Introduction

Myrmica Latreille 1804 ant colonies are notoriously prone to invasion by both socially parasitic ants (reviewed by Bourke and Franks, 1995) and social parasites from other insect taxa including Diptera, Coleoptera and Lepidoptera (e.g. Donisthorpe, 1927; Hölldobler and Wilson, 1990; Dettner and Liepert, 1994; Thomas and Settele, 2004; Settele and Kühn, 2009). Highly specific recognition mechanisms are essential for the evolution of the intricate interactions that form the basis of stable host–parasite systems in ants (Nash and Boomsma, 2008), particularly for butterflies of the genus Phengaris Doherty 1891, which are nest parasites of several Myrmica ant species. Females of Phengaris lay their eggs on particular food plants, on which the larvae feed and complete their development up to the fourth larval instar. After moulting into the fourth larval instar, the caterpillar leaves its food plant and moves to the ground, where it depends upon being adopted by foraging workers of a suitable Myrmica species, which will carry it into the host colony. Once in the colony the caterpillar will complete its development, pupate and leave the nest as an imago. Therefore, adoption and integration of the social parasite in the host colony are crucial moments, which can be regarded as a series of quite dynamic filters (encounter, infection and exploitation) that can determine local host specificity (Nash et al., 2011). Although signals such as size and tactility (Elmes et al., 2001), behaviour and sound (De Vries et al., 1993; Barbero et al., 2009a; Barbero et al., 2009b) apparently play a role in the initial adoption of Phengaris larvae, it seems that chemical signals are at the centre of the adoption process. Because chemical signals have been shown to be essential for nest-mate recognition in ants (e.g. Hölldobler and Wilson, 1990), Elmes et al. (1991) suggested that this is also the main mechanism for Phengaris adoption, based on chemical signals on the caterpillar body surface.

Variability in chemical signals is common in ant societies (Whitehouse and Jaffé, 1995) and cuticular hydrocarbons have been suggested to play an important role in nest-mate recognition (Lahav et al., 1999; Van Zweden and d’Ettorre, 2009). Nest-mate recognition in ants reflects the ability of workers to discriminate conspecific members from other colonies (Vander Meer and Morel, 1998; Van Zweden and d’Ettorre, 2009; Sturgis and Gordon, 2012).

The ability to recognize nest-mates is highly adaptive and particularly important to prevent intrusion of foreign conspecifics and robbing of the colony (Wilson, 1971; Stuart and Herbers, 2000). Colony brood recognition has been demonstrated in several ant species (Brian, 1975; Meudec, 1978; Isingrini et al., 1985; Hare and Alloway, 1987; Carlin and Schwartz, 1989; Bonavita-Cougourdan et al., 1989; Fénéron and Jaisson, 1992, 1995). Parental care of brood is a key factor in the social life of colonies as workers are able to assess the needs of the offspring such as grooming and feeding (Camargo et al., 2006). Concolonial brood is accepted and nourished by workers, whereas allocolonial brood may be rejected (Bonavita-Cougourdan et al., 1989). The brood has chemical cues that enable workers to recognize them as concolonial (Brian, 1975; Araújo et al., 1996; Viana et al., 2001). Nevertheless, ant workers are often more tolerant towards unfamiliar brood (and sometimes even heterospecific brood) than unfamiliar adults and a possible explanation is that they will become integrated into the colony workforce and also increase colony fitness (Haskins and Haskins, 1950; Elmes and Wardlaw, 1983; Errard, 1984; Plateaux, 1985; Goodloe and Topoff, 1987; Fouks et al., 2011). Adoption of unfamiliar brood does not, however, preclude the preference for nest-mate brood, as revealed in choice experiments (Fénéron and Jaisson, 1995).

In contrast, retrieving Phengaris larvae to the colony reduces colony fitness, setting the stage for co-evolutionary arms races between host recognition and parasite attractiveness. The high specificity in host and parasite signals may result in local adaptations similar to those reported for Microdon mutabilis Linnaeus 1758 (Schönrogge et al., 2006) and P. alcon Denis & Schiffermüller 1775 in their interaction with M. rubra Linnaeus 1758 (Als et al., 2001; Als et al., 2002; Nash et al., 2008). Further, new local hosts for P. alcon in Portugal (Arnaldo et al., 2011) and for P. teleius Bergsträsser 1779 in Poland (Witek et al., 2010), suggest host specificity is modulated by local Myrmica species availability.

To be adopted by ants, Phengaris larvae mimic ant brood (Thomas and Settele, 2004). However, the efficiency of and local adaptation in adoption of Phengaris larvae has not been compared to that of host larvae. In this study, we use P. nausithous and its host M. rubra to screen for local adaptations by analysing the workers’ adoption behaviour towards the butterfly larvae. We chose P. nausithous as a study species, because it shows high host specificity for M. rubra across Europe (Thomas et al., 1989; Elmes et al., 1998; Stankiewicz and Sielezniew, 2002; Tartally and Varga, 2005; Witek et al., 2008) in spite of occasional occurrences of P. nausithous larvae in M. scabrinodis (Elmes et al., 1998; Munguria and Martin, 1999; Witek et al., 2008) and M. ruginodis nests. Because P. nausithous has much higher initial survival in M. rubra nests than in those of any other host species (Patricelli et al., 2010), the mimetic efficiency should be exceptionally high among Phengaris butterflies. Other Phengaris species are more variable in their host use, and often it is not possible at all to identify a primary host especially for Phengaris predator feeders (Witek et al., 2008). This makes it difficult at best to screen for a signature of local co-evolution between host and parasite.

In this study, we compared a M. rubra population (called sympatric population) coexisting with P. nausithous with another M. rubra population (called allopatric) not exposed to P. nausithous to find out how the parasite affects the brood recognition and the acceptance in M. rubra ants. In case of local adaptation to the parasite, sympatric and allopatric workers should present different behavioural responses towards P. nausithous similar to those reported for P. alcon (Als et al., 2001; Nash et al., 2008) depending on who is leading the arms race: host or parasite. Therefore, our questions are: (i) do Myrmica workers recognize nest-mate larvae? (ii) do workers prefer ant larvae over P. nausithous larvae at colony and individual level? And (iii) are there differences in P. nausithous larvae choice between Myrmica ants from sympatric or allopatric parasite populations?

Materials and methods

We quantify the encounter behaviour of M. rubra workers with larvae from (i) their own nest, (ii) conspecific foreign nest or (iii) the parasite (P. nausithous) and we compare the retrieval of own brood versus parasite larvae into the M. rubra colonies.

Sampling and handling of ants

The study was conducted with 40 colonies of M. rubra collected from two different sites in Germany, 30 colonies from Halle (allopatric population, N51°30′E11°56′) and 10 colonies from Altenburg (sympatric population, N51°01′E12°28′). We collected the ant colonies from the wild, transferred them to artificial nests (gypsum nest 20 × 9.5 × 3 cm at 20 °C). All colonies consisted of 50 workers, a queen and some brood (≈15). Colonies were fed in a foraging arena attached to the nest with dry adult Drosophila and diluted honey (≈5 g in 40 ml water) in a paper ball placed in a small plastic dish (14 mm diameter) and water ad libitum. We changed water and food twice a week for all colonies. The ant species were identified using the morphometric key of Czechowsky et al. (2002).

Sampling and handling of butterfly larvae

We collected P. nausithous larvae from flower heads of Sanguisorba officinalis L. (food plant of P. nausithous) in Altenburg (N51°01′E12°28′). In the laboratory, we kept flower heads in a petri dish at room temperature until the fourth instar butterfly larvae emerged.

Recognition behaviour experiment

Behavioural experiments were conducted between April 2010 and September 2010 at the Department of Molecular Ecology at the Martin Luther University Halle Wittenberg. For these experiments we used 20 colonies of M. rubra from Halle. We performed three kinds of experiments, all based on behavioural observations in a Petri dish arena (7 cm diameter, 1 cm height), in which we presented one or two M. rubra larvae (nest-mate or foreign) in a small plastic dish (14 mm diameter). We observed worker behaviour under a binocular microscope with 32× magnification.

For each test, one worker was introduced into the arena to observe the acceptance of a M. rubra worker without nest-mate or nest material (which can affect ant tolerance towards foreign individuals; Buczkowski and Silverman, 2005).

Behavioural observations started 10 min after introduction into the arena in order to allow the test worker time to accommodate to the new environment. The trial consisted in an experimental encounter between an ant larva with a nest-mate or with a foreign worker in a fixed observation period. For each experiment we recorded the number of times a selected behaviour was observed (Table 1—modified from Human and Gordon, 1999). Every arena was carefully cleaned with ethanol before each trial.

Table 1 Categories and descriptions of recorded behaviour of Myrmica workers (after Human and Gordon 1999)
Full size table

We made 50 replicate tests for each experimental setup (see below), always using a different test worker to avoid habituation of the workers to the ant larvae. The ant larvae were used in five sequential tests, in total 20 M. rubra larvae (10 nest-mate and 10 foreign) were assayed in experimental setup 1. In experimental setup 2, 50 nest-mate M. rubra and 50 foreign M. rubra were used.

Experimental setup

  1. 1.

    Individual worker behaviour towards a single larva.

    We observed the behaviour of Myrmica workers either towards conspecific nest-mate or foreign larvae for 3 min to assess the worker behavioural repertoire (Fig. 1a).

    Fig. 1
    figure 1

    Experimental set up for: a individual worker behaviour towards a single larva; b individual worker behaviour towards nest-mate and foreign larvae in a simultaneous choice setting and choice behaviour of individual M. rubra workers between ant and butterfly larvae; c local host–parasite adaptation—adoption of P. nausithous larvae by M. rubra colonies from the sympatric and allopatric populations; d Local host–parasite adaptation—choice behaviour between ant and butterfly larvae by M. rubra colonies from the sympatric and allopatric populations

    Full size image
  2. 2.

    Individual worker behaviour towards nest-mate and foreign larvae in a simultaneous choice setting.

    By modifying experimental setup 1, we tested the preference of a M. rubra worker when given the choice between a nest-mate and a foreign conspecific larva. The test worker was observed for 10 min (Fig. 1b) to reveal any behavioural differences towards conspecific nest-mate and foreign brood.

Both experiments 1 and 2 were necessary to establish the feasibility of the arena bioassay to quantify the recognition of P. nausithous larva by M. rubra workers.

Choice behaviour of individual Myrmica workers between ant and butterfly larvae

We used a dual choice experiment to assess the discrimination of M. rubra workers between M. rubra and P. nausithous larvae (Fig. 1b). We used 20 randomly chosen workers (2 from 10 colonies each) from both the sympatric population of Altenburg and from the allopatric population of Halle. We simultaneously placed a M. rubra, a M. rubra larva, and a P. nausithous larva into the arena. Based upon the behavioural repertoire from the previous Myrmica recognition experiments, we performed our observations for 10 min. M. rubra workers and larvae of both P. nausithous and M. rubra were only used once to avoid habituation of the workers and chemical contamination of larva body surface via contact of M. rubra workers. Both the arena and plastic dish were cleaned with ethanol between the trials (20 trials per population).

Local host–parasite adaptation

We tested for local host–parasite adaptation by comparing worker recognition and adoption behaviour of ten M. rubra colonies each from Halle (allopatric with Phengaris) and from Altenburg (sympatric with Phengaris) using complete nests composed of 50 workers, a single queen and ≈15 larvae. M. rubra and P. nausithous larvae were used only once for each trial, to prevent potential effects of compound contamination resulting from contacts with M. rubra workers.

All interactions were recorded for 1 h with a digital camera (1,280 × 1,024 pixel) and the video sequences analysed with VirtualDub© (Lee 1998–2009). Since the maximum recorded adoption time for P. teleius was 20 h and for P. alcon was 23 h for the host and 47 h for the non-host, our observation period was arguably short (Als et al. 2001, Witek et al. 2011). Nevertheless, the chosen time window was sufficient to compare P. nausithous larvae adoption time with the much more quickly adopted M. rubra larvae, thereby reducing experimental artefacts resulting from extensive lab exposure.

  1. 1.

    Adoption of Phengaris

We screened for differential adoption behaviour of M. rubra workers coming from the sympatric or the allopatric population (Fig. 1c). We determined the number of trials with P. nausithous adoption, and larval survival within 1 h after adoption. We tested 10 colonies from each population with a total of 23 and 36 P. nausithous larvae, respectively.

  1. 2.

    Choice behaviour of colonies

For this second experiment we presented both a nest-mate and a P. nausithous larva simultaneously to the colony (Fig. 1d). Again we recorded the interactions between ants and both kind of larvae and noted the number of times and duration of adoption to assess possible differences between ant populations. We tested 10 colonies from the sympatric and 10 colonies from allopatric populations, respectively, with 40 P. nausithous/M. rubra larvae pairs. The number of adoptions were analysed with the χ 2 test and the means of adoption time with Mann–Whitney U test (STATISTICA© version 8).

Results

Recognition behaviour of individual ants

The behavioural repertoire of ants were identical in the experiments 1 and 2 and we grouped the behavioural traits into two categories: “inspection” (antennation, larva carrying and taking by mandibles) or “aggression” (biting and gaster contact). M. rubra workers show a different behavioural approach with the different kind of larvae (nest-mate, foreign and social parasite) resulting in statistically significant differences (χ 2 = 26.30, df = 2, P < 0.001, N = 127 nest-mate M. rubra, N = 40 P. nausithous and 130 foreign M. rubra larvae) (Fig. 2).

Fig. 2
figure 2

Behavioural trials of M. rubra workers towards: nest-mate M. rubra larva (white bar), P. nausithous larva (black bar) and foreign M. rubra larva (grey bar). Percentage (y-axis) and number of trials (above the bars) are given for each behavioural group; ***difference statistically significant; (χ2 = 26.30, df = 2, P < 0.001, N = 128 nest-mate M. rubra N = 40 P. nausithous and 130 foreign M. rubra larvae)

Full size image

Choice behaviour of individual ants

Antennation was the only behaviour of ants observed towards P. nausithous larvae. Workers from the allopatric population did not contact the social parasite in one out of 20 trials performed, in comparison the sympatric population did not contact P. nausithous 8 times out of 20. The sympatric population antennated mainly towards their own brood. The difference between populations in antennation behaviour displayed is statistically significant (χ 2 = 10.04, df = 2, P = 0.01, N = 40) (Fig. 3).

Fig. 3
figure 3

Choice behaviour of individual Myrmica workers between ant and butterfly larvae. Percentage (y-axis) and number of trials (above the bars) for M. rubra larva (white bar), P. nausithous larva (black bar) and both larvae (grey bar) relatively to the difference (***difference statistically significant) in antennation between the populations (χ2 = 10.04, df = 2, P = 0.01, N = 40)

Full size image

Colony-level experiments

Adoption of P. nausithous

The social parasite P. nausithous was adopted 10 times out of 23 trials by workers from the sympatric population. Hence, non-adoption was not significantly different from adoption (χ 2 = 0.39, df = 1, P = 0.53, N = 23; Fig. 4). In contrast, workers from the allopatric population adopted significantly less (10) times and left 26 larvae in the foraging area (χ 2 = 7.11, df = 1, P = 0.01, N = 36; Fig. 4). After adoption into the colony not all larvae survived. The rate of survival was similar in colonies of both populations. In the allopatric colonies 7 (out of 10) larvae survived after the adoption and 6 (out of 10) in the sympatric colonies (χ 2 = 0.22, df = 1, P = 0.64, N = 20; Fig. 4). The observed “adoption behaviour” is therefore not a guarantee for survival of P. nausithous in either host population because the ants kill the larvae once in the nest in this case at a rate of 40 %.

Fig. 4
figure 4

Percentage (y-axis) and number of trials (above the bars) for P. nausithous larvae not adopted (white bar), surviving (grey bar) and killed (black bar) relatively to the difference (***difference statistically significant) within population with and without P. nausithous. Difference in adoption rate in the sympatric population not statistically significant (χ2 = 0.39, df = 1, P = 0.53, N = 23); in the allopatric population statistically significant (χ2 = 7.11, df = 1, P = 0.01, N = 36). Difference in surviving after adoption between the populations not statistically significant (χ2 = 0.22, df = 1, P = 0.64, N = 20)

Full size image

Choice behaviour of colonies at population level

Colonies from both sympatric and allopatric populations preferred to adopt their own larvae rather than P. nausithous larvae in the simultaneous choice experiments (Fig. 5). The sympatric population colonies adopted 24 M. rubra larvae and 10 P. nausithous larvae (χ2 = 5.76, df = 1, P = 0.02, N = 34). The preference for the own larvae was even higher in the allopatric population (33 M. rubra larvae vs. 8 P. nausithous larvae) (χ 2 = 15.24, df = 1, P < 0.001, N = 41) but this difference was not significantly different between the two host populations (χ 2 = 1.00, df = 1, P = 0.32, N = 75). The mean of adoption times of M. rubra (7.31 ± 2.56 min) were significantly shorter than those of P. nausithous larvae (16.83 ± 4.29 min) in the sympatric population (Mann–Whitney U test, P < 0.01, N = 32). This difference was not significant in the allopatric population (M. rubra 11.93 ± 1.87, and P. nausithous larvae 19.71 ± 5.41 min) (Mann–Whitney U test, P = 0.14, N = 39), but the differences in adoption times between populations (7.77 vs. 8.52 min, respectively) were not statistically significant (Mann–Whitney U test, P = 0.75, N = 18).

Fig. 5
figure 5

Choice behaviour of colonies (number of adoption) Percentage (y-axis) and number of trials (given above the bars) for M. rubra larva (white bar) and P. nausithous larva (black bar) relatively to the difference (***difference statistically significant) in adoption within and between populations (sympatric population: χ2 = 5.76, df = 1, P = 0.02, N = 34; allopatric population: χ2 = 15.24, df = 1, P < 0.001, N = 41; χ2 = 1.00, df = 1, P = 0.32, N = 75)

Full size image

Discussion

Our results confirm previous findings from several ant genera that workers can distinguish between nest-mate and foreign conspecific larvae (Brian, 1975; Meudec, 1978; Isingrini et al., 1985; Hare and Alloway, 1987; Carlin and Schwartz, 1989; Bonavita-Cougourdan et al., 1989; Fénéron and Jaisson, 1992). Since the complete adoption of Phengaris larvae can take excessive time under laboratory conditions (>8 h Als et al., 2001; Witek et al., 2011), our low overall adoption rates might just be a reflection of the short observation interval. Nevertheless, excluding adoptions that took more than 60 min: the average of adoption time for P. nausithous in the sympatric population (16.83 min) was higher than for the cuckoo P. alcon in Denmark exploiting M. rubra (9.7 min, Als et al., 2001) and much less than for the predator P. teleius in Poland exploiting M. rubra (43.33 min, P. teleius adopted within 1 h Witek pers. comm.). The similarity in adoption times of P. alcon and P. nausithous adds to previous results about host specificity and integration of P. nausithous by Patricelli et al. (2010).

As expected, Myrmica workers are more aggressive towards foreign larvae—a trait typically observed within encounter experiments between workers (Dahbi et al., 1996; Errard and Hefetz, 1997; Lahav et al., 1999; Lucas et al., 2005). This behavioural assay can, however, also be used to quantify the interactions between P. nausithous and the ants. M. rubra workers are well able to distinguish between P. nausithous and M. rubra larvae in individual and colony encounters. P. nausithous larvae were more often inspected than foreign M. rubra larva, but antennation behaviour was less often released in the ants of the sympatric than in those of the allopatric population. These results can be interpreted in two ways:

  1. 1.

    Local adaptation of the allopatric host ants which avoid infestation by the parasite because they scrutinize the parasites more carefully and have a reduced adoption rate. This may have caused the parasite to temporarily disappear from the populations (host is ahead of parasite).

  2. 2.

    The higher adoption rate of the sympatric ants reflects a local adaptation of the P. nausithous larvae to better mime the sympatric host population (parasite is ahead of host), as already found for P. alcon (Als et al., 2002; Nash et al., 2008).

Foitzik et al. (2001; 2003) also found geographic variation in the response of Lepthorax longispinosus and L. acervorum host to two slave-making ant species, Protomognathus americanus and Harpagoxenus sublaevis. The social parasites were shown to induce behavioural differences in different host populations. L. acervorum host workers attacked H. sublaevis slavemakers from the same population less often than allopatric slavemakers (Foitzik et al., 2003). This may reflect a similar stage of a host–parasite arms race than in our case.

In conclusion, our data suggests that a local co-evolutionary arms race between P. nausithous and M. rubra resulted in a local adaptation. Given P. nausithous is the most specialized Phengaris species in terms of number of host ant species used, with a very strong preference for M. rubra (Thomas et al., 1989; Elmes et al., 1998; Stankiewicz and Sielezniew, 2002; Tartally and Varga, 2005; Witek et al., 2008; Patricelli et al., 2010) it would not be that surprising to find evidence for local adaptation. Although our results give support for local adaptation of adoption behaviour the co-evolution of host–parasite system are highly dynamic oscillating processes. Therefore, other populations might reveal alternative outcomes with the host or parasite having the lead and our results only represent a snap shot in the arms race between P. nausithous and M. rubra.