Leaf-cutting ants (Hymenoptera: Formicidae) have a mutualism with the fungus Leucoagaricus gongylophorus (Heim) Moeller (Agaricaceae), the primary food source of the colony (Schultz et al. 2005; Hölldobler & Wilson 2009a). These ants select and cut several plant species for the cultivation of their symbiotic fungus (Weber 1972). Consequently, some species of ants are known principally as pests of crops in the Neotropical region (Della Lucia & Souza 2011). Plant selection involves the proper choice of plant species that are suitable for development of the symbiotic fungus (Hölldobler & Wilson 1990), although the existence of an ant-fungus communication for the correct selection of plants is only hypothesized (Ridley et al. 1996; North et al. 1999; Green & Kooij 2018).

Studies investigating the existence of this communication have used a protein synthesis inhibitor, cycloheximide (CHX) (Obrig et al. 1971; Schneider-Poetsch et al. 2010). This substance, reported to be a fungicide, has been applied in behavioral studies on leaf-cutting ants. The effect of cycloheximide on the fungus garden was found to be responsible for changes in the foraging behavior of leaf-cutting ant workers, because cycloheximide is not identified as a harmful compound for the colony in the first contact with the workers (Ridley et al. 1996). North et al. (1999) hypothesized that a semiochemical emitted by the fungus garden regulates the selection of plants by leaf-cutting ant foragers, and is interpreted as communication between the symbiont fungus and leaf-cutting ant workers. The authors showed that the ants learn to reject plant material that contains chemicals injurious to the fungus. After an initial period of acceptance, ants from laboratory nests stopped harvesting granular bait containing a fungicidal agent (cycloheximide) and orange peel. Subsequently, many behavioral studies have used cycloheximide to understand the effects on the long-term olfactory memory of leaf-cutting ants. Some of these studies support the existence of an ant-fungus communication through a semiochemical volatile compound that has not yet been identified and whose presence is only hypothesized (Cazin et al. 1989; Ridley et al. 1996; North et al. 1999; Herz et al. 2008; Saverschek et al. 2010; Saverschek & Roces 2011; Thiele et al. 2014; Falibene et al. 2015; Arenas & Roces 2016a, b, 2017).

The objective of our study was to test the hypothesis proposed by North et al. (1999). This hypothesis suggests that a volatile semiochemical emitted by the fungus garden regulates the selection of plants by leaf-cutting ant foragers, which is interpreted as a communication between the symbiotic fungus and leaf-cutting ant workers. Specifically, we sought to identify the volatile compounds released by the fungus garden of leaf-cutting ants and to determine the behavioral response of workers to healthy and unhealthy fungus gardens.

Materials and Methods

CHOICE BIOASSAYS

Ant responses to healthy and unhealthy fungus gardens: Y-shaped choice system

To determine whether leaf-cutting ant workers are attracted to healthy and unhealthy fungus gardens, we monitored the worker choice in 6 experimental bioassays, 12 times per bioassay: (1) d 1, healthy fungus vs. healthy fungus; (2) d 1, healthy fungus vs. no fungus; (3) d 7, unhealthy fungus vs. healthy fungus; (4) d 7, unhealthy fungus vs. no fungus; (5) d 14, unhealthy fungus vs. healthy fungus; (6) d 14, unhealthy fungus vs. no fungus.

A simple Y-shaped choice system made of transparent glass was used, which consisted of branches of equal length (40 cm) and width (3 cm), with the upper branches arranged at an angle of 60° (Fig. 2). The worker container was coupled to the base branch of the Y trail. A plastic tray was installed on each upper branch where the choices were provided randomly in a plastic container.

Fifty ants with a mean head width of approximately 2.2 mm were introduced into the working container to allow bioassay choices containing 0.5 g (500 mg) of clean fungus, after removing ants, eggs, larvae, and pupae from the fungus. The number of workers was counted at 5 min intervals for 30 min, corresponding to the worker choice in each container. Each bioassay was repeated 12 times.

Ant responses to extracts of healthy and unhealthy fungus: simultaneous choice

The fungus extracts were obtained using pieces of the fungus from the colonies that received cycloheximide pellets and cycloheximide-free pellets as described above. The fungus pieces were cleaned with needles, removing all brood and workers, and about 500 mg of fungus was used (Viana et al. 2001). These fungus gardens were immersed in 5 mL dichloromethane for 2 h, filtered with filter paper (Qualy®, São José dos Pinhas, Paraná, Brazil), and kept at low temperature. These extracts were used in the bioassays.

Fig. 1.

Fungus garden. A: Fungus healthy; B: fungus with incorporation of pellets with cycloheximide on d 7 of the experiment; and C: fungus on d 14 of the experiment.

To determine whether leaf-cutting ant workers are attracted to healthy and unhealthy fungus extracts, filter paper disks (Qualy®, São José dos Pinhas, Paraná, Brazil) (0.6 cm diam) with the extract were offered in the foraging arena in 5 treatments: (1) 10 disks with healthy fungus extract; (2) 10 disks with unhealthy fungus extract from d 7; (3) 10 disks with unhealthy fungus extract from d 14; (4) 10 disks with dichloromethane; and (5) 10 disks without dichloromethane and extracts as control.

Fig. 2.

Olfactometer model used in the experiment of ant responses to healthy or unhealthy fungus in the Y-shaped choice system.

Four colonies were used and 7 trials were conducted per colony, totaling 350 disks offered per colony. The bioassay was terminated when the treatment was completely transported by foragers into the colony. The total number of transported disks was counted for each treatment.

Results

The fungus garden emitted 4 volatile substances, 2-methyl-propanal, 3-methylbutanal, pentanal, and hexanal, all of which are aldehydes (Fig. 3). The proportion of aldehydes was similar in the 4 colonies that received cycloheximide pellets and the colonies that received cycloheximide-free pellets (Fig. 4). We found no significant differences in the proportion of 2-methylpropanal [d 1 (U = 1.29; n = 4; P > 0.01), d 7 (U = 2.16; n = 4; P > 0.01), d 14 (U = 1.87; n = 4; P > 0.01)]; 3-methylbutanal [d 1 (U = 1.01; n = 4; P > 0.01), d 7 (U = 2.02; n = 4; P > 0.01), d 14 (U = 1.88; n = 4; P > 0.01)]; pentanal [d 1 (U = 2.30; n = 4; P > 0.01), d 7 (U = 0.86; n = 4; P > 0.01), d 14 (U = 0.86; n = 4; P > 0.01)]; or hexanal [d 1 (U = 0.57; n = 4; P > 0.01), d 7 (U = 0.10; n = 4; P > 0.01), d 14 (U = 1.29; n = 4; P > 0.01)] between colonies that received cycloheximide pellets and those that received pellets without cycloheximide.

Fig. 3.

Volatiles emitted by the fungus. A: Healthy fungus; B: fungus without cycloheximide for 7 d; C: fungus without cycloheximide for 14 d; D: fungus with cycloheximide for 7 d; E: fungus with cycloheximide for 14 d.

Fig. 4.

A: Mean percent area and standard deviation (confidence interval) of volatiles emitted by the fungus; B: healthy fungus and fungus with cycloheximide for 7 d; C: healthy fungus and fungus with cycloheximide for 14 d.

Analysis of ant responses to healthy and unhealthy fungus gardens using a Y-shaped choice system (Fig. 2) revealed a strong attraction to the fungus garden, regardless of its health status (Fig. 5). We observed significant differences between healthy fungus vs. no fungus on d 1 (U = 42; n = 12; P < 0.01), between unhealthy fungus vs. no fungus on d 7 (U = 28.8; n = 12; P < 0.01), and between unhealthy fungus vs. no fungus on d 14 (U = 41.5; n = 12; P < 0.01). On the other hand, there were no significant differences between healthy fungus vs. healthy fungus on d 1 (U = 0.54; n = 12; P > 0.01), between unhealthy fungus vs. healthy fungus on d 7 (U = 24.8; n = 12; P > 0.01), or between unhealthy fungus vs. healthy fungus on d 14 (U = 0.054; n = 12; P > 0.01).

As expected, a significant difference was detected among treatments with healthy and unhealthy fungus extracts (Fig. 6) (Kruskal-Wallis test, H = 27.15; df = 4; P < 0.001). The disks with healthy fungus extract were highly transported by foragers (175 disks), with a significant difference compared to disks with unhealthy fungus extract on d 7 (59 disks, P < 0.001) and disks with unhealthy fungus extract on d 14 (89 disks, P < 0.001). However, the number of transported disks with healthy fungus extract was not significantly different from the number of disks with dichloromethane (155 disks, P = 0.5542), or of disks without dichloromethane and extracts (182 disks, P = 0.7858).

Discussion

One of the major findings in this study is that the aldehyde compounds released by the fungus garden were the same in colonies that received cycloheximide pellets and colonies that received cycloheximide-free pellets on d 1, 7, and 14 (Fig. 3). This result does not corroborate the study suggested by North et al. (1999) that a volatile semiochemical emitted by the fungus garden regulates the selection of plants by leaf-cutting ant foragers, which is interpreted as a communication between the symbiotic fungus and leaf-cutting ant workers. They showed that the ants learn to reject plants containing chemicals that are harmful to the fungus. After an initial period of acceptance, ants from laboratory nests stopped harvesting granular bait containing a fungicidal agent (cycloheximide) and orange peel. In contrast, studies show that cycloheximide causes mortality to leaf-cutting ant and fungal workers, and thus may promote pellet rejection due to worker mortality and fungus garden breakage (Sousa et al. 2017).

Approximately 250 volatile organic compounds have been identified in 129 fungal species (126 Macromycetes and 3 Micromycetes pathogenic for plants: Puccinia), where they occur as aldehydes, ketones, alcohols, and phenols, among many others (Chiron & Michelot 2005). However, in our study, the fungus garden emitted 4 volatile substances, 2-methylpropanal, 3-methylbutanal, pentanal, and hexanal, which are aldehydes. The flowers, vegetative parts, and roots of plants are known to release more than 1,000 volatile organic compounds, principally 6-carbon aldehydes, alcohols, esters, and various terpenoids (Pichersky et al. 2006). In some cases, plants release volatile compounds when damaged by herbivores (Tumlinson et al. 1999). In the case of leaf-cutting ants, the leaves of plants are cut, processed into a pulp, and incorporated into the fungus garden surface by workers (Camargo et al. 2007; Garrett et al. 2016). It is likely that these leaf fragments in the fungus garden emit volatile substances, as observed in mechanically damaged leaves of poplar cuttings (Populus simonii Carrière and P. pyramidalis Salisb.) (Salicaceae) (Hu et al. 2008). These authors identified 16 aldehydes: acetaldehyde, butanal, pentanal, hexanal, heptanal, octanal, nonanal, decanal, undecanal, dodecanal, tetradecanal, (Z)-3-hexenal, (E)-2-hexenal, (E)-2-nonenal, benzaldehyde, and furfural. Interestingly, most of these substances were released 24 h after wounding (Hu et al. 2008), suggesting that the same may occur in the fragmented leaves of the fungus garden of leaf-cutting ants.

It is known that unhealthy plant parts, mechanical damage, pathogen infections, and herbivore injuries can induce the rapid synthesis of volatile compounds such as hexenal and hexanal (Arimura et al. 2000). Analogously, the symbiotic fungus degrades leaves as a substrate for its colonization, and may induce the release of some aldehydes. Leucoagaricus gongylophorus is a basidiomycetous fungus that produces specialized hyphae (gongylidia) for feeding ants and larvae (Hölldobler & Wilson 2009), as well as enzymes for lignocellulose degradation (Boyd & Martin 1975; Martin et al. 1975; Schiøtt et al. 2008, 2010; Aylward et al. 2013; Grell et al. 2013; Kooij et al. 2014, 2016). The degradation of lignin by basidiomycetes produces a variety of extracellular aromatic metabolites (Jong et al. 1994), especially aldehydes (Gallois et al. 1990) such as hexanal, heptanal, 2-butenal, 2-methyl-2-butenal, 4-nonenal, and 2,4-decadienal. These findings support the production of some aldehydes by the fungus garden during lignocellulose degradation.

The worker responses to healthy and unhealthy fungus gardens indicated an attraction to the fungus garden (Fig. 5). This attraction might be attributed to the presence of ant hydrocarbons and large quantities of n-alkanes in the fungus garden that constitute a signal for leaf-cutting ant workers (Viana et al. 2001). This fact also may explain why the workers did not distinguish between healthy and unhealthy fungus gardens (Fig. 5), although the worker in a colony can take the unhealthy fungus to the waste chamber (Sousa et al. 2017). This observation was corroborated when the healthy and unhealthy fungal extracts were tested, suggesting that workers cannot distinguish the health status of the fungus (Fig. 6). It is likely that behavioral results reinforce the lack of a volatile semiochemical emitted by the fungus garden, as suggested by North et al. (1999), because workers must avoid breakdown products from unhealthy or dead fungi. Besides, the choice of these ants could have been affected by many other factors, such as the type of workers, the way the extracts were produced, the way the healthy and unhealthy parts were selected, etc.

Fig. 5.

Percent and confidence interval of ant choice in the Y-shaped choice system. A: Chamber with healthy fungus and chamber without fungus; B: chamber with healthy fungus and chamber without healthy fungus; C: chamber containing fungus with cycloheximide for 7 d and chamber without fungus; D: chamber with healthy fungus and chamber containing fungus with cycloheximide for 7 d; E: chamber containing fungus with cycloheximide for 14 d and chamber without fungus; F: chamber with healthy fungus and chamber containing fungus with cycloheximide for 14 d.

Fig. 6.

Means and confidence interval of disks transported containing healthy and unhealthy fungus extract into the colony.

Performing behavioral experiments, the study of North et al. (1999) concluded that highly volatile compounds emitted by the fungus garden directly affect the foraging behavior of leaf-cutting ant workers. The major flaw in their study is that they did not identify any volatile compounds from unhealthy or dead fungi because of the effect of cycloheximide. At high concentrations, as used by Ridley et al. (1996), cycloheximide causes the death of workers and the fungus garden (Sousa et al. 2017, 2018). Foraging behavior is probably affected by the toxicity of cycloheximide to workers and the fungus garden, and is not a behavioral response to the volatile semiochemical.

In conclusion, we observed that healthy and unhealthy fungus gardens emitted 4 volatile substances at the same proportion, and that workers did not discriminate between them. Thus, we hypothesize that non-volatile semiochemicals may be involved.