) for this species. The results also indicate that P. teleius does not show either species-specificity or local specialization to host ants. Additionally, our findings clearly show a high local specialisation of P. alcon populations with M. scabrinodis being the primary host. This Myrmica species is the most common host ant of P. alcon in East-central Europe, although other species such as M. vandeli or M. salina are used occasionally (Sielezniew & Stankiewicz, 2001;Tartally, 2005;Tartally et al., 2008). Similar to Pech et al. (2007) we think that the regional host species pool is one of the most important factors affecting the host ant specificity of Phengaris butterflies, and that stating that a butterfly is specialised to a specific ant at a particular site, if is not supported by data "may lead to erroneous management prescriptions" (Pech at al., 2007, page 15). But, we also argue that insufficient information on host specificity may have the same effect (as was shown for P. arion by Thomas 1980, 1995. Therefore it is important to further investigate this aspect of the Large Blues' biology in order to improve management concepts (compare Johst et al., 2006;Drechsler et al., 2007), while in parallel we might be forced to mimic land-use systems as surrogate for the ants' requirements as long as these are not sufficiently known.