Ants have paired metapleural glands (MGs) to produce secretions for prophylactic hygiene. These exocrine glands are particularly well developed in leaf-cutting ants, but whether the ants can actively regulate MG secretion is unknown. In a set of controlled experiments using conidia of five fungi, we show that the ants adjust the amount of MG secretion to the virulence of the fungus with which they are infected. We further applied fixed volumes of MG secretion of ants challenged with constant conidia doses to agar mats of the same fungal species. This showed that inhibition halos were significantly larger for ants challenged with virulent and mild pathogens/weeds than for controls and Escovopsis-challenged ants. We conclude that the MG defence system of leaf-cutting ants has characteristics reminiscent of an additional cuticular immune system, with specific and non-specific components, of which some are constitutive and others induced.
The metapleural glands (MGs), paired structures at the posterolateral margin of the mesosoma, are found only in ants and are one of the defining apomorphies of the family Formicidae [1,2]. The secretions of MGs may have many functions, but the production of anti-microbial compounds for general nest sanitation is the most widespread and general [3–6]. The evolution of glands that produce secretions with generalized prophylactic functions seems a logical adaptation for ants, because colonies are often densely packed and individuals interact continuously with the nest substrate (usually soil) and food (often cadavers), where micro-organisms abound [7,8]. Despite their likely importance for the evolution and diversification of the ants, few studies have investigated the details of MG function. The taxonomic coverage of these studies is scattered, and most have primarily focused on ants with exceptionally derived MG functions [9,10].
The leaf-cutting ants (Attini) are a partial exception to this dearth of information, as MG reservoir size can be measured relatively easily, and these sizes show interesting allometries across worker castes [11–13]. The anti-microbial role of MG secretions in leaf-cutting ants is potentially of particular significance, because these ants have to protect both themselves against entomopathogens, and their mutualistic fungus gardens against parasites and competitors , which appears to have led to significantly enlarged MGs in Atta and Acromyrmex leaf-cutting ants , where the secondary evolution of queen multiple-mating has also allowed considerable genetic variation for MG size to be maintained [15,16].
Studies on MG secretions in attine ants have shown that the production of these secretions is metabolically costly , that they contain both a diversity of carboxylic acids of various chain lengths and proteinaceous compounds [18,19], and that their functioning may be subject to metabolic trade-offs [20,21]. However, specific tests of antimicrobial function have remained rare, mainly due to the technical difficulties of extracting the very small quantities of these glandular secretions [9,18,19]. Hence, the antimicrobial role of MG secretions has usually been inferred rather than measured, or has focused on single compound testing rather than condition-dependent variation in natural MG secretions .
MG secretions were believed to mainly spread passively over the ant cuticle , but a recent study by Fernández-Marín et al.  showed that active acquisition of small quantities of secretion with the front legs followed by focused grooming frequently occurs across the fungus-growing ants. Most bacterial, viral and protozoan disease propagules must be ingested by ants to become infective, but pathogenic fungi enter the ant hosts via the cuticle . This has led to the expectation that MG secretions in leaf-cutting ants may function predominantly as an antifungal defence, a notion that was corroborated by Fernández-Marín et al. [21,23], showing that a number of attine species increase their active grooming rate with MG secretion after being exposed to fungal conidia (asexual spores). The ants also apply MG secretions to fungus garden infections, after which the compromised but treated mycelial debris is stored in the infrabuccal pocket (a cavity just behind the ant mouthparts), where any remaining conidia are likely to be killed by the mixture of MG and labial gland secretions before debris-pellets are discarded [23,24]. However, it has remained unknown whether any infection-induced increase in grooming with MG secretions is accompanied by an increase in quantity or adjustment in the specific potency of the secretions.
The present study examined whether the MG secretions of Acromyrmex octospinosus leaf-cutting ant workers can be adjusted according to the fungal conidia to which the ants are exposed, using five species of fungi representing three broad categories of threat (entomopathogen, generalist saprophyte and fungus garden pathogen).
2. Material and methods
(a) Fungal inoculation experiments
Four representative colonies of Acromyrmex octospinosus, collected in Gamboa, Panama, in 2005 (colony Ao273) and 2007 (colonies Ao404, Ao482 and Ao492), were used in the experiments. Colonies were picked because of their similarity in size, with approximately 1 litre of fungus garden and 50–100 large garden workers being easily available at each of the consecutive bouts of sampling. The colonies were kept under standardized conditions in a climate room at 25°C and 70 per cent relative humidity at the University of Copenhagen. To avoid age- or caste-specific variation in the composition of the MG secretions, we used only large garden workers of approximately the same intermediate age class .
Two entomopathogenic fungi (Beauveria bassiana and Metarhizium brunneum—previously called Metarhizium anisopliae, but now distinguished as a sibling species ), two saprophytic fungi with the potential of causing low pathogenicity (Aspergillus niger and Gliocladium virens) and a specialized parasite that attacks the fungal cultivar of the ants (Escovopsis weberi) were chosen as disease agents. Beauveria bassiana (KVL 03-90) and M. brunneum (KVL 04-57) were obtained from the stock collection of the Department of Agriculture and Ecology, University of Copenhagen, whereas A. niger (DSM 1957) and G. virens (DSM 1963) were purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH. The E. weberi strain was collected from an A. octospinosus nest in Gamboa by Hermogenes Fernández-Marín in early 2010, and identified based on morphological characters .
We first tested the pathogenicity and virulence of these five fungal species on A. octospinosus workers by inoculating ants with fungal conidia and monitoring their survival. From each colony, we collected a random sample of 60 major workers, of which we inoculated half and used the other half as controls. Inoculations were conducted by gently grasping individuals with a pair of sterilized soft forceps, and pipetting 2 μl of fungal (approx. 107 conidia ml−1) suspension in 0.05 per cent Triton-X (to avoid conidia clumping) onto the propodeum. Inoculated individuals were placed separately into plastic pots (diameter 2.5 cm, height 4 cm), where they were maintained at 25°C with an ad libitum supply of 10 per cent sucrose water. The other half of the workers (control group) were treated with 2 μl 0.05 per cent Triton-X solution applied in the same way. Subsequently, ant mortality was assessed daily for 14 days (survival data in electroic supplementary material, table S1). Ants that died during this period were surface-sterilized by rinsing with 70 per cent ethanol followed by 96 per cent ethanol to avoid fungal growth from external contaminants. The surface-sterilized ants were placed in a plastic pot lined with damp filter paper. Cadavers were inspected after 10 days to score the presence or absence of hyphal growth on the cuticular surface, and to check whether these hyphae produced conidia of the fungus with which the ants were inoculated (sporulation data in electronic supplementary material, table S2). The fungal inoculations and controls were carried out simultaneously so that ant mortality assessment and cadaver inspections were completed within two weeks from the start of the experiments.
The workload of the experiments was such that the treatments could only be handled in parallel by same person (the first author) for two colonies at a time. The experiments with colonies Ao273 and Ao404 were therefore carried out six months before the experiments with colonies Ao482 and Ao492. This reduced error variance as the variance across colonies could be partialled out in subsequent statistical analyses (where it turned out to be insignificant).
(b) Extraction and quantification of metapleural glands secretions
To examine the quantitative responses of MG secretion in inoculated ants, we set up a second experiment structured as two-way repeated measures factorial design, to measure the amount of MG secretions from inoculated and control ants for each of the five different fungal species. To inoculate ants with each of the five fungal species and obtain sufficient amounts of the glandular secretions after these treatments, we took out 30 major workers from each colony at a time and divided these randomly into six groups of five individuals to be inoculated with the five different kinds of conidia or to serve as controls.
The inoculation procedure was as mentioned before, but the ants were killed by freezing in liquid nitrogen 12 h after inoculation, and gland extractions were carried out immediately after killing. Dipping in liquid nitrogen serves two purposes in glandular extractions: (i) to prevent passive flow or active grooming of MG secretions during the extraction process, and (ii) to halt the chemical reactions of MG secretions so that all secretions collected in the reservoir (a sclerotized atrium where glandular secretions were stored ) represented the same state (after fungal inoculations or control solution application) at the time of extraction. The amount of secretion was measured by piercing the MG reservoir (externally visible as the bulla ) with a very fine insect-mounting needle, after which a graduated 10 μl Hamilton syringe was inserted to extract accumulated secretion. Secretions from both reservoirs of the paired glands were extracted in the same syringe (i.e. pooled), and their volume recorded (MG quantity data in electronic supplementary material, table S3). To ensure that all secretions present were extracted with our method, we dissected the reservoir from a sample of ants after glandular extractions with a fine razor blade and examined the reservoir under the microscope. These reservoirs were found to be empty and dry . Similar dissections were carried out on a sub-sample of ants prior to glandular extractions, and these reservoirs were found to be either wet or with secretions ‘oozing’ from the punctured point. Extraction and quantification procedures were done for all five individuals of the same treatment group in quick succession (usually spanning 10–20 min). Secretions extracted from the same treatment for all five individuals were deposited as droplets on a sterilized microscope slide, after which 10 μl of the pooled sample of MG secretions was collected and immediately dissolved in 1 ml of pentane, capped to prevent evaporation, and used for testing the quality of these MG secretions.
(c) Testing quality of metapleural gland secretions after challenging with fungal conidia
The efficacy of MG secretions in inoculated and control ants was measured with in vitro inhibition assays measuring zone of inhibition. All five fungal species were grown on potato dextrose agar (PDA) plates. Conidia (asexual spores borne on hyphae) were harvested, cleaned and diluted to approximately 105 conidia ml−1, a concentration low enough to prevent fungal overgrowth that could mask inhibition halos. Prior to using the fungal species for these inhibitive bioassays, assessments of conidia germination were carried out to ensure that the fungal conidia were viable. All fungal species had more than 90 per cent germination (M. brunneum 99%; B. bassiana 96%; A. niger 97.8%; G. virens 93.6%; E. weberi 91.5%).
To achieve an even fungal growth, 1 ml volumes of fungal conidia suspensions were pipetted onto the surface of PDA plates and evenly spread out using a Drigalski spatula. PDA plates were divided into six equal segments, and a disc of sterilized filter paper (5 mm diameter) was placed at the centre of each segment. About 10 μl of the pooled MG secretion extracted from ants after inoculation was dissolved in 1 ml pentane (see §2b) and 10 µl was then pipetted onto the paper discs. The sixth segment was used as a control and received 10 μl of pentane. To examine whether the inhibitive effects of MG secretions were due to enzymatic activities, we performed a sub-sample of inhibition assays (MG secretions from ants treated with M. brunneum and control solution) by incubating the MG secretions in a water bath at 100°C for 5 min before application to denature the proteinaceous components.
All agar plates were incubated at 25°C for 24 h, after which the halos in the fungal mats caused by the inhibitive action of the applied secretions were measured (see panel in figure 3). The zone-of-inhibition assay was replicated on ten PDA plates for each fungal species (inhibitive halo radius data in electronic supplementary material, table S4).
(d) Statistical analyses
The survival of inoculated ants over 14 days was analysed using a proportional-hazards model, with colonies, fungal treatments and their interaction as main effects. Surviving individuals were included as censored cases. Post hoc pairwise differences between colonies and fungal treatments were based on risk-ratio tests, with the significance level being adjusted with the Bonferroni procedure to correct for multiple comparisons. The proportion of sporulating ants was analysed using a generalized linear model with binomial error structure, with fungal treatments and colonies as main effects.
Differences in the amounts of MG secretion produced by ants inoculated with different fungal conidia were analysed using two-way ANOVA, testing for differences between both colonies and fungal treatments, with each ant providing a data point consisting of pooled secretion volumes of its two MGs. In the case of overall significance, post hoc multiple-comparison Tukey's tests were performed to examine which treatments made ants produce significantly more MG secretion. Approximately one-third of the worker MG reservoirs have previously been reported as being filled with secretion under unchallenged conditions , so we assumed that we would be able to measure any increases of secretion volume relative to such controls.
To test the differences in inhibitive efficiency between MG secretions from ants inoculated with different fungal species, we used a repeated-measures multivariate analysis of variance (MANOVA), with halo area of fungal bioassay species as the dependent variable set, and fungal species and colony as independent variables. This allowed us to evaluate whether the identity of inoculated fungal species had an effect on the inhibitive activity of MG secretions, and whether any such effect was specific to the fungus tested.
To test the differences in inhibitive efficiency between boiled MG secretions and non-heated MG secretions, we performed a Student's t-test for the available comparisons. Analyses were carried out using jmp software (v. 9.02, SAS institute).
The survival of ants differed significantly between the fungal treatments (likelihood-ratio, LR: χ2 = 654.18, d.f. = 5, p < 0.0001), with treatment explaining approximately 60 per cent of the variation in survivorship (R2LR = 0.597). Ants inoculated with the entomopathogenic fungi M. brunneum and B. bassiana consistently suffered greater mortality than those inoculated with the control solution. Ants inoculated with A. niger also consistently suffered greater mortality than those inoculated with the control solution, but significantly lower mortality compared with ants inoculated with both entomopathogenic fungi (figure 1). Mortalities of ants inoculated with G. virens and E. weberi did not differ significantly from mortalities of ants inoculated with the control solution (figure 1), confirming previous findings of E. weberi conidia not being harmful to ants . The overall survival of ants differed significantly between the four experimental colonies (LR: χ2 = 14.47, d.f. = 3, p = 0.0023), reflecting resistance variation between colonies. There was also a significant interaction between colony and fungal treatment (LR: χ2 = 36.41, d.f. = 15, p = 0.0015), reflecting small but consistent differences in how each colony responded to each fungus (figure 1), although colony and the interaction between colony and treatment each explained less than 5 per cent of the variance in survival (R2LR = 0.020 and 0.049, respectively). The proportion of dead ants sporulating differed significantly between the fungal treatments (LR: χ2 = 150, d.f. = 5, p < 0.0001), but not between experimental colonies (LR: χ2 = 8.58, d.f. = 3, p = 1.0000). No sporulation was detected for G. virens, E. weberi and control treatments, except minor contamination from unknown saprophytic fungi that also occurred in the controls and were therefore ignored. However, all cadavers from ants exposed to B. bassiana and M. brunneum, and (depending on colony) between 33 and 63 per cent of the cadavers of ants exposed to A. niger, produced characteristic conidia.
Workers from the four colonies did not differ significantly in the quantities of MG secretion that their large workers produced (F3,120 = 0.08, p = 0.97), nor in their respective responses to the different fungal treatments (F15,120 = 0.77, p = 0.70), so data were pooled for presentation in figure 2. There were, however, significant differences in the amount of MG secretion produced following the different inoculation treatments (F5,120 = 19.45, p < 0.0001), with B. bassiana and M. brunneum treatments eliciting about twice as much MG secretion compared with the controls. Treatments with A. niger, G. virens and E. weberi elicited significantly less secretion than M. brunneum and B. bassiana, and did not differ significantly from the controls (Tukey's tests; figure 2).
Across all five treatment and control groups, we extracted on average 3.6 ± 0.1 μl (mean ± s.e.) of secretion from 120 workers MG reservoirs. From the control treatments, we extracted 2.3 ± 0.1 μl of secretion from 20 worker MG reservoirs. Using the scale of MG reservoir content developed by Poulsen et al. , we estimated that these corresponded to approximately one-third-filled reservoirs on average, whereas the 40 M. brunneum- and B. bassiana-challenged ants had 4.8 ± 0.3 μl and 4.8 ± 0.3 μl of secretion in their respective reservoirs, indicating that they were approximately two-third-filled on average. The 40 ants inoculated with A. niger and G. virens had 3.7 ± 0.1 μl and 3.0 ± 0.2 μl of secretion in their respective reservoirs, which amounts to them being roughly half-filled. The 20 E. weberi inoculated ants had 2.8 ± 0.2 μl of secretion in their reservoirs on average, which was not significantly different from the control ants. In our total sample, four ants (3%) had an empty reservoir for unknown reasons: one each from the M. brunneum, B. bassiana, G. virens and E. weberi treatment group (figure 2). Using our volume approximation, four workers had their reservoir almost fully filled with secretion (approx. 6.5 μl)—three from the B. bassiana and one from the M. brunneum treatment group (figure 2)—indicating an induced response to infection with entomopathogenic fungi.
There were highly significant differences in the efficiency of antifungal activity of MG secretions from ants treated with different fungal species (table 1 and figure 3). Secretions taken from ants inoculated with B. bassiana, M. brunneum, A. niger and G. virens exhibited higher antifungal activity compared with MG secretions taken from E. weberi and control treatments. There were no significant between-colony differences in efficacy of MG secretions or interactions between colony and fungal treatment, indicating that the response was colony-independent. Across colonies, however, there were significant differences in the sensitivity of the different fungi to MG secretions (table 1), with G. virens being somewhat more sensitive and B. bassiana less sensitive to MG secretions than the other species (figure 3). There was also a significant interaction between the species of fungus to which worker ants were exposed and the sensitivity of the fungus to the MG secretions of those ants (table 1), which was primarily due to a relatively constant response of the bioassay fungi to MG secretions for controls and ants treated with E. weberi, and a different pattern of efficacy of MG secretions from ants inoculated with M. brunneum or B. bassiana (figure 3).
There were no significant differences in the inhibitive efficiency between boiled MG secretions and non-heated MG secretions (t = 0.94, d.f. = 1, p = 0.350 for boiled control versus control; t = −0.77, d.f. = 1, p = 0.443 for boiled secretions treated with M. brunneum inoculations versus non-heated secretions treated with M. brunneum inoculations), indicating that the antifungal ingredients were not enzymatic. This result concurs with an earlier study of MG secretion antibiotic properties in Myrmecia gulosa ants .
(a) Functional plasticity in metapleural gland responses
Infections with virulent M. brunneum and B. bassiana entomopathogens triggered both more MG secretion and more efficient inhibition per unit of MG secretion. These results are consistent with those of Bot et al. , who showed that both conidia and hyphae of these pathogens are inhibited similarly by seven different classes of chemical compounds from the MG secretions of A. octospinosus. The mild pathogen A. niger has an interesting intermediate position, as infections resulted in more potent MG secretions (figure 3) but without increasing the secretion volume (figure 2). This may be related to conidia germination, but not hyphal growth, of both A. niger and G. virens being less efficiently inhibited by several compounds in the MG secretion than conidia of the two virulent entomopathogens .
While previous authors [28,31] concluded that ant MGs seemed to be somewhat disappointingly simple, because no muscles were present to directly regulate the emergence of secretion to the cuticular surface, our present data suggest that the production of MG secretion is remarkably plastic and appears to be both quantitatively (figure 2) and qualitatively (figure 3) adjusted to specific fungal infection threats. Given these conditional responses, the question now seems almost the reverse: why have behaviourally controlled muscular release mechanisms for MG secretion not been favoured by selection?
Our results suggest that the MGs of leaf-cutting ants may function as analogues of a simple prophylactic immune system that operates on the cuticular surface of ant workers. This hypothesis is based on the evidence provided here that MG secretions have both general and specific functions, and that their production is partly constitutive and partly induced. We will discuss this hypothesis below, using a modified version of a framework diagram from Schmid-Hempel & Ebert  that maps immune defences along perpendicular specificity and inducibility axes (figure 4). We also include our current understanding of MG chemistry and function in Atta leaf-cutting ants, which are the sister genus of Acromyrmex, but different in several key life-history traits. We use this analogy with the immune system proper because it helps to interpret our findings in a coherent conceptual framework emphasizing evolutionary adaptation, not because we would claim that the mechanisms and metabolic pathways involved are similar or even related to those of the normal immune system of these insects.
(b) Constitutive, non-specific defence components of leaf-cutting ant metapleural glands
The bottom-left quadrat of figure 4 summarizes the most basal antimicrobial MG functions, which were the first to be discovered . General antimicrobial activity probably stems from the high acidity (pH 2.5–4) of MG secretions [3,9,30]. This acidity and its pH-reducing effects in fungus gardens are well documented in a series of studies on leaf-cutting ants [5,33], and are probably due to the abundant presence of organic acids in MG secretions .
The second component in this quadrat is that at least some secretion is normally present in the MGs of all ants, although the amount in the reservoirs of unchallenged ants varies, probably as a function of nutrition and age , so that some ants (3% in our study) end up having reservoirs that are scored as empty [3,17]. There is thus almost always secretion present for the ants to work with, even under benign circumstances, and this constitutive defence amounts to about one-third of the maximal holding capacity of the MG reservoirs, on average  (figure 2).
The third component is the direct supply of MG secretion to the external environment via the rather large opening of the reservoir . The lack of any autonomous or behaviourally controlled muscles to control release of the secretion  indicates a constitutive and non-specific defence component of the MGs.
(c) Induced, non-specific defence components
The bottom-right quadrat of figure 4 summarizes the two known mechanisms by which the application of MG secretion for non-specific defence can be induced. The first is active grooming. The location of the MG openings just above the hind-legs allows secretions to be picked up by leg movements as soon as they emerge. This form of dispersing MG secretion  is particularly employed when leaf-cutting ants are challenged with fungal conidia [21,23].
The frequency of MG grooming and the target of application is known to vary between Acromyrmex and Atta, with Atta substantially increasing their MG grooming rate and expanding their grooming targets to the garden, and Acromyrmex maintaining much lower MG grooming rates and primarily targeting the brood . This is consistent with Atta workers relying more on MG grooming and less on antibiotic production by mutualistic actinomycete bacteria, whereas the opposite combination appears to apply in Acromyrmex  (see §4d for more details).
(d) Induced, specific defence components
The top-right quadrat of figure 4 summarizes two known mechanisms in Acromyrmex and Atta leaf-cutting ants by which MG secretion appears to be differentially induced depending on the type of challenge. The fact that A. octospinosus workers significantly increase the quantity of MG secretion only after being challenged with directly lethal insect pathogens such as B. bassiana and M. brunneum indicates that the ants are able to classify pathogens according to the degree of threat to themselves and their nest-mates. The same results (figure 2) suggest that this recognition process is a continuum, rather than an all-or-nothing response, as the amount of MG secretion was also elevated somewhat after challenges with G. virens and A. niger.
Interestingly, induction specificity also included increases in potency of the MG secretion, and here responses were generally similar for all fungal infections except Escovopsis, in spite of the variation in threat across these challenges (figure 3). This underlines that the MG secretion responses are analogous to immune system functioning, because challenges both increase overall investment in defence (cf. lymphocytes) and induce the production of specific defence agents (cf. antibodies) after the kind of challenge has been identified . There was difference in how the fungal species responded to MG secretions that were induced by particular fungal conidia (table 1), which suggests that there is potential for changing MG secretions to target particular threats. However, we found no match between which fungal species had induced a particular MG secretion and the susceptibility of that species to the secretion (figure 3). From our heat-denaturing testing and an earlier study , we infer that proteinaceous/enzymatic compounds are unlikely to be important as antifungal ingredients. Hence, the identities of compounds that make MG secretion more potent after infections remain unknown, but will be interesting targets for future research.
The specific defence responses obtained in our present study are similar to those obtained by Fernández-Marín et al. [21,23], who showed that MG grooming rates in Atta colombica were elevated when inoculated with conidia of Metarhizium and Escovopsis but not by inert talcum powder controls. Whether this response is mediated by quantitative and/or qualitative change in MG secretion is unknown, but it seems reasonable to infer that an increase in grooming rate would only make sense if at least the amount of available MG secretion would also be proportionally increased. There is a significant difference between Atta and Acromyrmex in that MG grooming in the former also has a major function for infection control in the fungus garden, whereas this is not so in the latter . This is consistent with Acromyrmex maintaining cultures of actinomycete bacteria on their exoskeleton to control Escovopsis infections, whereas Atta has abandoned this form of biological control (possessed by most basal attine ants ) in exchange for chemical control via MG grooming . It is therefore not surprising that Atta has inducible specific MG defences against Escovopsis and Acromyrmex has not.
(e) Constitutive, specific defence components
The upper-left quadrat in figure 4 largely represents a ‘black box’, but there is enough evidence for this defence component to hypothesize that this aspect of the immune system analogy is also likely to be important in fungus-growing ants [5,19]. There are major compounds of MG secretions that are likely to qualify as specific constitutive defences, which are different in Atta and Acromyrmex in spite of other compounds being present in the MG secretions of both genera . The prime candidate compound in Atta is phenylacetic acid, which makes up 72 to 80 per cent of the total MG secretion, but is absent in the MG secretion of Acromyrmex. However, Acromyrmex MG secretion has indoleacetic acid (IAA) as a major component (24–25% of total secretion), whereas this compound is found only in trace amounts in the MG secretion of Atta. Further work will be needed to unravel the specific targets of these major components. The role of IAA seems particularly intriguing, as this compound is a well-known plant growth hormone , but initial tests have shown that it does not inhibit fungal conidia or hyphae .
It has been repeatedly found that in addition to chemical compounds that make MG secretion acidic, there is a significant fraction of proteinaceous compounds of unknown identity in the MG secretions of leaf-cutting ants [3,18]. However, the results of our boiling assay indicate that any such proteins do not seem to have the specific functions found in other insects [37–39]. This underlines that secretions of the MGs of leaf-cutting ants might have a combination of adaptive traits that are specific for this clade of ants only.
We thank Louise Lee Munk Larsen for technical assistance in rearing the fungal species and performing the infection assays, Michael Poulsen for sharing his skills on working with Escovopsis weberi, and the reviewers for their comments. All authors were supported by a grant from the Danish National Research Foundation.
- Received June 25, 2012.
- Accepted August 2, 2012.
- This journal is © 2012 The Royal Society