Changes in host ant communities of Alcon Blue butterflies in abandoned mountain hay meadows

Land use intensification is a general threat to biodiversity, but many species depend on low‐intensity agricultural ecosystems. One example is European mountain meadow ecosystems, traditionally managed by hay harvesting or livestock grazing. Abandoning management often causes population declines, local extinctions and biotic homogenisation in these meadows. We studied changes in the Myrmica host ant communities of the xerophilic form of the ant‐parasitic Alcon Blue butterfly (Maculinea alcon) in four hay meadows in the Bükk mountains of Hungary between 2000–2007 and 2018. Abandonment started in this region in the 1970s, accelerated in the 1980s, and culminated in the 1990s. We found that the Myrmica ant community has changed substantially in less than two decades. Diversity of the ant community always decreased, and species composition became more homogeneous at two sites. Habitat suitability for Maculinea butterflies decreased at three sites and increased at only one site, where management was restarted 20 years after abandonment. The number of M. alcon caterpillars and pupae, the rate of infestation of ant nests and the mean number of caterpillars and pupae per ant nest decreased between the two periods, whereas host ant specificity did not differ from random in either period. We conclude that the unfavourable changes in the host ant community due to abandonment have negative consequences for the persistence of Maculinea populations. Our study highlights the need for detailed monitoring, and the maintenance of low‐intensity management by mowing or grazing to avoid the decline of biodiversity dependent on low‐intensity agriculture.


Introduction
Land use change is one of the major drivers of biodiversity loss in Europe (Rounsevell et al., 2018). In most of Europe, intensification of land use, for example, increased inputs of chemicals in agriculture, is associated with negative impacts on the biodiversity of natural habitats. However, many species and habitats have adapted to long-term low-intensity farming, and farmland biodiversity has also declined considerably in the last few decades (Kleijn et al., 2009). Mountain hay meadows and pastures are prime examples of human-maintained habitat types in landscapes dominated by forests. These open habitats host unique plant and animal species that are characteristic of both lowland steppes and mountain meadows or grasslands, leading to the formation of highly diverse communities packed with species of biogeographic and conservation importance. For example, mountain hay meadows in Transylvania, Romania hold the world record for plant species richness in 0.1 m 2 and 10 m 2 areas (Wilson et al., 2012). These meadows have been maintained for centuries by regular human use, for example, for harvesting hay or livestock grazing, which effectively prevented the overgrowth of these open habitats by woody plants such as shrubs and trees (Vera, 2000). The cessation of traditional land use often results in succession starting with the appearance of woody plants, leading to scrub, then patchy woodland and eventually to closed forest (Baur et al., 2006). In central and eastern Europe, the number of grazing livestock started to decrease in the 1970s and 1980s and fell drastically in the 1990s due to socioeconomic changes and changes in agricultural practices (Prokopová et al., 2018), leading to large-scale abandonment of mountain meadows. As a result, the vegetation and invertebrate fauna of mountain hay meadows have changed considerably since the 1990s (Baur et al., 2006). The maintenance of the biodiversity of these unique, human-maintained habitat types thus requires active management to halt succession and to keep them in their highdiversity state (Stoate et al., 2009).
In many parts of Europe, Maculinea (= Phengaris) butterflies (Lepidoptera: Lycaenidae) are flagship species due to their complex life cycle and sensitivity to the status of their habitats (Settele et al., 2005). Mountain meadows and semi-dry grasslands in central Europe are often inhabited by the xerophilous form of the Alcon Blue butterfly (Maculinea alcon [Denis & Schiffermüller], 1775), which we will refer to as M. alcon X (Tartally et al., 2019). This butterfly has an unusual life cycle in which the caterpillars start their development by feeding on the developing seeds of gentian plants. In the final instar, the caterpillars are taken into their nests by Myrmica (Hymenoptera: Formicidae) ants, and are raised together with the host brood (Thomas et al., 1989) for 1 or 2 years (Thomas et al., 1998a). The caterpillars mimic the odour Thomas et al., 2013) and the sound ) of the ants and are mostly fed by the Myrmica workers by trophallaxis , although they also prey directly on the ants' brood (Tartally, 2004). A multitude of studies suggest that different Alcon Blue populations have adapted to different Myrmica species as their local hosts (Tartally et al., 2019). There are both multiple-host-ant using populations, typically in central Europe, and more host-specific populations, typically in western Europe (Tartally et al., 2019). The hygrophilic form of M. alcon (see 'Study sites and populations' below) uses mainly M. rubra and M. ruginodis in the north-west of Europe, and mainly M. scabrinodis and its close relatives in the remainder of Europe. M. alcon X generally exploits M. sabuleti or M. schencki and has a more diverse host ant usage in central Europe than in the western or northern periphery of its range (Tartally et al., 2019). The M. alcon form rebeli has been recorded only from M. sulcinodis (Tartally et al., 2014). Not all the available Myrmica species are suitable hosts of a given M. alcon population (Thomas et al., 1989). Some Myrmica species, for example, usually kill the caterpillars after adoption, especially under food stress Thomas et al., 2005). This life cycle makes Maculinea butterflies highly sensitive to changes in the local Myrmica community. The Myrmica communities in turn are sensitive to changes in their microhabitats . Therefore, successional changes and/or inappropriate management of a site can cause the decline of the local host ant species, which in turn can result in the local extinction of the Maculinea population (Thomas, 1980;Thomas et al., 2009). For these reasons, Maculinea butterflies are widely recognised as indicator and umbrella species of their habitats (Settele et al., 2005), and are protected in most countries of Europe (Munguira & Martín, 1999).
Understanding the links between land use or management and Myrmica ant communities is thus of fundamental importance in the conservation of species-rich mountain hay meadows and pastures, as well as the persistence of Maculinea butterflies. In addition, study of the host ant specificity of Maculinea butterflies also has evolutionary relevance. Following the pioneering work of Thomas et al. (1989), we have obtained extensive information on the host ant use of Maculinea butterflies in different regions of Europe (Tartally et al., 2019), largely as a result of the MacMan collaborative research programme (Settele et al., 2002). These insights (Tartally et al., 2019) aptly demonstrate the spatial dimension of the geographic mosaic of coevolution hypothesis (Thompson, 1999) in the Maculinea-Myrmica coevolved system. However, to our knowledge, the temporal dimension, that is, the dynamics of this system in time has not yet been sufficiently addressed.
The aim of this study was to evaluate whether and how changes in land use and management influence Myrmica ant communities and their infestation by M. alcon X caterpillars and pupae in four meadows in the Bükk-mountains of northern Hungary. Because many of the former mountain hay meadows and pastures have already disappeared due to land abandonment, lack of management and shrub encroachment, it was not possible to obtain an adequate sample size to systematically examine different management histories. Instead, we use the four study sites as case studies of different management histories. Traditional land use was abandoned at different times at these four sites and at least one 10-year period of no management occurred on each of them since the 1970s (see Materials and Methods section, Table 1). A previous publication (Tartally et al., 2008)  reported on our studies of Myrmica communities and Maculinea host ant use carried out between 2000 and 2007 at these four sites and several others. In 2018, we resampled these four sites and evaluated the impact of habitat management and vegetation succession on host ant communities, Maculinea numbers and infestation rates, and host ant specificity. We compare these results to those of our previous study (Tartally et al., 2008) and develop recommendations for habitat management to increase the chance of persistence of Maculinea butterfly populations.

Study sites and populations
Current literature (Tartally et al., 2019) distinguishes three forms of M. alcon, the hygrophilic (M. alcon H), the xerophilic (M. alcon X) and the Alpine (M. alcon form rebeli). For a long period M. alcon X was widely referred to as 'M. rebeli'; however, this assertion has been shown to be misleading (Habeler, 2008;Kudrna & Fric, 2013;Tartally et al., 2014;Koubínová et al., 2017;Bereczki et al., 2018). In this study, we focused exclusively on habitats of M. alcon X. This form occurs in drier or mesic areas, often at intermediate and higher elevations, and mostly uses Gentiana cruciata L. as a host plant for egg-laying and early larval development.
In all four sites, we selected patches with high G. cruciata density for detailed examination. Patches were selected haphazardly from across the site, but were chosen so as to also leave other nearby patches with high G. cruciata density undisturbed. When other host plants were found within a patch (i.e., Gentiana pneumonanthe L. and Gentianella austriaca (A. & J. Kern) Holub, see below), that patch was not sampled.
This small meadow, surrounded by sparsely populated residential areas was regularly mowed and grazed by cattle from at least 1970. Combined land use by early summer mowing and late summer cattle-grazing continued longest at this site, and the meadow was abandoned in the late 1990s. The vegetation of the site had very few woody plants before abandonment, and a substantial part of the site has been overgrown by shrubs since. The area of the selected patches at this site was 3161 m 2 in total.
This large meadow was regularly mowed at least from 1960, but this management ceased at the end of the 1970s. The meadow was subsequently mowed a single time around 2010 by Bükk National Park, making this the site with the longest period without management. This site still had rather open vegetation in 2018, presumably due to the grazing of wild ungulates. The area of the selected patches at this site was 7373 m 2 . This large meadow was regularly mown by farmers for hay and was also regularly grazed since at least 1970. The meadow was abandoned in the early 1990s. Management was re-started in 2010 by regular annual mowing and low-intensity cattle-grazing.
Gentianella austriaca is an additional host plant of the local M. alcon X population. The area of the selected patches at this site was 6189 m 2 .
This large meadow was regularly mowed at least from 1960. At the end of the 1970s, management by horse-grazing was started at the site. Horse-grazing was terminated around 2010 and the site has not been managed since, making this site the most recently abandoned. Gentiana pneumonanthe is an additional host plant of the local M. alcon X population. This population does not have clearly separated flying periods associated with the different flowering periods of the two host plants (János Tóth & Tamás Korompai, pers. comm.), unlike the Romanian site R ascruci. At that site there is strong genetic differentiation between the xerophilous form (M. alcon X) using Gentiana cruciata and the hygrophilous form (M. alcon H) developing on the G. pneumonanthe and the plants (Tartally et al., 2016). The area of the selected patches at this site was 5330 m 2 .

Sampling
To count Maculinea alcon X, we searched all Myrmica nests less than 2 m distant from every stalk of G. cruciata in each selected patch for caterpillars, pupae and pupal exuviae. This area corresponds to the approximate space where foragers from Myrmica nests are likely to encounter caterpillars of M. alcon X emerging from the host plants . Ant nests were examined in the early summer, immediately before the flight period of each population, to ensure that any M. alcon X caterpillars and pupae found had survived the winter, most likely as an integrated part of the colony investigated . As no 2-year developing larvae (Schönrogge et al., 2000;Witek et al., 2006) are known from Hungary (Tartally, 2005), and all larvae found in this study were of similar size, we did not exclude any larvae from the analyses.
We recorded Myrmica nests, opened them carefully and counted the number of M. alcon X caterpillars and pupae. Five to 10 Myrmica workers per colony were collected and stored in 70% ethanol. We identified worker ants to species in the laboratory [according to the keys of Seifert (1988), Radchenko and Elmes (2010), Czechowski et al. (2012)], using a Leica MZ125 microscope (magnification 10-160×).
Sampling was carried out every year on each site during the 2000-2007 period, but different patches were sampled each year, and the data of the different patches and years were pooled for analysis. All the patches investigated during 2000-2007 were re-sampled in 2018, so the exact areas surveyed were comparable between the two periods.

Data analysis
The Myrmica communities (proportion of nests of different Myrmica species) were compared between the two collection periods at each site using a Fisher's exact test. We used the R 'stats' package for calculating Fisher's exact tests for contingency tables with two columns (periods) and N rows, where N was the number of species. We used Shannon's diversity index calculated using PAST v. 3.20 (Hammer et al., 2001) to characterise Myrmica species diversity in the two periods. We visualised similarity of and changes in Myrmica community with multidimensional scaling (MDS) ordination in JMP 14.3 (SAS institute, 2019) using a matrix of pairwise Bray-Curtis dissimilarities between Myrmica communities calculated using PAST and based on the relative abundance of each Myrmica species.
We characterised the suitability of the ant community for the Maculinea population at a site using the following Myrmica community suitability index (MCSI):

MCSI =
N Maculinea caterpillars and pupae N Myrmica nests: where N Maculinea caterpillars and pupae was the number of all Maculinea caterpillars and pupae recorded in all the examined Myrmica nests at a site in a study period, and N Myrmica nests was the number of the examined Myrmica nests. This index gives the average number of M. alcon X caterpillars and pupae per Myrmica nest, and as such we assume that the higher the value of the index, the more suitable is the Myrmica community for the Maculinea population. It is important to note that this simple index does not take into account any host ant specificity, so is not generally applicable to all Maculinea sites. Our previous study, however, found little or no host ant specificity in the studied populations (Tartally et al., 2008). For number of caterpillars and pupae and infestation rates, although the distribution of most variables departed from normality, we present means AE SDs for comparability with previous studies. We used the median test to statistically compare medians of distributions.
We calculated host ant specificity, that is, the deviation from a random occurrence of Maculinea caterpillars and pupae in nests of different Myrmica species in two ways: (i) using Fisher's exact tests of heterogeneity of whether nests of different Myrmica species are infected by M. alcon X or not, and (ii) by estimating the likelihood that the numbers of M. alcon X caterpillars and pupae would be distributed as heterogeneously as observed between nests of different Myrmica species by constrained randomisation of ant nests between species using the software MacSamp (Tartally et al., 2008). The latter takes into account the number of caterpillars and pupae found within nests as well as whether nests are infested or not.
Distinguishing between the closely related M. sabuleti Meinert, 1861 and M. lonae Finzi, 1926 can be problematic (Seifert, 2000;Czechowski et al., 2012). These two similar ants are discussed here as separate species, but their relationship needs clarification (Radchenko & Elmes, 2010;Seifert, 2018). Sharing of mitochondrial haplotypes suggests that they may not be separate species (Ebsen et al., 2019), as does their similarity in cuticular hydrocarbons (Guillem, 2014). Because of this taxonomic uncertainty and the difficulty of identification of these two ants, we repeated all analyses assuming that M. lonae nests identified in 2000-2007 were actually M. sabuleti nests. We note differences only if the results of this repeated analysis differed from the original analysis.

Ant species and community composition
In the two periods combined, we examined 625 nests of six Myrmica species (M. scabrinodis Nylander, 1846; M. sabuleti;  (Table 2), but there was no consistent change in ant nest density between periods, as more nests were found in the first period at two sites and at the second period in the other two. Myrmica scabrinodis was the most abundant species at three sites (Bükkszentkereszt, Lóf} o-tisztás, Nagy-mez} o) and M. sabuleti Table 3. Indices and probabilities for comparison of sites between survey periods. 'P same' is the two-tailed probability from Fisher's exact tests comparing the distribution of ant nests amongst Myrmica species between the two periods. P1 is the two-tailed probability from Fisher's exact tests of heterogeneity in infestation of host ant nests. P2 is the probability from randomisation tests of ant nests between species (using MacSamp  The distribution of ant nests amongst Myrmica species differed between the two periods at three of the four sites (Table 3). When M. lonae individuals found in 2000-2007 were considered as M. sabuleti, this difference remained significant for Nagy-mez} o (P < 0.001) and became marginally non-significant for Kecskeláb-rét (P = 0.096) and also for Lóf} o-tisztás (P = 0.052), where these proportions did not differ in the first analysis (Table 3).
The proportion of ant species considered as hosts (Tartally et al., 2008) (Table 3). When data from all sites were combined, MCSI decreased from 0.399 in 2000-2007 to 0.198 in 2018. MDS ordination suggested that species composition changed in different ways at the sites between 2000-2007 and 2018 ( Fig. 1; Table 2). Communities in Lóf} o-tisztás and Nagy-mez} o became more similar in 2018 because the number of species decreased from five to three and six to four respectively, and both communities became dominated by M. scabrinodis. Both of these communities converged on that found at Bükkszentkereszt, where M. scabrinodis was already dominant in 2000-2007. However, the total number of Myrmica species found at this latter site increased from two in 2000-2007 to four in 2018. Finally, the ant community at Kecskeláb-rét was dominated by M. sabuleti in both periods and changed in parallel with the other three sites because two of the species found in 2000-2007 were not found in 2018 and one species not found in 2000-2007 appeared at the site in 2018. These patterns of change were largely consistent regardless of whether M. lonae was treated as a separate species or combined with M. sabuleti (Fig. 1).

Larval numbers and infestation rates
In 2000-2007, we found 147 M. alcon X caterpillars and pupae in 18 of the 368 ant nests studied, whilst in 2018, we found 51 caterpillars and pupae in 11 out of the 257 nests. The decrease in the total number of M. alcon X caterpillars and pupae (ca. 65%) was much larger than the decrease in the number of nests found (30%), indicating that the decrease in the number of caterpillars and pupae could not be attributed only to differences in sampling effort.

Host ant specificity
Extended Fisher's exact tests showed no evidence of host ant specificity in infection rates at any site in either period, although that for Bükkszentkereszt for 2018 was close to statistical significance (P = 0.063, Table 3). Randomisations of ant nests between species by MacSamp also did not show any specificity of numbers of M. alcon X caterpillars and pupae to ant species (Table 3). Taken together, these results are consistent with our previous finding of no heterogeneity in host ant use at these sites (Tartally et al., 2008).

Discussion
Our study provides two key findings. First, several results showed that the Myrmica ant community changed slightly between 2000-2007 and 2018. Shannon diversity, infestation rate, and the mean number of caterpillars and pupae per nest decreased despite the increasing relative abundance of host ant species. Second, the changes in relative abundance and species composition differed amongst the sites, for example, MSCI decreased at three sites and increased only at Lóf} o-tisztás (Table 3). This is the only site where management was restarted after two decades of abandonment (Table 1).
There are some notable changes in the structure of the Myrmica communities, such as variation in dominance, species loss or substitution. It is possible that some of these changes are driven by climate changes as well as changes in management, although the pattern seen here is not consistent, as both warmloving (M. schencki) and more cold-tolerant (M. ruginodis) species  were only recorded in the second survey at some sites. The apparent absence of M. lonae in 2018 is intriguing, as the quantitative differences used to distinguish this species from M. sabuleti (Seifert, 2000) were clearly present in 2000-2007, but could not be found in 2018. Whether this represents true species loss or a temporal change in a plastic phenotype remains to be clarified. The continued or increasing dominance of M. scabrinodis on three of the sites is noteworthy, but is difficult to interpret as there seem to be at least two divergent lineages within this species that differ in their habitat requirements (Ebsen et al., 2019) and may also differ in their suitability as hosts of Maculinea butterflies (Tartally et al., 2019).
In general, results from our sites with little or no management suggest negative trends in Maculinea populations in the last decade. In Bükkszentkereszt and Kecskeláb-rét, the MCSI was an order of magnitude lower two decades after abandonment than only a decade after. In Nagy-mez} o, MCSI was already low in 2000-2007 (only two infested ant nests) and it decreased to zero (no infested nests) by 2018. This result suggests that lowintensity horse-grazing was not successful in maintaining host ant and Maculinea populations, and that 9 years after abandonment the Maculinea population had likely gone locally extinct. This was surprising because a conspicuously abundant flight of M. alcon X was observed in the summer of 2012 (Table 1), in a year when this site was ravaged by a wildfire in early spring. This observation suggested that fires in the early spring do not damage overwintering, still inactive populations of Myrmica and Maculinea species, a finding also reported by Nowicki et al. (2009). Although Myrmica ants are sensitive to changes in the cover of vegetation (Grill et al., 2008), fire can even have beneficial effects on Myrmica and indirectly on Maculinea species by removing dead plant material and changing the microstructure of the vegetation. The potential use of fire as a management tool (prescribed burning) should be explored in further studies.
There was no evidence for host ant specificity for M. alcon X on the four sites examined, despite reasonable sample sizes of Myrmica nests of different species, which often limits the ability to detect heterogeneity in host use (Tartally et al., 2019). This may also reflect the particular Myrmica community composition on these sites, as although there were large changes in the relative abundance of different species, the proportion of species that had been recorded as local hosts of M. alcon X remained high. Changes in Myrmica ant community are likely to have an even larger effect if the proportion of host species changes, which will be more likely for sites where there is more host any specificity, as is generally the case for other Maculinea such as M. arion, M. nausithous and even M. alcon H (Tartally et al., 2019). There was also no clear relationship between Myrmica nest density on the four sites that we examined and the presence and number of M. alcon X caterpillars and pupae. This suggests that it is not simply the density of potential host ants that is important for the maintenance of M. alcon X populations, but other factors associated with management. These could be related to Myrmica diversity per se, perhaps through the selection pressures imposed by the geographic mosaic of coevolution (Thompson, 1999;Tartally et al., 2019), or to other factors that we did not measure, such as the size of host ant nests or changes in host plant populations (Mouquet et al., 2005).
In summary, our results suggest that abandonment, that is, the termination of management, leads to changes that are unfavourable to the xerophilous form of M. alcon populations in the Bükk Mountains. The same trend was clear in the nearby Aggtelek region, but was successfully reversed by resuming management ( Arnyas et al., 2006). It should be noted that caution is warranted in the interpretation of our results because the management history of the study sites was sketchy and more details on, for example, livestock grazing pressure or frequency of mowing could have further refined the trends found. However, such detailed information was not available for the sites. Moreover, caution is also warranted because data referred to as '2000-2007' were collected during eight summers (Tartally et al., 2008), whereas data from 2018 were collected in a single summer. The size of Maculinea populations tends to be quite variable between years, often showing something approaching cyclic changes in the adult population (see e.g. Nowicki et al., 2009), so the decrease in the number of larvae in one single monitoring event (2018) could be due to a low adult population the year before that is not linked to environmental causes. As no information was available about the adult and/or egg densities in the previous years, we cannot conclude that the background of this decrease of M. alcon larvae by 2018 is entirely the change of the Myrmica community. All of these uncertainties emphasise the importance of more detailed and long-term monitoring of Maculinea populations in the future.
The patterns that we observed suggest that management of mountain grasslands (former hay meadows) is crucial for maintaining rich and diverse communities of Myrmica ants, which may also be crucial for M. alcon X populations. The observed changes in Myrmica communities combined with what is known of the histories of the sites suggest that the best way to manage these habitats is a combination of regular low-intensity mowing and low-intensity grazing. Such management has been started recently at Lóf} o-tisztás, and changes in the ant community and Maculinea numbers attest that such management can lead to favourable changes in the habitat. Further monitoring is necessary to establish the exact mechanisms of these favourable changes. It is, however, important to note that management proven successful in one type of habitat or population may not necessarily be successful for other habitats or populations. For example, Maculinea populations are also affected by large-scale factors such as latitude, altitude and climate (Thomas et al., 1998b).
Our study shows that it is important to conduct long-time monitoring of the effects of site management on host ant use of Maculinea butterflies. Because land abandonment occurs on large spatial scales in central and eastern Europe, such monitoring should be extended to as many Maculinea populations as possible to compare these effects and to design conservation actions for Maculinea butterflies and other species of mountain hay meadows.

Conclusion
We conclude that abandonment, that is, termination of regular management, of mountain hay meadows presents significant risks to the persistence of Maculinea populations. Re-starting management by low-intensity mowing and low-intensity grazing can increase the suitability of habitats to Maculinea butterflies and their host ants. Long-time monitoring of the effects of site management is important for the successful protection of Maculinea populations. on the manuscript during the revision process. Research was funded by the European Commission within the RTD project 'MacMan' (Settele et al., 2002) and also by the EFOP-3.6.1-16-2016-00022 project. AT was supported by a János Bolyai Scholarship of the Hungarian Academy of Sciences and by the ÚNKP-18-4 New National Excellence Program of the Ministry of Human Capacities. SL was funded by two grants from the National Research, Development and Innovation Office of Hungary (NKFIH-OTKA K106133, GINOP 2.3.3-15-2016-00019).