Acoustical mimicry in a predatory social parasite of ants

Effective communication is fundamental to the ability of social insects to live in complex hierarchical societies, in which different castes or individuals perform different activities, yet each type of behaviour is so well integrated that the colony functions as a ‘superorganism’ (Hölldobler and Wilson, 2009). In ants, the primary method of communication involves chemical and, to a lesser extent, tactile cues (Hölldobler and Wilson, 1990; Hölldobler and Wilson, 2009). Acoustics also plays a role among the adults of four ant subfamilies (Ponerinae, Nothomyrmecinae, Pseudomyrmecinae, Myrmicinae) that can stridulate (Markl, 1973; Taylor, 1978), and others that drum the substrate, by inducing specific behaviours in receiving individuals (Markl, 1965; Barbero et al., 2009) or by amplifying or modulating the effects of pheromones (Markl and Hölldobler, 1978; Baroni-Urbani et al., 1988; Hölldobler, 1999).

Although acoustics has generally been regarded as ‘weakly developed’ in ants (Hölldobler and Wilson, 1990; Hölldobler and Wilson, 2009; Keller and Gordon, 2009), it is also the least studied of their communication systems (Barbero et al., 2009). Indeed, knowledge of the structures involved in the production and reception of signals is incomplete. Most sounds or vibrations are produced by a minutely ridged stridulating organ (pars stridens) situated on the mid-dorsal edge of the fourth ‘abdominal’ segment, and by an embossed spike (plectrum) projecting from the rear edge of the post-petiole (Giovannotti, 1996; Grandi, 1966; Pavan et al., 1997; Grasso et al., 1998; Ruitz et al., 2006). When an ant raises and lowers her gaster, the two structures rub together and emit a series of ‘chirps’ (Hölldobler and Wilson, 1990; Roces and Hölldobler, 1996; Ruitz et al., 2006). Stridulations are defined by the transmitting medium, and may be sounds transmitted by air or vibrations transmitted by the substrate. Many myrmecologists maintain that ants cannot hear the aerial component of stridulations and perceive only substrate-transmitted vibrations (Fielde and Parker, 1904; Roces and Tautz, 2001; Hölldobler and Wilson, 2009), an idea supported by the discovery of a subgenual organ in Camponotus ants (Menzel and Tautz, 1994). However, evidence for the perception of air-transmitted sounds is accumulating, at least over distances of a few centimetres (Hickling and Brown, 2000; Hickling and Brown, 2001; Roces and Tautz, 2001). Possible receptors of air-borne sounds are the Johnston's organ (Masson and Gabouriaut, 1972) or the trichoid organs on the tips of the antennae (Dumpert, 1972), although the latter, in ants, differ in structure from the filiform hair trichoid sensilla used by other insects and spiders for auditory reception (Tautz, 1977; Kumagai et al., 1998; Barth, 2000).

However sounds are detected, ant acoustics frequently signal antagonistic or distress behaviours between workers in a colony, including alarm (Markl, 1965; Markl, 1985; Roces and Hölldobler, 1995), intimidation (Stuart and Bell, 1980; Ware, 1994), aposematic ‘threatening’ (Santos et al., 2005), or a call for rescue after a cave-in of the nest (Markl, 1985). In addition, when combined with pheromones or in isolation, several species stridulate or drum during foraging to enhance worker recruitment to food sources (Markl and Hölldobler, 1978; Baroni-Urbani et al., 1988; Roces et al., 1993; Hölldobler and Roces, 2000): in Atta stridulations are most frequent when leaves of the highest quality for fungal cultures are found (Hölldobler and Roces, 2000). Myrmica workers frequently stridulate during trophallaxis, particularly the receiving worker when food decreases (Stuart and Bell, 1980; Zhantiev and Sulkanov, 1977).

Inter-caste acoustical communication has been recorded in only a few instances. Mating queens of Pogonomyrmex badius stridulate to signal to males when their spermathecae are full (Markl et al., 1977), whereas in Atta those workers that cut leaves stridulate when they are ready to return to the nest. This recruits individuals of the smallest minim caste to climb onto the leaf fragment from where they protect their larger sisters from attack by phorid flies during the journey home (Roces and Hölldobler, 1995). Until recently, there was no direct evidence that different members of an ant society produced distinctive caste-specific sounds to induce appropriate patterns of behaviour in fellow or other castes, although this was implied when Markl (Markl, 1968) found that the major workers of Atta cephalotes produce more intense sounds, that carry further, than their smaller nestmates, and by Grasso and colleagues' (Grasso et al., 1998) demonstration that the space between the ridges of the pars stridens in queens of Messor species was greater than in the workers. Working with Myrmica schencki, we reported that the difference in width between the ridges of workers and queens was larger than in Messor, and that despite a considerable overlap, the queens made distinctive sounds from the workers (Barbero et al., 2009). When recordings of unstressed adult M. schencki were played back to laboratory cultures of workers, the sounds of both castes induced benign responses including aggregation and antennation at the speaker. Moreover, when workers were played their queen's sounds, they stood ‘on guard’ on the speaker to a much greater extent than when worker sounds were played, each holding the characteristic posture adopted by a Myrmica worker when protecting an object of high value to the colony (Barbero et al., 2009).

The chemical component of ant communication systems is highly specific, enabling colony members to recognise kin ants as well as conspecifics. Nevertheless, approximately 10,000 other invertebrate species live as social parasites within ant colonies, where they exploit the rich resources concentrated inside nests (Thomas et al., 2005a). The mechanism whereby they penetrate and often integrate with their host society has been studied in very few social parasites, but generally involves corrupting the honest signals of the ants (Hölldobler and Wilson, 1990; Akino et al., 1999; Lenoir et al., 2001; Thomas et al., 2005a; Hojo et al., 2009). European species of Maculinea butterflies are among the better understood examples. The adult butterfly is free-living and oviposits on a specific food plant(s), on which the first three larval instars feed. On entering the fourth and final instar, weighing just 1-2% of its ultimate biomass, the small larva (hereafter called a caterpillar) falls from its plant and secretes semio-chemicals that mimic the surface hydrocarbons of Myrmica ants, causing any forager that encounters it to carry the caterpillar to the brood chambers of the underground ant nest (Thomas, 1984; Thomas, 2002; Elmes et al., 1991a). Two strategies have evolved in Maculinea to exploit Myrmica resources. Caterpillars of the cuckoo species, Maculinea rebeli and M. alcon, remain in the brood chambers for 11-23 months, where they are tended and fed directly by the nurse ants on regurgitations and other food (Elmes et al., 1991a; Elmes et al., 1991b). Caterpillars of M. arion and M. teleius are predators of ant larvae, and inhabit peripheral cells from which they periodically foray to binge-feed in the brood chambers (Thomas and Wardlaw, 1992).

With a few apparent exceptions (Barbero, 2007; Tartally et al., 2008), all Maculinea species or populations are specific to one primary host species or local genotype of Myrmica at a regional scale (Thomas et al., 1989; Thomas et al., 2005a; Thomas et al., 2005b; Nash et al., 2008). In cuckoo species, host specificity is explained by chemical mimicry of the host's communication system. Pre- and newly adopted caterpillars of cuckoo Maculinea are weak chemical mimics of their regional host (Akino et al., 1999; Nash et al., 2008), but within a few days are given preferential care. Thus, in western Europe where its host is Myrmica schencki, Maculinea rebeli caterpillars are rescued ahead of the ant brood when a colony is disturbed, and are fed in preference to M. schencki larvae when food is scarce (Thomas et al., 1998). This deeper infiltration coincides with the secretion of additional hydrocarbons that more closely mimic the distinctive odour of M. schencki, but inevitably make this genotype of M. rebeli a poor (and seldom tolerated) mimic of other Myrmica species (Elmes et al., 2004; Schönrogge et al., 2004).

Neither chemical mimicry nor their begging behaviour explains why M. rebeli caterpillars are treated in preference to host ant brood. Instead, we have suggested that acoustical cues are employed (Barbero et al., 2009). It is well known that certain pupae and caterpillars of Lycaenidae produce sounds; the former from tooth-and-comb stridulatory organs between the fifth and sixth segments (Downey, 1966; Downey and Allyn, 1973; Downey and Allyn, 1978; Pierce et al., 2002), whereas caterpillar sounds probably emanate from muscular contraction and air compression through the trachaea (Schurian and Fiedler, 1991). The acoustics of mutualistic lycaenid species do not obviously mimic ant stridulations, and an attraction to ants has been demonstrated only in the pupae of one extreme mutualist (Travassos and Pierce, 2000; Pierce et al., 2002). The calls of socially parasitic Maculinea caterpillars, however, more closely resemble Myrmica worker stridulations, although the putative mimicry appeared to be modelled on the genus rather than a host species of Myrmica (DeVries et al., 1993). The same study showed that Myrmica larvae are mute, suggesting that in this trait Maculinea caterpillars are mimicking an adult ant cue.

The recordings by DeVries et al. (DeVries et al., 1993) were restricted to distressed worker ants and caterpillars, and were not played back to the ants. Using modern equipment, we recently found that unstressed Maculinea rebeli caterpillars and pupae were close acoustical mimics of Myrmica schencki, and that the sounds produced by both butterfly stages were significantly closer to those of the queen ants than the workers (Barbero et al., 2009). Playing M. rebeli sounds back to laboratory cultures of M. schencki workers resulted in similar enhanced ‘on guard’ (and other) benign behaviours as when the ants were played the stridulations of their own queens, especially when pupal calls were played. We concluded that although chemical mimicry is used to gain and maintain acceptance in a M. schencki society, the social parasite simultaneously employs acoustical mimicry inside the nest to advance its status to that of the highest member in its host's hierarchy.

DeVries et al. (DeVries et al., 1993) showed that caterpillars of predatory Maculinea species also produce sounds that appear to mimic Myrmica (worker) stridulations, although in nature they are less closely integrated with their host's society (Thomas et al., 2005a). We speculated that they might be less perfect acoustical mimics of their hosts. Here we test this idea by comparing the acoustics of unstressed Maculinea arion caterpillars and pupae with those of the queens and workers of its host ant, Myrmica sabuleti, and with our published data for Maculinea rebeli and Myrmica schencki. We also compare the worker and queen sounds of M. sabuleti, and those of another ant Myrmica scabrinodis, to determine whether the distinctive acoustical communication made by different castes of M. schencki exists in its congeners.

Myrmica colonies were excavated in the field and set up as laboratory ant colonies of >100 workers in 28 cm×15 cm×10 cm Perspex containers and maintained on a diet of sugar and Drosophila larvae. We collected Myrmica schencki Emery 1894 colonies (N=12) in May 2006 at Colle di Tenda, Piedmont, Italy where this species is host to Maculinea rebeli (Barbero et al., 2009). Two colonies of M. sabuleti Meinert 1861 were collected from Italy and four from Somerset, UK from sites where they are host to Maculinea arion. Eleven colonies of M. scabrinodis Nylander 1846 were taken from Dorset, UK from sites with no Maculinea parasites, although this ant is host to Maculinea alcon and M. teleius elsewhere in Europe (Thomas et al., 2005a). The post-adoption larvae and pupae of Maculinea arion Linnaeus 1758 were collected in Somerset, UK and kept with M. sabuleti colonies until they were recorded.

M. sabuleti and M. scabrinodis ants (three queens and three workers of each species) were kept in 70% ethanol and dissected between the post-petiolum and the abdomen to expose the pars stridens and the plectrum. The two ant parts were mounted on the same steel stub, coated with gold, and the distances between adjacent ridges of the pars stridens were measured automatically (10 measurements per individual) using a Cambridge Stereoscan S360 scanning electron microscope (SEM). The SEM operated at 20-25 kV.

Recordings were made of individual workers and queens of the three Myrmica species (Table 1) using the procedures described by Barbero et al. (Barbero et al., 2009). We also recorded four pre-adoption caterpillars of Maculinea arion, i.e. before they came into contact with the ants, three post-adoption caterpillars and three pupae. The sounds for M. rebeli and Myrmica schencki used in the analyses are the same as those used in Barbero et al. (Barbero et al., 2009) (Table 1).

The recording equipment consisted of a 12.5 cm×8 cm×2 cm recording chamber with a moving-coil miniature microphone attached through the centre. A second microphone of the same type was used to record ambient noise but in anti-phase. An amplifier was attached to each microphone and calibrated to maximise the noise cancellation of ambient noise from the two microphones, leaving the signal from the recording chamber. The resulting signal was processed through a two-stage low-noise amplification before being recorded digitally on a laptop computer using Audacity v. 1.2.4 (http://audacity.sourceforge.net/). To further reduce ambient noise and interference, the equipment was powered by a 12 V gel cell battery, and the recording chamber and microphones were placed inside an anechoic chamber. Sounds were recorded for 15 min periods starting 5 min after an insect was introduced and had become calm.

The sound parameters used by DeVries et al. (DeVries et al., 1993) and Barbero et al. (Barbero et al., 2009), dominant frequency (DF; Hz), pulse repetition frequency (the reciprocal of the duration of one pulse; PRF; s⁻¹) and pulse length (PL; s), were measured using Audacity 1.2.4. To test whether sound differed between groups we calculated the pairwise normalized Euclidean distance over all three parameters and used a nested (‘Individual’ within ‘Group’) ANalysis Of SIMilarity implemented in Primer v. 6.1.12 (Primer-E Ltd.). Groupings were visualised using non-parametric multi-dimensional scaling (nMDS) with sound fragments averaged within individuals. To further test whether the overall sounds produced by butterfly pupae and larvae were more similar to queens or workers of the two host ants M. sabuleti and M. schencki, we estimated mean Euclidean distances between groups and used Student's t-test to estimate the significance of the differences. Student's t-test was also used to establish the differences in the morphometric measurements on the stridulation organs of M. sabuleti and M. scabrinodis queens and workers.

The workers of all Myrmica species recorded to date have been shown to stridulate. We recorded both workers and queens of Myrmica sabuleti, M. scabrinodis and M. schencki to compare species and casts. Average measurements for the three sound parameters PL, PRF and DF are listed in Table 2. Using a multivariate approach over the three sound parameters, the normalised Euclidean distance within each group were (mean ± s.e.m.): M. sabuleti queens: 0.68±0.05, workers: 0.97±0.08; M. scabrinodis queens: 1.26±0.19, workers: 1.38±0.22; M. schencki: queens: 0.97±0.06, worker: 1.74±0.05. The sounds of M. schencki workers are quite diffuse (Fig. 1), whereas those of the other workers are more clustered. However, the sounds of queens of all three species were very similar and closely overlap, as seen in Fig. 1. This is also reflected in the results of a nested ANOSIM to test for pairwise differences between groups. Among the ant groups, M. schencki workers were significantly different only from their own queens (ANOSIM R=0.12, P=0.04), and did not separate from the workers or queens of the other species (R<0.12, P>0.17). More intriguingly, there was no significant difference in the sounds made by the queens of the three species tested (M. sabuleti versus M. scabrinodis R=0.05, P=0.22; M. sabuleti vs M. schencki R=0.002, P=0.43; M. scabrinodis vs M. schencki R=0.11, P=0.15). All other pairwise comparisons among groupings show significant separation (R≥0.49, P≥0.01). In other words, as previously demonstrated in M. schencki, the queens of both M. sabuleti and M. scabrinodis made distinctive sounds that were significantly different from those of their workers. This difference reflects the distinct stridulatory organ morphology of the two castes (Fig. 2). As shown in M. schencki, the distance between the pars stridens ridges in M. sabuleti (queens: 1.61±0.18 μm, workers: 1.35± 0.11 μm) and M. scabrinodis (queens: 1.86±0.50 μm, workers: 1.27±0.48 μm) differs between queen and worker ants (two-sample t-test t_M.sabuleti=6.53, d.f.=58, P<0.001; t_{M.scabrinodis}=4.66, d.f.=58, P<0.001).

Using a nested ANOSIM we found that sounds from M. arion pupae differ from those of pre-adoption larvae (R=1, P=0.029), yet they apparently do not differ from the post-adoption larvae (R=1, P=0.1; Fig. 3). However, we interpret the results in Fig. 3, as an artefact of small sample size. Since data were available for only three pupae and three post-adoption larvae, there were only 10 possible permutations, and in one instance R, after permutation, was greater than the observed R, giving a level for P of 0.1.

The acoustical signals of all three groups of butterfly differed significantly from both M. sabuleti queens and workers (Fig. 3). t-Tests comparing the between-group normalised Euclidean distances of the butterfly to workers and queens showed that the acoustics made by all three stages of M. arion were significantly closer to the stridulations of the host queens rather than workers (Pupae: distance_queens=1.98±0.07, distance_worker=3.33±0.09, t=11.3, d.f.=41, P<0.001; pre-adoption larvae: d_queens=2.32±0.08, d_worker=3.13±0.08, t=7.18, d.f.=61, P<0.001; and post-adoption larvae: d_queens=3.33±0.09, d_worker=4.22±0.11, t=6.13, d.f.=45, P<0.001).

Owing to their cuckoo feeding life style, M. rebeli larvae, although not the pupae, are in more frequent and closer contact with ant workers than those of M. arion. We therefore predicted that M. rebeli might be a closer acoustical mimic of its host M. schencki than M. arion of M. sabuleti. In the absence of behavioural data, we compared the normalised Euclidean distances of sounds of the larval and pupal stages of both Maculinea species to the stridulations of both Myrmica species' queens and workers (Table 3).

We found no evidence that M. rebeli is a closer mimic of M. schencki than M. arion is to M. sabuleti. In fact, with one exception (M. arion pre-adoption larvae and M. rebeli pupae are equidistant to M. schencki workers), in all comparisons between M. arion pupae or pre-adoption larvae and M. rebeli larvae or pupae with queens and workers of either ant species, the acoustics of M. arion were significantly closer to Myrmica stridulations than were those of M. rebeli. Only the acoustics of M. arion post-adoption larvae were found to be more distant than either stage of M. rebeli to both castes of either ant species (Table 3). Overall, the acoustics made by the immature stages of M. arion were 19.7% closer to Myrmica stridulations than those from M. rebeli (max=39.3% M. arion pupa and M. rebeli larva vs M. sabuleti workers, min 6.5% M. arion pre-adoption larva and M. rebeli larva vs M. sabuleti workers). This comfortably encompasses the range of acoustical variation found between M. rebeli larvae and pupae, where differences in the worker responses to butterfly sounds have been demonstrated (Barbero et al., 2009).

Our results demonstrate that stridulating queens from two additional Myrmica species make distinctive sounds to those of their workers using morphologically distinct organs. Indeed, intra-specific inter-caste differences were more clear-cut in M. scabrinodis and M. sabuleti than we previously reported for M. schencki (Barbero et al., 2009). Less expected was the fact that queen stridulations from the three Myrmica species were indistinguishable, as were worker stridulations in two of the three pairs of species tested. This suggests that acoustics plays little or no part in the cues used by Myrmicato distinguish between non-kin or other species of ant and members of their own society, and indeed numerous studies demonstrate the predominant role of chemical cues and the gestalt odour in colony recognition or physiological states within an ant society (Hölldobler and Wilson, 1990). However, our recent results suggest that acoustical communication, in isolation, is capable of signalling at least the caste and the status of a colony member, and of inducing appropriate behaviour towards it by the workers (Barbero et al., 2009). It is possible that caste differences are yet more distinctive than the data presented here (Figs 1, 2), since to date we have measured and compared only the three attributes most commonly studied in ant—butterfly acoustics: dominant frequency, pulse repetition frequency and pulse length. Furthermore, when analysing our recordings of Myrmica acoustics, we noted that a wide variety of sound sequences was made by an individual queen or worker (see Barbero et al., 2009) (Fig. 2). During bioassays, we played the full repertoire of a test ant to cultures of workers, but it is possible that different component phrases induce one or other of the three behaviours observed in these tests (aggregation, antennation or guard attendance). It is also probable that individual ants can alter the rhythms, speed and intensity of stridulations to communicate other information under more natural or different conditions to those imposed by our experiments. For example, our equipment and procedures ensured that the ants were unstressed both at the time of recording and when receiving signals; and we observed only benign responses, with none of the antagonistic or alarm behaviours induced in early experiments, when ants were often distressed; and vice versa (Hölldobler and Wilson, 1990; Barbero et al., 2009).

Yet our experimental procedures were far from natural, and we suspect that a wider spectrum of information might involve acoustical communication. To draw a parallel from vertebrates, the main benefit of acoustical communication is its flexibility, allowing many short-lived signals to convey different information in a short time, by changing the pitch, harmonies or volume of the sound (Krebs and Davies, 1993; Greenfield, 2002). We have not yet studied whether different castes of Myrmica ant respond differently when exposed to the same sounds, although this seems probable, because queen Myrmica schencki respond aggressively when introduced to Maculinea rebeli pupae (which mimic queen sounds) whereas the workers tend them gently (Barbero et al., 2009) (Supporting Online Material: http://www.sciencemag.org/cgi/content/full/323/5915/782/DC1). And it is of course probable that the distinctive acoustical signals made by different members of an ant society modulate chemical and tactile cues in different ways.

Confirmation that the queens in other Myrmica societies make distinctive sounds from their workers indicates that this model is available to other mimetic Maculinea species or races that parasitise other Myrmica species, as we confirm here for Maculinea arion and Myrmica sabuleti. However, the demonstration that queens make the same sound in all three host ants studied suggests that acoustical mimicry functions strictly to raise the hierarchical status of a social parasite once it has been successfully accepted as a chemical mimic by a host society. In other words, acoustical mimicry is genus rather than species specific, as DeVries et al. (DeVries et al., 1993) concluded, albeit after recording highly stressed ants and Maculinea caterpillars. Although this presumably conveys a considerable benefit when resources in a nest are scarce — a frequent occurrence when Myrmica colonies are parasitised by supernumerary caterpillars of predatory (Thomas and Wardlaw, 1992) or cuckoo Maculinea caterpillars (Thomas et al., 1993) — it does not influence host specificity. For example, despite their similar acoustics, mortality of Maculinea arion caterpillars is more than five times greater in Myrmica scabrinodis nests than when adopted by M. sabuleti, whereas mortality of (western) Maculinea rebeli caterpillars is more than 30 times higher when adopted by M. sabuleti than by their primary host, M. schencki (Thomas et al., 1989; Thomas et al., 2005a).

In our earlier study of M. rebeli—M. schencki interactions, we found that the pupal stage of the social parasite was a closer acoustical mimic of host queens than the caterpillar. This was demonstrated by worker ant behaviour, where pupal and queen sounds elicited characteristic ‘on guard’ behaviour at equal frequency, as did larval sounds, but to a slightly lesser degree (Barbero et al., 2009). That and this study also illustrate the amount of variation in acoustic signals found within and between the groups. The only guidance currently available of how much difference in acoustical similarity results in a difference in worker behavioural response comes from the original study, where, for instance, there were 27% more differences between the pupal and the worker calls than between the pupal and the queen calls, resulting in behaviour frequencies that were the same as towards queens, but significantly different from workers (Barbero et al., 2009). In this study, between-group differences in acoustical similarity ranged from ~6% to ~40%, and those of M. arion pupae and larvae tended to be closer to the queens than those of M. rebeli pupae and larvae, suggesting that the predatory species is at least as good a mimic as the cuckoo-feeding one. Further behavioural work should focus on the response norm of the receiving partner, i.e. the variation in the cue that is still accepted to trigger a behaviour, which in this case would be both workers and queens. This could be achieved by varying particular aspects of the sounds used for behavioural assays by computer manipulation. That M. arion caterpillars are apparently closer mimics of queen Myrmica would suggest that this interaction evolved as a basal trait in the Phengaris—Maculinea clade of Lycaenidae, for molecular studies suggest that the predatory life style preceded the evolution of the cuckoo forms (Als et al., 2004). Behavioural assays and further recordings will allow acoustic signalling to be used to address these evolutionary hypotheses.

As in ant communication, we suspect that the role of acoustics has been underestimated in the few studies made to date of the adaptations with which an estimated 10,000 species of invertebrate social parasite succeed in cheating ant societies. Promising taxa for future research are Lepidochrysops and 11 other lines of lycaenid butterfly that have independently evolved social parasitism from mutualistic, presumably sound-producing, ancestors (Fiedler, 1998): Myrmecophila species of cricket; the stridulations of staphilinid beetle social parasites; and the many inquiline ‘queen’ ants that parasitise other ants, including their close relatives.

We thank Luca Casacci for help during fieldwork and Mark Charles for designing and building the acoustic equipment.