The re-discovered Maculinea rebeli (Hirschke, 1904): Host ant usage, parasitoid and initial food plant around the type locality with taxonomical aspects (Lepidoptera, Lycaenidae)

Abstract The taxonomy of the myrmecophilous Maculinea alcon group (Lepidoptera: Lycaenidae) is highly debated. The host-plant and host-ant usage of these butterflies have conventionally been important in their identification. Maculinea ‘rebeli’ has generally been considered to be the xerophilous form of Ma. alcon (Ma. alcon X hereafter) with Gentiana cruciata as initial food plant. However, the type locality and all other known sites of Ma. rebeli are found above the coniferous zone, and are well separated from the lower regions where Ma. alcon X sites are found. Furthermore, no food plant and host ant data for the nominotypic Ma. rebeli have yet been published. Our aim was therefore to identify the host ant(s) of Ma. rebeli around the type locality and compare this with the host ant usage of nearby Ma. alcon X. Nests of Myrmica spp. (Hymenoptera: Formicidae) close to the host plants were opened on one Ma. alcon X (host plant: Gentiana cruciata) and two Ma. rebeli (host plant: Gentianella rhaetica, first record, confirmed by oviposition and emerging larvae) sites just before the flying period, to find prepupal larvae and pupae. Three Myrmica species (My. lobulicornis, My. ruginodis, My. sulcinodis) were found on the two Ma. rebeli sites, which parasitized exclusively My. sulcinodis (22 individuals in 7 nests). On the Ma. alcon X site Myrmica sabuleti and My. lonae were found, with My. sabuleti the exclusive host (51 individuals in 10 nests). Ichneumon cf. eumerus parasitized both butterflies. The results highlight the differentiation of Maculinea rebeli from Ma. alcon X, from both conservation biological and ecological points of view. Thus, it should be concluded that Ma. rebeli does not simply represent an individual form of Ma. alcon but it can be considered as at least an ecological form adapted to high mountain conditions both in its initial food plant and host ant species. In addition, it should be emphasized that Ma. alcon X (= Ma. rebeli auct. nec Hirschke) cannot be synonymised with Ma. rebeli (Hirschke, 1904).


Introduction
Although Maculinea van Eecke, 1915 (Lepidoptera, Lycaenidae) has been synonymised with Phengaris Doherty, 1890 (see , Kudrna and Fric 2013 we continue to use the well-established generic name here, since (i) the case is still undecided by the International Commission on Zoological Nomenclature (ICZN 2012) and strong arguments for the precedence of Maculinea over Phengaris have been made (Balletto et al. 2010, and Comments on this Case); and (ii) Ugelvig et al. (2011) have shown that Maculinea represents a monophyletic sub-clade and we agree with their conclusion: "We recommend that the nomenclatural debate is delayed until irrefutable evidence is provided".
Maculinea (abbreviated as "Ma." below) butterflies are endangered species (Munguira and Martín 1999, Maes et al. 2004 and their larvae are social parasites of Myrmica Latreille, 1804 (Hymenoptera: Formicidae; abbreviated as "My." below) ant colonies (Thomas et al. 1989). Larvae of Maculinea alcon ([Denis & Schiffermüller], 1775) initially feed on the seeds of gentian plants for about a month and are later taken in and raised by Myrmica colonies (Thomas et al. 1989). They have a "cuckoo" strategy, in which they mimic the odour (Thomas and Settele 2004, Akino et al. 1999) and sound (Barbero et al. 2009) of the host ant species, and are therefore fed like an ant larva , although they also sometimes prey directly on the ant brood (Tartally 2004). The caterpillars develop within the host ant nest during the autumn, winter and spring, and will then either pupate there, or continue development for an additional year (Schönrogge et al. 2000, Witek et al. 2006). There are regional and local differences between the host ant usage of different populations (e.g. Elmes et al. 1994, Als et al. 2002, Sielezniew and Stankiewicz 2004, Stankiewicz et al. 2005, Witek et al. 2008, Tartally et al. 2013. Knowledge of the local host ant species is vital to understand the biogeography, conservation and evolutionary biology of Maculinea spp. Kühn 2009), and may also have some taxonomic implications. Furthermore, the host ant and host plant usage of Ma. alcon has been thought to be important for taxonomic separation (Thomas et al. 1989) of the hygrophilous ('Ma. alcon H' below) and the xerophilous ('Ma. alcon X') form of Ma. alcon, although recent studies have been unable to show any consistent genetic separation between these two butterflies (Als et al. 2004, Bereczki et al. 2005, Ugelvig et al. 2011.
Despite this, Ma. alcon X is still generally (and erroneously, see Kudrna and Fric 2013) referred to as 'Ma. rebeli (Hirschke, 1904)' (e.g. Elmes et al. 1998, Stankiewicz et al. 2005. Some Balkan, Caucasian and Central Asian mountainous populations of Ma. alcon have also been referred to as Ma. 'rebeli' (Kolev 2002, Savela 2014, Tuzov 2000. However, a recent paper (Habeler 2008) called our attention to the fact that Ma. alcon X (= Ma. rebeli auctorum nec Hirschke, 1904) is not likely synonymous with Ma. rebeli described by Hirschke (1905) from the type locality, based on external morphological features and the different habitat types in which they occur. It has now been demonstrated that Ma. rebeli auct. does not represent a distinct species, but rather the xerophilous form of Ma. alcon (Ma. alcon X). It feeds initially typically on G. cruciata (Bereczki et al. 2005, Pecsenye et al. 2007, Sielezniew et al. 2012) then using mainly Myrmica schencki Viereck, 1903, My. sabuleti Meinert, 1861 and some other available Myrmica species (e.g. Elmes et al. 1998, Meyer-Hozak 2000, Stankiewicz et al. 2005. However, following the paper of Habeler (2008), we realised that no reliable host ant and initial food plant data have been published for the nominotypic Ma. rebeli (called simply Ma. rebeli hereafter).
The main reason for our study is that the type locality and the other suggested sites of Ma. rebeli are above the coniferous zone and well-separated by spruce forests (see Figure 7 in Habeler 2008) from the lower regions where Ma. alcon X occurs in grasslands (mostly extensive pastures) with G. cruciata. Furthermore, from our personal observations (see Materials and Methods), Ma. rebeli uses Gentianella rhaetica (A. Kern. & Jos. Kern.) Á. Löve & D. Löve rather than G. cruciata at higher altitudes, as the latter plant is not available in that zone (see the questions about host plant usage of Kudrna and Fric 2013). Similarly, we reasoned that the available Myrmica species are likely to be different for Ma. rebeli and Ma. alcon X within the same region because of changes in available niches with increasing altitude , Glaser 2006. It is therefore important to know the host ant use and specificity of Ma. rebeli and Ma. alcon X around the type locality of Ma. rebeli to help in answering the question of whether these two butterflies should be treated as the same or different forms or species.

Methods
Three sites within the Hochschwab area (Styrian Alps, Austria; Figs 1-4) were visited between the 22 nd and 24 th June 2012 (just before the flying period). Two Ma. rebeli habitats above the coniferous zone were investigated, one at Zeiritz (47. Ma. alcon X site at Sankt Ilgen within the coniferous zone (photo: AK).

3 4 2
Universalmuseum Joanneum; it belongs to the Gentianella germanica agg.: Greimler et al. 2011) on both sites, and caterpillars readily developed on and emerged from this plant (AK and AT, pers. observ.). The third site is a calcareous grazed meadow (Fig. 4) within the coniferous zone (Sankt Ilgen: 47.562N, 15.153E, ca 810 m a.s.l.) where Ma. alcon X uses Gentiana cruciata as initial host plant (an adjacent similar site for Ma. alcon X has already been described by Bereczki et al. 2005 andPecsenye et al. 2007).
The host ant specificity of the Ma. alcon X population at Sankt Ilgen was checked by searching for Myrmica spp. nests within 2 m of G. cruciata plants (see details: Tartally et al. 2008). However, this method was not possible in the case of the two Ma. rebeli sites (at Präbichl and Zeiritz) as Gentianella rhaetica plants are rather small before the flying period of Ma. rebeli, and were only rarely visible. We instead worked within patches where G. rhaetica plants were found in the previous year (although it should be noted that we were often not sure of the actual presence of plants in 2012, as G. rhaetica is biannual). When a Myrmica nest was found, it was opened as carefully as possible to search for Maculinea larvae and pupae. Five to ten workers were collected from each ant nest and preserved in 75% ethanol for identification in the laboratory (following keys in Seifert 1988, Radchenko and Elmes 2010, Czechowski et al. 2012. Host ant specificity was calculated based on the number of fully grown butterfly larvae, pupae and exuvia found (i.e., not including small larvae: see results) in two ways (see details: Tartally et al. 2008): p1 is the 2-tailed probability from a Fisher exact test of heterogeneity in infection of host ant nests (as implemented at http://www.quantitativeskills.com/sisa/), and p2 is the probability from a randomization test of ant nests between species (using the program MacSamp; Nash 2014). We should emphasize that our data on the host ant specificity of Ma. rebeli and Ma. alcon X should be compared cautiously, as our certainty of the presence of host plants differed for Ma. rebeli and Ma. alcon X (see above). Furthermore, excavating complete nests of Myrmica spp. was relatively easy in the deep humus of the Ma. alcon X site, but was much more difficult in the rocky Ma. rebeli sites.

Results
Altogether 53 Myrmica nests were found on the three sites ( be two-year developing larva after their first winter. There were 15 pupae in total on these two sites, nine of which (60%) proved to be infected with Ichneumon cf. eumerus Wesmael, 1857. Ten nests of My. sabuleti were infested at Sankt Ilgen with 53 Ma. alcon X larvae, pupae and exuvia, but only two caterpillars were so small as to be two-year developing larvae. There were nine "orphaned" Ma. alcon X pupae in the empty nest, which were dirty and showed visible signs of external mould. There were 36 pupae in total, 28 of which (77.8%) were infected with Ichneumon cf. eumerus.

Discussion
Based on the character of the habitat, the figures published by Habeler (2008) and photographs of Ma. rebeli imagos taken by AK (Figs 8-9), we are confident that we have re-discovered two populations that correspond to the description of Hirschke (1905). Our data confirm that Ma. rebeli is (at least) an ecological form found at high altitudes, and with host ant and initial food plant distinct from the adjacent populations of Ma. alcon X (Ma. rebeli auct. nec Hirschke). It is unlikely to be a mere individual form characterised "by the presence of a band of whitish or silver-greyish spots in the blue ground colour in the submarginal band adjacent to the marginal black line on the upper side of wings, particularly well pronounced on the hind wings and especially in the ♀" (see Kudrna and Fric 2013 page 117, which is a misinterpretation of the original German text of Hirschke (1905): 109-110: "Am Innenwinkel der Vorderflügel zeigt sich in Zelle 2 und 3 eine auffallende Aufhellung, welche sich auf die Hinterflügel fortsetzt und gegen den Innenwinkel derselben sogar die Form einer Antemarginalbinde gewinnt."). Furthermore, the specimens figured by Habeler (2008) also clearly show this character of Ma. rebeli but they are simply darker than the type series of Hirschke (1905) due to the circumstance that they were captured quite recently  while the type series was collected in 1904.
To avoid further confusion, we argue that the true Ma. rebeli occurs in the Hochschwab area above the coniferous zone (between c.a. 1500 and 2000 m) on slightly covered stony and sunny slopes with uncovered chalk rocks, where G. rhaetica is the only recorded initial host plant, and My. sulcinodis is the only recorded host ant species so far. We still need to confirm the observation of Habeler (2008) that these high altitude habitats are usually small, fragmented, and often located on steep and inaccessible (see Fig. 1) slopes, but it does seem that these populations have remained unknown to (or neglected by) most authors (see Kudrna and Fric 2013).
As far as we know, our observations are the first data on the host ant specificity of Ma. rebeli around the type locality, and hence, we maintain, of the host ant specificity of Ma. rebeli. Earlier data about the host ant specificity of Ma. 'rebeli' (e.g. Elmes et al. 1998, Meyer-Hozak 2000, Stankiewicz et al. 2005) should instead be considered as data about the host ant specificity of Ma. alcon X (see Introduction and Habeler 2008). While Ma. rebeli was found exclusively  in My. sulcinodis nests at Zeiritz and Präbichl, the works just cited mention Ma. alcon X generally from nests of Myrmica schencki or My. sabuleti and from some other locally available Myrmica species (My. lonae;My. scabrinodis Nylander, 1846;My. specioides Bondroit, 1918;My. rugulosa Nylander, 1849;and My. ruginodis). However, we cannot rule out the discovery of other host ants of Ma. rebeli in the Hochschwab region, following more intensive work. H.C. Wagner (see Kudrna and Fric 2013) also found Myrmica sulcinodis, My. ruginodis and My. lobulicornis as available Myrmica species on Ma. rebeli sites, plus one nest of My. scabrinodis, and also suggested My. lobicornis Nylander, 1846 as a potentially available but unrecorded species. Jutzeler (1989) has previously recorded My. sulcinodis interacting with a full-grown larva of Ma. 'rebeli' in the Swiss alps, although it was unclear whether this was a host nest or opportunistic predation (see also : Kudrna and Fric 2013). Although he did not record the altitude of this observation, it is notable that he also observed oviposition by Ma. 'rebeli' on Gentianella germanica (Willd.) E.F.Warb. on the same site (Jutzeler 1988). This region is distant and isolated by high mountains from the type locality of Ma. rebeli. Genetic studies and more knowledge about the host ant specificity of this Swiss population would be necessary to decide whether it is really Ma. rebeli (and e.g. not Ma. alcon X) and if its main host ant is really My. sulcinodis. Here it should be noted that specimens, phenotypically completely identical with the type specimens of Ma. rebeli, have also been collected from some high mountain habitats of Switzerland (Pontresina, 21. 07. 1907, leg. F.W. Dadd, coll. Zoological State Collection, Munich, checked by ZV).
Our results for Ma. alcon X, showing My. sabuleti as host, are in agreement with previous studies (e.g. Meyer-Hozak 2000, Tartally et al. 2008) that My. sabuleti is one of the most important host ants of this butterfly, and which has been already identified as the main host in other Austrian sites by Steiner et al. (2003).
The presence of I. cf. eumerus in Maculinea pupae appears to be a good indication that a given ant species is a suitable host ant of Maculinea butterflies (Thomas et al. 2002), so the relatively high infestation rate by the parasitoid is consistent with My. sabuleti being a suitable host of Ma. alcon X, and with My. sulcinodis being a suitable host of Ma. rebeli in the sites studied here.
The one empty Myrmica nest found to be infested with Ma. alcon X confirms that colonies of various Myrmica species can migrate (Radchenko and Elmes 2010). Therefore it could potentially happen that a host Myrmica colony moves out and leaves the Maculinea pupae in the nest and another Myrmica species occupies the empty nest, complete with Maculinea pupae, leading to an erroneously recorded 'secondary host ant' .
It is almost certain that Manica rubida Latreille, 1802 is not available as a potential host for the populations studied here, despite the presence of this ant in this region. Manica rubida is closely related genetically (Jansen and Savolainen 2010), and similar chemically (Errard et al. 2006), to Myrmica spp. and shows potential suitability as a host of the Ma. alcon group in the laboratory (Tartally 2004). This ant is common in the nearby sites of both butterflies, typically in the border of spruce forests, but was not found in the patches with Gentiana cruciata or Gentianella rhaetica. It also should be noted that despite the common presence of other gentian species (e.g. Gentiana asclepiadea L., see also: Kudrna and Fric 2013) in this region, Maculinea eggs were never observed on them.
The unusual host ant use of Ma. rebeli in the Alpine sites could be constrained by the distinctive composition of the Myrmica community found there. Although Maculinea populations can adapt to locally available Myrmica species (e.g. Nash et al. 2008, Tartally 2005, Tartally et al. 2013, the unique combination of host ant and host plant usage of these Ma. rebeli populations means that the nominotypic Ma. rebeli should be treated as a distinct Conservationally Significant Unit (see Maes et al. 2004), separated biologically and ecologically from both Ma. alcon X and Ma. alcon H. Further studies on the host ant and host plant specificity of Ma. rebeli and genetic comparison of these butterflies is necessary to qualify the degree of ecological and genetic relationships within the Ma. alcon group, and also to draw well-supported taxonomical conclusions on the status of the nominotypical Maculinea rebeli versus Ma. rebeli auct. (Ma. alcon X).

Conclusion
It should be concluded that Ma. rebeli does not simply represent an individual form of Ma. alcon (as considered by Kudrna and Fric 2013) but it should be considered as at least an ecological form adapted to high mountain conditions both in its initial food plant and host ant species. In addition, it should be emphasized that Ma. alcon X (= Ma. rebeli auct. nec Hirschke) cannot be synonymised with Ma. rebeli (Hirschke, 1904).