Foliar endophytic fungi (hereafter “endophytes”) live within leaves and other above-ground plant tissues without causing any apparent signs of disease[1]. Previous work in temperate areas has demonstrated that some endophytes defend their host plants by making their leaves less palatable to insect herbivores[2–5]. Leaf-cutting ants (genera Atta and Acromyrmex, Myrmicinae) maintain an obligate symbiosis with their fungal cultivar (Basidiomycota, Leucocoprinus gongylophorus)[6, 7]. The ants defoliate a wide diversity of plants and often have an enormous effect on local flora and distribution of nutrients[8–10]. The worker ants cut leaves, carry them to the nest, clean them, and use them as compost to cultivate their symbiont fungus. The ants’ fungal symbiont then partially degrades the leaf material, converting leaf biomass to fungal food for the ants.

While the ant-cultivar symbiosis is relatively well studied, the potential interactions among the ants’ cultivar and fungal endophytes are only recently receiving attention. In the temperate zone, one previous experiment with grasses found evidence of endophyte toxicity toward leaf-cutting ant queens for some grass-endophyte combinations, suggesting a defensive mutualism between the grass and endophyte[11]. A few descriptive studies have suggested that some fungal endophytes can enter and persist in the ant’s gardens. Endophytes were isolated from the gardens of naturally occurring Acromyrmex sp. nests[12], and from laboratory colonies of Atta cephalotes; in the latter endophyte composition changed when ants were offered a new food source[13]. Despite these observations, little is known about how ants and their fungal cultivar interact with endophytic fungi in their forage material[14–17].

Previous research with laboratory colonies of Atta colombica (Guérin-Méneville) (Myrmicinae; Attini) in Panama has shown that (1) ants spend 45% more time cutting and removing leaf pieces with high versus low endophyte loads, (2) ants reduce the amount of endophytic fungi in leaves before planting them in their gardens and (3) the ants’ fungal cultivar inhibits the growth of most endophytes tested using in vitro bioassays[14]. Moreover, when given a choice, leaf-cutter ants cut nearly 50% more leaf area from seedlings with low endophyte relative to high endophyte loads[17]. These short-term experiments, however, did not test whether the extra time spent cutting and preparing leaves translated into a cost for the colony, such as reduced fungal growth and overall colony development. The overall growth rate of the colony is likely to depend on many factors, including the maturity of the colony, environmental conditions, and the amount of time necessary for worker ants to cut, process and plant material in the fungal gardens. Moreover, the ants’ fungal cultivar is likely to interact with plant material from different host plants in idiosyncratic ways, with some host plants providing a superior substrate for cultivar growth.

Here we report a 20-week long experiment that monitored the development of A. colombica colonies foraging on leaves with high or low endophyte loads. The experiment was conducted to provide evidence for or against the hypothesis that foliar endophytes reduce defoliation to plants that host them by negatively affecting leaf-cutting ant colonies and their fungal symbiont. Previous observations with A. colombica suggested that endophytes were not directly toxic to the ants or their fungi (Van Bael unpublished). We therefore predicted that endophytes would exact a cost on ant colonies by slowing their overall development rate. This prediction assumes that early ant colony development is limited by how much leaf material the workers can cut.

An alternative hypothesis is that leaf-cutting ants and their cultivar benefit from the activity of certain endophytic fungi in their fungal gardens. In this case, leaf-cutting ants could choose to remove some strains of endophytes, but actively plant and encourage the growth of others. There may be an unrecognized benefit to maintaining diverse endophytic fungi in leaf-cutter ant gardens, as suggested for microbiome diversity in general (e.g.,[18]). If such benefits outweigh the cost of extra time needed to manipulate endophytes, we would expect to see greater survivorship and growth rates in colonies that foraged on leaves with high endophyte loads.

We observed a trend toward a greater fungal garden mass per unit worker when the ants were fed a low endophyte diet, but the difference was only marginally significant over the whole experiment. The trend was driven by the colony measurements from the first sampling, two weeks after worker ants emerged and began cutting leaves. A foundress queen rears her first brood of young using trophic eggs, and nourishes the fungal cultivar with secretions[6, 7], so we did not provide queens with leaves for cutting. Thus, this result of smaller colonies and fewer workers is solely due to effects of endophytes on workers, not the queen. This suggests that endophyte loads may limit colony productivity for very young colonies with few, naïve workers, but that within a month of cutting the colony productivity (as measured by fungal mass per unit worker) was similar between low and high endophyte treatments.

H. virens, a fungal parasite and the teleomorph of Trichoderma virens[24], was sequenced from several of our laboratory colony isolates (Table3). It may have been present in the laboratory colonies at higher levels than in the field colonies. Further, our sampling of fungal gardens after 16 weeks of leaf-cutter ant foraging showed there were greater levels of non-symbiont fungal infection in E_low relative to E_high laboratory colonies. One possible interpretation is that the presence of high endophyte loads in leaf material brought by the ants may actually help the leaf-cutter ants’ symbiont defend against fungal parasites in the garden. This echoes the hypotheses in human agriculture that polycultural practices help defend crops against herbivores and diseases[25], or that a diverse microbiome should be less susceptible to successful invasion by a pathogen[18]. Interestingly we also observed that fungal gardens freshly harvested from the field had higher bacteria and lower non-symbiont fungal loads than our laboratory colonies, suggesting that the microbial community in laboratory gardens was disturbed and not representative of natural communities.

Our experiment was limited by several factors. First, the laboratory setting protected all colonies from soil pathogens, predators, fungal parasites, and potential fungal mutualists that the ant colonies would have experienced in nature. Second, due to the practical difficulties of producing large quantities of forage material for our experimental colonies, we provided the ants with an artificially narrow diet relative to what they would forage on in nature[9], though colonies near monoculture (human) farms would naturally have lower diversity of forage material. We chose the plants C. sativus and M. esculenta because they were experimentally tractable, relative to rainforest plant species that would not produce enough leaves rapidly enough to support our colonies. Finally, the levels of endophytes in our E_low treatment were higher than we expected with respect to previous experiments where “forest inoculations” were used. For example, in Bittleston et al.[17], Cordia alliodora leaves with low endophyte treatments resulted in 13% of the leaf area containing endophytes (vs. 39% in this experiment). Thus, the difference between treatments was less than ideal to see whether endophytes play an important role in fungal garden development.

Considering that endophytes are symbionts that obtain resources from their hosts and grow within their hosts, it is plausible that endophytes of woody plants have evolved ways to defend their hosts, and thus themselves, from being eaten or carried away and decomposed. Plants generally defend themselves from herbivore attack using: (i) inducible defense, whereby herbivore attack elicits changes in leaf chemistry that reduce palatability, (ii) constitutive or “quantitative” defense, in which leaf toughness, fiber, leaf-chemistry, and/or low nutrient availability collectively work to increase an herbivore’s development time on a particular food plant (reviewed in[26, 27]), (iii) or a combination of inducible and constitutive defense. Constitutive defenses that slow herbivore growth lead to the prediction that longer development time of herbivores will increase their exposure to predators and parasites[26]. We suggest that endophytes are functionally analogous to constitutive defenses toward ants and their cultivar, in that increased development time (slower growth rates) of ant colonies leads to greater rates of mortality among incipient colonies.

Detailed longitudinal studies in natural populations are lacking for most leaf-cutter species, as is true for ants in general[28]. For Atta capiguara, fewer than 2% of incipient nests survive through the claustral stage (i.e., when the first workers emerge)[29], but successful nest establishment also depends on the local density of mature colonies. A review of the limited data for Atta gives comparable figures for other species[28], although factors other than disease may also be important. In A. sexdens, for example, probability of nest survival is also contingent on duration of digging activity[30]. In a savannah species (A. laevigata), about 3% of inseminated queens successful established a nest ([31], cited in[32]). After the first workers emerged, nest mortality was approximately 45% per month for the next 25 months, then monthly mortality rates dropped to single digits[32]. At our study site, where Atta spp. are abundant, ~6% of the incipient A. colombica queens resulted in productive colonies. Given these mortality schedules, our data raise the possibility that there is a differential cost of endophytes to the colony during the early weeks of establishment—an important time interval for determining colony survival[28]—relative to already-established colonies. Since our laboratory colonies tended to survive this mortality window despite endophyte loads, however, this hypothesis assumes that the cost of endophytes in ant forage material interacts with other environmental or biotic factors to cause less productivity or greater mortality. For example, workers of incipient colonies are monomorphic, while those of mature colonies are highly polymorphic[6, 7]. Tiny minor workers, which have relatively larger antibiotic- producing glands, are efficient at removing pathogenic spores from interstices deep within in a garden, and are recruited to sites of infection by larger major workers[33, 34], but these size-based efficiencies cannot be realized by a small, monomorphic, work force. Similarly, the production of costly antibiotic compounds and the extra time needed for cleaning leaf material with endophytes, which can increase by 40% on average when leaf material is high in endophyte abundance[16], could decrease the developing colony’s ability to properly patrol and clear the garden from microbial invaders[35, 36]. Such an effect would be strongest when there are few workers available to perform colony defensive duties. Accordingly, previous studies have suggested that the number of workers present will influence the colony’s ability to defend itself against pathogens and parasites[37–39].

In summary, our experiment provided evidence for a cost of endophytes on fungal garden productivity, but only during the first weeks of leaf-cutter ant foraging. Combined with high mortality rates in the field for incipient colonies, this provides partial support for our hypothesis that foliar endophytes defend their host plants from leaf-cutter ant attack. Our alternative hypothesis, that endophyte diversity could actually help colony productivity, was not supported by the measurements of ant colony productivity in this experiment. The microbial communities in fungal gardens, however, did change as a result of the high and low endophyte diets. Whether these changes in the microbial community affect the ant colonies’ defense against pathogens and parasites deserves further attention.