Chemical and behavioural strategies along the spectrum of host specificity in ant-associated silverfish

Abstract

Background: Host range is a fundamental trait to understand the ecological and evolutionary dynamics of symbionts. Increasing host specificity is expected to be accompanied with specialization in different symbiont traits. We tested this specificity-specialization association in a large group of 16 ant-associated silverfish species by linking their level of host specificity to their degree of behavioural integration into the colony and to their accuracy of chemically imitating the host's recognition system, i.e. the cuticular hydrocarbon (CHC) profile.

Results: As expected, facultative associates and host generalists (targeting multiple unrelated ants) tend to avoid the host, whereas host-specialists (typically restricted to Messor ants) were bolder, approached the host and allowed inspection. Generalists and host specialists regularly followed a host worker, unlike the other silverfish. Host aggression was extremely high toward non-ant-associated silverfish and modest to low in ant-associated groups. Surprisingly, the degree of chemical deception was not linked to host specificity as most silverfish, including facultative ant associates, imitated the host's CHC profile. Messor specialists retained the same CHC profile as the host after moulting, in contrast to a host generalist, suggesting an active production of the cues (chemical mimicry). Host generalist and facultative associates flexibly copied the highly different CHC profiles of alternative host species, pointing at passive acquisition (chemical camouflage) of the host's odour.

Conclusions: Overall, we found that behaviour that seems to facilitate the integration in the host colony was more pronounced in host specialist silverfish. Chemical deception, however, was employed by all ant-associated species, irrespective of their degree of host specificity.

Keywords: Host specialization; Hydrocarbon - inquiline; Messor - myrmecophily; Specialization; Symbiont.

Fig. 1

Heat map displaying the behaviour of ant hosts towards different species of silverfish (ant behaviour) and the behavioural repertoire of the silverfish (silverfish behaviour, see also Table 2). The hosts of the silverfish are abbreviated (following Table 1). Some silverfish species were associated and tested with multiple hosts. N number of behavioural trials per host-silverfish pair. The darkness of the colour in the heat map positively correlates with the frequency that a behaviour in a silverfish-ant pair is observed compared to other silverfish-ant pairs. Significant differences (P < 0.05) among the four functional groups for each behaviour are indicated with an *

Fig. 2

CHC similarity among silverfish and host ants. NMDS plot displays the Bray-Curtis similarities for all detected CHCs (N = 199). Filled symbols represent the host ants (congeneric species have the same colour), open symbols represent silverfish associated with ants. For clarity, the seven Messor specialist species are not specified on this plot but grouped as Messor specialists (green open symbols). A detailed NMDS plot with each Messor specialist species specified is given in Additional file 21. The colour and shape of the silverfish symbols correspond with the colour and shape of their host ant. Species identity of myrmecophilous silverfish is given on the plot, except for Proatelurina pseudolepisma which is indicated with a letter code (PP) for clarity. Unassociated silverfish are represented with a letter code: Allacrotelsa kraepelini (AK), Ctenolepisma ciliatum (CC), Ctenolepisma nicoletii (CN), Lepisma baetica (LB – one individual not associated with ants), Ctenolepisma guadianicum (CG)

Fig. 3

A hierarchical cluster analysis (average linkage method) on the matrix with the Bray Curtis similarity distances between the CHC profiles of the ants and silverfish. To avoid overloading of the tree, we grouped the samples of the silverfish found with the same host species and the ant samples per species (by averaging the BC similarites in the matrix), number of samples in each group/branch in brackets. To assess the statistical support of the clusters, we applied multiscale bootstrapping (1000 bootstraps) with the modified pvclust package for the Bray-Curtis similarity matrix. The approximately unbiased P-values are given for each cluster. Values greater than 95% are considered significant

Fig. 4

Representative plastic hydrocarbon profiles of the host generalist Neoasterolepisma curtiseta associated with different hosts: a) Camponotus pilicornis, b) Aphaenogaster iberica, c) Iberoformica subrufa and d) Tapinoma nigerrimum. Peak identities of the CHCs can be found in Additional file 19

Fig. 5

Representative cuticular hydrocarbon chromatogram of the Messor specialist Tricholepisma aureum, a moulted T. aureum individual, and the host Messor barbarus. Peak identities of the CHCs can be found in Additional file 19. The dissimilarities in the CHC profiles are displayed with a NMDS plot. Silverfish are represented by coloured circles (grey: associated with the host, blue: isolated individual and moulted) around a letter code. Ant individuals are depicted by a letter code without coloured circle. The letter code refers to the host colony (see Table 3)

Fig. 6

Moulted individuals of the generalist N. curtiseta have different CHCs than their host (compare with the matching profiles of ant-associated N. curtiseta in Fig. 4). Peak identities can be found in Additional file 19. The peak labelled as non-HC is not a hydrocarbon, but likely a steroid

Fig. 7

Bar plot comparing the CHC concentrations (±SE), i.e. total CHC mass (μg) / dry body weight (mg), for all analysed silverfish. The degree of host specificity of the silverfish is indicated with a colour code