Introduction

Ant societies may function as small ecosystems for various other animals (Hölldobler and Wilson 1990). Especially large ant colonies offer ecological niches free of enemies and rich in food resources for organisms that have managed to intrude into the host’s social system. Although such symbionts frequently partake in host resources, the term symbiosis originally did not imply a particular positive or negative interaction between the partners (Goff 1982). Symbionts that live in close association with ants are also frequently referred to as myrmecophiles. To attain social integration, myrmecophiles apply adaptive strategies of appeasing or deterring the host or of circumventing the nest recognition system (Lenoir et al. 2001). Specifically with migratory ants, myrmecophiles require additional mechanisms of maintaining contact with their mobile hosts. Although many army ants impose such challenges by being highly migratory, forming only temporary nests (bivouacs) that are shifted frequently, they exhibit an extraordinarily diverse fauna of myrmecophiles (Gotwald 1995). This is probably due to the large colonies that offer numerous ecological niches and to the availability of considerable amounts of high-quality food that is retrieved during regular and highly efficient raiding activities. The Southeast Asian ponerine, Leptogenys distinguenda (Emery) (Hymenoptera: Formicidae), which reaches colony sizes of more than 50,000 workers and exhibits army ant behavior (Maschwitz et al. 1989), is also inhabited by numerous myrmecophiles, including beetles, phorid flies, springtails, silverfish, woodlice, spiders, and snails (Witte 2001; Witte et al. 2008). By depleting colony resources or by direct predation on the host, myrmecophiles might exert fitness costs on their host (Kistner 1982), and under these circumstances, selection may have occurred such that the host now displays counterstrategies to minimize the costs of symbionts. In this study, we analyzed the specific interactions for two selected myrmecophiles and their host, L. distinguenda, in order to identify potential host countermeasures that would reduce symbiont pressure. We compared the integration levels of the oonopid spider, Gamasomorpha maschwitzi (Wunderlich 1994), and the newly described silverfish, Malayatelura ponerophila gen. n. sp. n. (Mendes et al., unpublished). Both of these symbionts occur in most L. distinguenda colonies, and since they consume some of the host’s resources, they are costly for the host (Witte et al. 2008). In this study, we compared the chemical and behavioral integration strategies of these two morphologically and systematically different symbiont species.

Chemical cues are known to play a major role in social insect recognition mechanisms; social insects generally use characteristic hydrocarbon profiles on the cuticle to recognize nestmates (Reese 1982; Vender Meer et al. 1982; Howard and Blomquist 2005; Hefetz 2007). These surface chemicals are transferred constantly among nestmates, passively by physical contact and actively by grooming and trophallaxis (Soroker et al. 1995). Myrmecophiles have been found to mimic the chemical profiles of their hosts either by biosynthesis of cuticular hydrocarbons or by acquiring the host chemical profiles passively (Dettner and Liepert 1994). The term “chemical mimicry” is used in this paper according to the original definition of the term “mimicry,” as one organism’s resemblance to another organism’s properties (Vane-Wright 1976). Since chemical profiles of ant colonies may change considerably over time (Hölldobler and Wilson 1990), intruders must be able to update their profiles constantly in order to avoid detection. We hypothesize that myrmecophiles will fail to match host profiles if they do not continuously renew their chemical cues. At some point, profiles will diverge to such an extent that hosts can detect, expel, or even kill myrmecophiles. Once a certain state of mismatch in recognition cues has been reached, it must become increasingly difficult to approach host workers and to update the chemical profiles for organisms relying on passive chemical mimicry. By negative feedback, this process might even lead to complete social rejection, which would be fatal for obligate symbionts. Hence, symbionts might pay high costs for the exploitation of their host. We show in this study that considerable differences exist in the social acceptance of two common symbionts and that their social integration is a well-balanced and potentially fragile system.

Methods and Materials

We used a fourfold approach to study the behavioral and chemical integration of the myrmecophiles into the colony of their host: (1) studies of behavioral interactions between host ants and myrmecophiles in two colonies, which included their native myrmecophiles, but were reduced in size to enable laboratory observations; (2) studies of recognition and aggression between colonies with different cuticular hydrocarbon profiles; (3) identification of cuticular hydrocarbons in free-living host colonies based on the samples of highest concentrations; and (4) analyses of chemical similarities between hosts and myrmecophiles based on a large dataset of relative quantities of the principal cuticular hydrocarbon components from six free-living colonies.

Behavioral Observations

We studied the host ant L. distinguenda in a well-recovered secondary, dipterocarp, lowland rainforest ecosystem at the Field Studies Centre of the University Malaya in Ulu Gombak, Malaysia (03°19.4796′ N, 101°45.1630′ E; elevation, 228.8 m). Studies were carried out for a total of 9 weeks, 3 weeks each in March and September 2006 and another 3 weeks in March 2007. Host colonies were located during their activity phase at night by retracing raiding columns to the nest sites. Nest sites were numbered, marked with tape, and checked every 30 min between 9:00 p.m. and 3:00 a.m. for ongoing activities. Army ants shift their nest sites frequently, carrying all larvae and pupae in a continuous column to a new location (migration). With L. distinguenda, such migrations last several hours. When a colony was found emigrating, G. maschwitzi and M. ponerophila were collected with aspirators directly from the emigration column. Adult and young teneral adult host workers (also known as callows), larvae, and pupae also were collected to assemble simplified observation nests for behavioral studies (two colonies) and for exchange experiments (two additional colonies) at the field station (see below). Six additional collections, again from different free-living host colonies, were subjected to cuticular hydrocarbon extraction and chemical analyses (see below). To permit investigation of potential influences of nest composition on the interactions between the host and each symbiont, laboratory colonies were assembled with a different demographic composition. Worker numbers ranged from 83 to 140, callows from 3 to 34, larvae from 15 to 28, and pupae from 9 to 53. Ant colonies were housed in clear plastic containers (1 × 14 × 20 cm) with a 1-cm wide entrance, shaded with cartridge covers, and placed upside down on a moistened plaster floor in an arena (9 × 25 × 32 cm). Arenas were covered with plastic lids when no observations were carried out, and the side walls were treated with paraffin oil to prevent escape of the ants. These nests were for behavioral studies only, so the paraffin treatment did not affect our chemical analysis.

Interactions among spiders, silverfish, and host ants were monitored during nightly observation sessions of defined length, on average, 12 sessions of 5 min per night. Observation times of the four laboratory nests summed up to more than 38 hours and included 20 spiders and 41 silverfish. Observations were carried out both before and after feeding. Each day before the observation sessions, total numbers of all nest inhabitants were counted. For statistical analysis, the behaviors observed between the host and the spider and between the host and the silverfish were standardized in a behavioral index (F r). This was necessary to compare laboratory nests of different composition, considering that the odds of interactions are influenced by the number of potentially interacting individuals and the observation time. The index (F r) standardizes the frequency of a particular behavioral pattern between the objects A and B relative to the number of individuals in the focal groups (A and B) and observation time with \(F_{\text{r}} = \left[ {{{F_{\text{o}} } \mathord{\left/ {\vphantom {{F_{\text{o}} } {\left( {N_{\text{A}} \times N_{\text{B}} \times t} \right)}}} \right. \kern-\nulldelimiterspace} {\left( {N_{\text{A}} \times N_{\text{B}} \times t} \right)}}} \right]\), where F o is the observed frequency, N A is the number of animals in group A, N B is the number of animals in group B, and t is the observation time. Behavioral indices were evaluated with nonparametric statistical methods (Wilcoxon-matched pairs test, Mann–Whitney U test, and Kruskal–Wallis test). Furthermore, the data were standardized to 100% for each symbiont species to compare the overall proportions of behavioral patterns.

To study the dependency of social acceptance on colony-specific chemical cues, reciprocal exchange experiments were carried out between two L. distinguenda laboratory colonies, including 40 host ants, ten spiders, and three silverfish. Unfortunately, spiders and silverfish could be exchanged only in smaller numbers due to their overall rarity, their low social acceptance, and particularly the mortality of silverfish from aggressive behavior by the host. Twenty ants of a closely related (undescribed) Leptogenys species also were tested as a control. This species resembled L. distinguenda morphologically and showed similar behavior, including army ant mass raiding. Specimens of this species were deposited at the Ludwig-Maximilians Universität (LMU, Munich, Germany). For intercolonial exchanges, individuals from one colony were captured by using a padded aspirator and introduced carefully into a different observation colony without causing disturbance. All interactions were monitored for the following 10 min and forward jerking toward an introduced animal; chasing, mandible grabbing, snapping, biting, or stinging were all classified as “aggression.” All kinds of direct physical contact between two focal objects were considered as “contact.” In addition to these behavioral observations, cuticular hydrocarbons were extracted as described below from the experimental colonies as well as the control species.

Chemical Analyses

We examined the cuticular hydrocarbon profiles of the Leptogenys colonies that were used for laboratory exchange experiments to correlate observed aggression levels directly with chemical similarities/dissimilarities. In addition, chemical analyses were conducted with six L. distinguenda field colonies including their spider and silverfish symbionts to unravel the chemical resemblance between symbionts and their hosts in natural colonies.

Cuticular hydrocarbons were extracted by placing individual insects (or spiders) in 2-ml vials with polytetrafluoroethylene septa and treated for 5 min with 200 μl of pentane (HPLC grade, Sigma-Aldrich). The crude extracts containing both polar and non-polar components were analyzed at the LMU by coupled gas chromatography and mass spectrometry (GC-MS) on an Agilent Technologies 6890N GC and 5975 MSD (70 eV, EI) equipped with a Restek Rxi-5MS column (30 m length, 0.25 mm i.d., 0.25-μm film thickness). Injections were performed in the pulsed splitless mode over 1.0 min at 280°C, including an initial pressure pulse of 16 psi for 0.5 min, followed by automatic flow control at 1.0 ml/min with helium as the carrier gas. The oven program began isothermally at 120°C for 2 min, then increased at 25°C/min until 200°C was reached, followed by a temperature ramp of 4°C/min until the final temperature of 300°C was reached. The transfer line was held constantly at 310°C. A range of 50–500 amu was scanned after an initial solvent delay of 3.8 min.

Identification of Compounds

In the most concentrated extracts, individual compounds, mostly hydrocarbons, were identified by their mass spectra, their retention indices, and comparison with library spectra (Wiley7N). As far as possible, concentrations were evaluated semiquantitatively according to the relative contribution of each peak to the total peak area of the sample. Structures of saturated unbranched and branched hydrocarbons were assigned according to well-established procedures (Carlson et al. 1998; Schulz 2001). Double-bond positions in alkenes and alkadienes were identified by derivatization of the crude extract with dimethyldisulfide (DMDS; Francis and Veland 1981; Vincenti et al. 1987; Carlson et al. 1989; Schulz and Nishida 1996). The resulting adducts allowed assignment of the (E)-or (Z)-configuration only in cases when both diastereomers were present in the extract. Because of the completely stereoselective reaction of DMDS, the adduct derived from the (Z)-compound eluted before the (E)-adduct on apolar gas chromatographic phases (Leonhardt and Devilbiss 1985). Double-bond positions in tri-, tetra-, and pentanes were identified tentatively by analysis of mass spectra and rationalization of fragmentation patterns according to known reference compounds. The general appearance of the mass spectra points to a homoconjugated arrangement of double bonds (Blumer et al. 1970; Youngblood et al. 1971; Karunen 1974; Steiger et al. 2007). The mass spectra of these compounds show the characteristic fragments a, a′, and b, which, together with the molecular mass, allow assignment of double bonds (Fig. 1). The spectra are shown in the Supplementary material (S1).

Fig. 1
figure 1

Characteristic fragments for the identification of homoconjugated polyenes. Ions of the general formula [C n H2n − 4]+.are preferentially formed (a, a′), resulting in intense even-numbered ions. Less abundant were ions resulting from other characteristic cleavages next to the polyene system, such as cleaving off the shorter alkyl chain (b) or cleavage in the distal end of the polyene system resulting in less abundant ions (b′), often not more intense than other ions of the same ion series. For example, by this method, 3,6,9-pentacosatriene (1, a = 108, a′ = 290, b = 317 = M-29) can be readily differentiated from 6,9,12-pentacosatriene (2, a = 150, a′ = 248, b = 275 = M-71). Similarly, 6,9,12,15-pentacosatetraene (3, a = 150, a′ = 206, b = 273 = M-71) can be differentiated from the 3,6,9,12-isomer (a = 108, a′ = 248, b = 315 = M-29). Finally, the spectrum of 3,6,9,12,15-pentacosapentaene (4, a = 108, a′ = 206, b = 313 = M-29) was identical to the published spectrum of the all-Z isomer (Leal et al. 2005), also showing an identical retention index. With the exception of 1, all tri-, tetra-, or pentenes were present in trace amounts only (Table 2). Not all trace components could be detected in samples from spiders and silverfish because they usually contained much less material, making identification of trace compounds difficult

Full size image

Chemoinformatics

Characteristic cuticular hydrocarbons for “within-nest groups” of animals (workers, callows, larvae, pupae, spiders, or silverfish of a particular host colony) were identified and compared for potential similarities with multivariate statistical methods. To consider a GC peak characteristic for a particular within-nest group, at least 50% of the samples belonging to this group had to show the corresponding GC peak. Peaks had to be characteristic for at least one group; otherwise, they were considered not consistent and excluded from analysis. Consistency of 50% of the samples within one group is a rather conservative restriction and therefore minimizes the risk of excluding meaningful chemicals from analysis. Contaminants were excluded, and only hydrocarbons of 21 C-atoms or more were included in the statistical evaluation, since these are considered typical insect cuticular hydrocarbons (Carlson et al. 1998). Absolute peak area of these substances was fourth root transformed, and each peak analyzed in a sample was standardized by dividing its area by the transformed area of the maximum peak in the sample to evaluate relative proportions of compounds only. We used this procedure because we had no a priori hypothesis about the composition of profiles between host and symbionts, and standardization by maximum peak area appears most appropriate for comparing profiles containing different numbers of peaks. Non-parametric statistical procedures robust to data type and distribution (Clarke 1999) were applied by using the Primer 6 software package (Primer-E Ltd.). Bray–Curtis distances were calculated, which were then subjected to the following statistical procedures. A similarity percentage (SIMPER) procedure was used to calculate relative contributions of each compound to the similarity of a within-nest group. Non-metric multidimensional scaling (NMDS) was applied to visualize chemical distances. These were analyzed further by hierarchical cluster analysis (CA), with group average cluster mode, and the results were combined with the NMDS plots to visualize similar groups of samples. A priori hypotheses about differences between sample groups were tested for significance by the analysis of similarities (ANOSIM).

Results

We collected myrmecophiles from ten different L. distinguenda colonies and recorded a median of three spiders (range, 1–13) and 14 silverfish (range, 3–47). This difference in abundance is significant (P = 0.008, exact Wilcoxon matched-pair test, N = 10 each). It is important to note that emigrations could not always be sampled from the very beginning so that higher total numbers of myrmecophiles are possible and even likely. Nevertheless, silverfish are found typically in the earlier emigration phases and spiders in later phases, so that the number of silverfish was likely underestimated relative to the number of spiders. Consequently, sampling error should not affect the observed difference in abundances. On three emigrations, larval packets carried by worker ants were collected, which contained silverfish larvae (size, 1.0–1.2 mm) hidden between host larvae, as well as silverfish subadults (size, <4 mm) attached to the outside of the packets. This suggests that the entire life cycle of this myrmecophile takes place within the host colonies.

Behavioral Integration

Surprisingly, we observed considerable differences in survival rates of the myrmecophiles in our laboratory colonies, which contradict the naturally observed abundances. Within the first 6 days of observation, 92–100% of the spiders survived (N = 20), which was similar to host worker survival of 80–100%. In contrast, silverfish (N = 41) disappeared almost completely within these 6 days (0–30% survival), and this loss was attributed to increased host aggression. In all laboratory nests, aggression increased toward silverfish, and active killings of myrmecophiles were frequently observed (Fig. 2). Thus, the observed mortality of silverfish was not a result of inappropriate laboratory handling, but rather a consequence of host aggression (see below). We observed significantly more attacks toward silverfish than spiders (P < 0.001, U = 5211.5, Mann–Whitney U test, N = 54).

Fig. 2
figure 2

Myrmecophiles are attacked and killed by host Leptogenys distinguenda ants. Left Myrmecophilic spider, Gamasomorpha maschwitzi; right myrmecophilic silverfish, Malayatelura ponerophila

Full size image

Overall, the two myrmecophile species showed strongly dissimilar behavioral patterns (P < 0.001, Χ 2 test, Table 1). Spiders interacted more with adult ants (P < 0.001, Mann-Whitney U test, U = 3132.0, N SP = 271, N SF = 147), whereas silverfish interacted more frequently with callow adults (P = 0.001, U = 2,265.5, Mann–Whitney U test, N SP = 174, N SF = 72). Silverfish were frequently observed moving their body surface directly over the cuticle of adult and callow worker ants. This type of close physical contact to the host may be a mechanism that obtains host cuticular chemicals (Lenoir et al. 1997, 2001). This characteristic behavior of silverfish was more frequent during interaction with callow adult workers than with adult ants (P < 0.001, U = 4,842.9, Mann–Whitney U test; N = 31). In accordance with this observation, aggression toward myrmecophiles (N = 87) was never initiated by callows but instead by older workers. Contact between silverfish and worker ants occurred most frequently when the silverfish were below the ants (Table 1) and when the ants were occupied with other tasks such as feeding or brood care. Spiders, in contrast, were typically observed crawling on top of adult workers or callows and actively maintained contact by following worker ants through the nest (Table 1). From the on-top position, spiders had permanent physical contact and frequently moved their legs actively over the cuticle of the host (N > 20). Both myrmecophiles spent significant time in areas where ants were actively feeding and also were observed directly in contact with food (Table 1). The rare occasions when myrmecophiles directly contacted food with their mouthparts (Fig. 3) lasted only a few seconds each.

Fig. 3
figure 3

Myrmecophile contact with food lasts only a few seconds. Left Myrmecophilic spider Gamasomorpha maschwitzi gets brief access to a prey item from the top. Right myrmecophilic silverfish Malayatelura ponerophila participates from the top at larval feeding

Full size image
Table 1 Relative proportions of typical behaviors observed for spiders, Gamasomorpha maschwitzi, and silverfish, Malayatelura ponerophila, in two Leptogenys distinguenda laboratory colonies
Full size table

Aggressive interactions were regularly observed between spiders; however, these interactions did not result in injury or fatality. On encounter, one spider typically displaced another individual after a short period of leg-struggling. Spiders occasionally constructed small, flat webs (0.5–1 cm diameter) inside the ant nests, but prey capture was never observed with these webs. Passing host ants also frequently damaged these webs; therefore, they were of short durability and might serve purposes other than foraging.

In exchange experiments (see “Methods and Materials”), L. distinguenda recognized conspecifics from other colonies as alien; however, they were treated with low aggression (Fig. 4a). In addition to inspection behavior, which included intense bouts of antennation (N = 330), dominance behavior was displayed, which included upright positioning of the resident ant, mandible grabbing and holding down of the subdominant, and antennal boxing (N = 13). These interactions ceased within 5 min so that introduced workers could finally move undisturbed and were adopted into the alien colony. In contrast, allospecific ants (Leptogenys sp.) were treated aggressively (Kruskal–Wallis test, P < 0.001, N = 128). Resident L. distinguenda workers displayed defensive and aggressive behavior, including bouts of forward jerking (N = 214) and biting (N = 165). Allospecific workers were found dead, although this happened long after the 10-min observation period had ended. In contrast, and similar to conspecific workers, introduced myrmecophiles from other L. distinguenda colonies were treated with lower aggression (Fig. 4a). Both myrmecophilic species sought direct physical contact with resident ants, possibly thus obtaining chemical signatures (Fig. 4b, Kruskal–Wallis test, P < 0.001, N = 82). However, behavioral responses to host aggression differed considerably between the two myrmecophilic species, which explains observed differences in contact frequencies. Spiders remained stationary, tolerated ant dominance behavior, and tried to maintain host contact (N = 58). Silverfish, in contrast, attempted to escape when encountered, a behavior that provoked aggressive chases and attacks by ant workers, similar to the response to introduced allospecifics (N = 8). If resident ants were able to capture an alien silverfish, it was subsequently killed. Thus, the survival of transferred silverfish was zero (N = 3), whereas eight of ten spiders were finally accepted.

Fig. 4
figure 4

Differences in aggressive behavior directed by and contact received by Leptogenys distinguenda workers during 10-min observation periods relative to conspecific workers (LD), allospecific workers (Lsp), and myrmecophilic silverfish (Sf) or spiders (Sp). a Aggression of resident ants toward introduced animals. b Physical contact initiated by introduced animals to resident ants. The plus symbol indicates mean values. Different letters denote significant differences according to a Kruskal–Wallis test, P < 0.001

Full size image

Chemical Integration

We identified 109 different cuticular hydrocarbons among the host and two myrmecophiles (Table 2). Many of the substances occurred in low amounts, so that only characteristic principal compounds were included in the multivariate analysis (see “Methods and Materials”). The analysis showed, however, that spiders and silverfish did not carry any compounds that were uniquely different from those on the host ants (Table 2). Furthermore, the presence or absence of cuticular hydrocarbons on the myrmecophiles depended on their concentrations on the host ants. Only compounds that occurred as traces on the host were found significantly less frequently on the myrmecophiles than expected under the hypothesis of a total match of compounds (P < 0.001, chi-square goodness of fit test, N = 75) (see Supplementary material, Table S2). All compounds that occurred at >0.1% of total peak area were not found less frequently than expected (P > 0.545, chi-square goodness of fit test, N = 119, 27, and 12, respectively).

Table 2 Hydrocarbons identified in workers of the host ant Leptogenys distinguenda and the myrmecophilic spider, Gamasomorpha maschwitzi, and silverfish, Malayatelura ponerophila
Full size table

The hydrocarbon profiles from workers of three different L. distinguenda laboratory colonies showed weak (Fig. 5), but significant, colony-specific structure (ANOSIM, pairwise R values ranged between 0.23 and 0.67, overall P < 0.002). Hydrocarbon differences were more pronounced between colonies of L. distinguenda and Leptogenys sp. (ANOSIM, R > 0.94, P < 0.001; Fig. 5). Further quantification is shown in the Supplementary material (Table S3).

Fig. 5
figure 5

Non-metric multidimensional scaling (NMDS) plot visualizing chemical similarities of three Leptogenys distinguenda colonies (LD11, LD12, and LD13) and one closely related Leptogenys sp. colony (Lsp) as an outgroup. Hierarchical cluster analysis also demonstrated the similarity among groups at the 80% level. N LD11 = 15, N LD12 = 18, N LD13 = 17, and N Lsp = 9

Full size image

The hydrocarbon profiles of myrmecophilic spiders and silverfish from six field colonies resembled the patterns from their host to a high degree; however, colony-specific matching was not observed. This might be due in part to the small sample sizes of the myrmecophiles. More likely and consistent with the results above, L. distinguenda inter-colonial differences are too weak to be detected through their myrmecophilic fauna. Instead of showing colony-specific structure, the profiles of within-nest groups were distinct across all L. distinguenda colonies (R = 0.376, P < 0.001, two-way nested ANOSIM, within-nest groups nested in colony groups). Thus, we further compared chemical similarities among workers, callow adults, larvae, pupae, as well as myrmecophilic spiders and silverfish of field colonies and found a high chemical resemblance of the two myrmecophile profiles with callow and adult host profiles (Fig. 6). Nine cuticular hydrocarbons accounted for more than 98% of the similarity within each of these groups, despite being present in different proportions (see results of SIMPER analysis in the tabular insert of Fig. 6).

Fig. 6
figure 6

Similarity of cuticular hydrocarbon profiles from the ant, Leptogenys distinguenda, the silverfish, Malayatelura ponerophila, and the spider, Gamasomorpha maschwitzi. Total ion chromatograms of (from top to bottom) callow (Ca), adult worker (Ad), myrmecophilic silverfish (Sf), and myrmecophilic spider (Sp). Principal compounds are numbered: 1 9-tricosene, 2 tricosane, 3 tetracosane, 4 6,9-pentacosadiene, 5 9-pentacosene, 6 7-pentacosene and 3,6,9-pentacosatriene, 7 pentacosane, 8 9- and 11-methylpentacosane, 9 9-heptacosene and 6,9-heptacosadiene, 10 7-heptacosene and 3,6,9-heptacosatriene; 11 9-nonacosene and 6,9-nonacosadiene; 12 7-nonacosene and 3,6,9-nonacosatriene. The table insert shows relative contributions of the nine principal compounds to the similarity of within-nest group chemical profiles according to a similarity percentage (SIMPER) procedure (see text for details)

Full size image

Both myrmecophiles were chemically more similar to adult and callow adult host ants than to host ant larvae and pupae, which typically lack most cuticular hydrocarbons (Fig. 7). Based on hierarchical CA, 98% of adults, 97% of callow adults, 84% of silverfish, and 57% of spiders were arranged in a 70% similarity group, whereas only 38% of the larvae and none of the pupae were sorted into the same group. Instead, all pupae, most larvae, and the remaining samples were spread in six separated groups. On a higher similarity level of 80%, there were still 88% of adults and callows, as well as 57% of silverfish, and 21% of spiders sorted into one group. A second group contained few host ants (6% of workers and 9% of callows) and a considerable number of spiders (36%) and silverfish (19%). An ANOSIM revealed more details, which cannot be observed directly in Fig. 7. Spiders were less similar to both callows (R = 0.54) and adults (R = 0.58), whereas silverfish showed higher resemblance to callows (R = 0.12) and adult host ants (R = 0.22) (P < 0.002 for all pairs, ANOSIM). Thus, in the six field colonies, silverfish were chemically more similar to their hosts than were spiders, whether we performed the ANOSIM on selected cuticular hydrocarbon groups only (i.e., unsaturated or saturated) or on the most abundant hydrocarbons.

Fig. 7
figure 7

Non-metric multidimensional scaling (NMDS) plot visualizing chemical similarity of 49 adults (A), 32 callow adults (C), 21 larvae (L), 11 pupae (P), 37 myrmecophilic silverfish (SF), and 14 myrmecophilic spiders (SP) from six different Leptogenys distinguenda colonies (=164 samples). In the hierarchical cluster analysis, the 70% similarity group contained 77% of all samples (=126 samples). Pupae (P) data points are out of range of the shown frame

Full size image

Discussion

Social Integration Strategies of Myrmecophiles

If mimicry is regarded as the resemblance of the properties of a model organism (Dettner and Liepert 1994), our results demonstrate chemical integration by mimicry of host cuticular hydrocarbon patterns for both myrmecophilic species. To the best of our knowledge, this is the first report of chemical mimicry in silverfish. A well-known example of chemical mimicry in spiders is with Cosmophasis bitaeniata, which feeds on larvae of the weaver ant, Oecophylla smaragdina, and acquires a colony-specific chemical signature (Allan et al. 2002; Elgar and Allan 2004, 2006). Nonetheless, social integration of most myrmecophilous spiders remains poorly studied, even though myrmecophily is not uncommon among spiders (Cushing 1997). Our study supports the notion that a passive mechanism of chemical mimicry through maintenance of physical contact (Lenoir et al. 1997, 2001) was used by both myrmecophiles. Specifically, a passive transfer of chemicals predicts a match of major host compounds with only host compounds of low concentrations falling below the detection threshold on the myrmecophiles. Furthermore, it appears evolutionarily unlikely that spiders and silverfish experienced all mutations to biosynthesize all or most of these specific Leptogenys host compounds (Dettner and Liepert 1994). A system of passive chemical mimicry by adoption of nest odor requires a constant renewal of the myrmecophile cuticular hydrocarbon profile, and such efforts were apparent in their behavior. In both species, over 40% of all monitored behaviors were species-specific attempts to get in close contact with host ants. Direct physical contact with legs (spiders) or body surface (silverfish) was observed frequently. Nevertheless, our chemical analysis of free-living colonies demonstrated considerable variation in the chemical profiles within each group of myrmecophiles, suggesting different success in maintaining mimicry among individuals. Chemical dissimilarity even reached such a high degree that several individuals were not grouped together with their host ants by CA. Such significant deviations from the host colony odor could provide sufficient recognition cues for the myrmecophile to be detected and rejected, which also explains the observed killings of myrmecophiles in our laboratory colonies.

In native colonies, silverfish resembled host profiles better than spiders, no matter which class of cuticular hydrocarbons the analysis was based on. This finding suggests that silverfish should thrive well in free-living host colonies, and in fact, they were found in high numbers. In stark contrast to this, silverfish showed strikingly high mortality rates in our laboratory colonies. This could be attributed only to the altered nest structure of our observation nests. Full-sized L. distinguenda colonies show a different and distinct nest structure (Witte 2001). Adult workers occupy peripheral areas of the nest, whereas large piles of pupae and dense aggregations of callow adults with larvae in their mandibles are found in the center. Silverfish prefer these central regions, where they find enough space to hide between brood, and to update their chemical profiles from defenseless callows without disturbance. Furthermore, they are able to hide their offspring in the larval clumps that are located in these central nest regions. Brood and callows were not arranged in such distinct spatial structure in our small laboratory colonies, so that silverfish were probably unable to find their natural niche. Encounters with aggressive mature workers were frequent, which constantly disturbed the silverfish and probably hindered the updating of their cuticular hydrocarbon profiles by physical contact. Consequently, they were detected, attacked, and finally killed. Mortality due to worker aggression was also observed in myrmecophilous silverfish associated with ecitonine army ants (Rettenmeyer 1963). In contrast to silverfish, most of the spiders were accepted in our L. distinguenda laboratory colonies probably due to their behaviorally different integration strategy, which depended less on the nest structure and interactions with defenseless callows. Continuous contacts with callow and adult host ants and frequent rubbing of their filigree legs over the host cuticle while resting directly on their bodies facilitated a constant update of the chemical signature. Spider cuticular profiles were presumably more stable under laboratory conditions, and these myrmecophiles were better suited to remain socially accepted.

Countermeasures by the Host

Even if amplified by our laboratory conditions due to an artificial change in nest structure, the possibility of recognition and aggressive rejection of myrmecophiles is an important finding. This demonstrates that the host ant L. distinguenda is able to detect and regulate its symbiont fauna to a certain extent. Myrmecophiles live under the potential risk of being detected, expelled, or even killed. Accordingly, active host regulation can explain, on the one hand, the overall limited numbers of myrmecophiles within L. distinguenda colonies and, on the other hand, the large variation in myrmecophile abundances. Due to colony-specific circumstances and species-specific integration mechanisms, it might be temporarily more or less challenging for myrmecophiles to exploit a host colony. Although field conditions were different from laboratory conditions, nest demography also varies in free-living L. distinguenda colonies. We suggest that silverfish can also experience integration difficulties under natural conditions (e.g., after demographic changes due to resource shortages, in small colony propagules after nest fission, or simply if a silverfish fails to locate callow adult aggregations in a larger colony or a widely spread nest arrangement). The considerable variation of silverfish abundance (three to 47 individuals) is in accord with this assumption. Although exploitation of colony resources by symbionts can be tolerated when they occur in low numbers, cost of symbionts to the colony increases if they occur in higher abundances. Consequently, there should be selection on the host for symbiont recognition and rejection mechanisms. The question remains to be explained why silverfish occur regularly in higher numbers than spiders in natural L distinguenda colonies. We hypothesize that the reproductive rate of silverfish mainly is responsible, because, in contrast to spiders, we repeatedly detected juvenile silverfish. Furthermore, spiders might encounter stronger intraspecific competition as they frequently showed intraspecific aggressive interactions.

Nestmate Recognition and Aggression

Somewhat surprisingly and atypically for ants, nestmate recognition appeared to be developed weakly in L. distinguenda, which contradicts an effective symbiont recognition system. Both variation in cuticular hydrocarbon profiles between different colonies and the level of inter-colonial aggression were surprisingly low in L. distinguenda. This phenomenon might be explained by the mode of reproduction in this species. Winged males disperse to enter alien colonies for mating, whereas queens are wingless. Similar to our exchange experiments with conspecifics, male intrusion into alien colonies elicits initial aggression that ceases over time so that males are finally able to enter (Witte, personal observation). Furthermore, males show similar profiles to workers (Witte, unpublished data), and this suggests a trade off in chemical discrimination and aggressive behavior in L. distinguenda. Cuticular hydrocarbon profiles might be just sufficiently different to recognize aliens and to maintain colony integrity but not so different to elicit high aggression levels, which in turn would complicate the intrusion and acceptance of males. Nevertheless, social integrity is guaranteed, as our analysis confirmed that chemical differences were greater between different Leptogenys species, accompanied by elevated aggression. In addition to chemical recognition and rejection, there are other lines of defense against social parasites, such as the extraordinarily high migration frequency of L. distinguenda (Maschwitz et al. 1989) that places additional adaptive challenges on the symbiont fauna. However, both spiders and silverfish appear well-adapted in this respect, as they are able to follow host emigration trails independently and without obvious difficulties (Witte et al. 2008). Thus, the mechanisms based on chemical recognition and rejection that we described in this paper appear to be of particular importance concerning the regulation of these myrmecophile species.