Study site and study species
The Gamboa field station (9° 07′N, 79° 42′W) of the Smithsonian Tropical Research Institute (STRI) in the Republic of Panama borders lowland, wet tropical forest and hosts a high abundance and diversity of plants, endophytic fungi and leaf-cutting ants. The most common leaf-cutting ant species in the forest margins and open areas of Gamboa is Atta colombica. We grew cucumber plants (Cucumis sativus) from seeds and cassava plants (Manihot esculenta) from cuttings as both plants grew rapidly in greenhouse conditions and provided a constant source of forage material for laboratory colonies.
Endophyte treatments
To provide Elow and Ehigh forage material to the feeding groups, we manipulated endophyte density and diversity of leaf material using two different techniques. Initially, plants were maintained free of endophytes by planting seeds or cuttings in sterile growth chambers (2 weeks). Then all C. sativus plants were moved to plastic enclosures inside of greenhouses (2–4 weeks), while M. esculenta plants remained in the growth chambers. We then introduced fungi in two ways. For M. esculenta, laboratory inoculations involved growing pure cultures of Colletotrichum tropicale ([19], strain Q633) conidia in broth, concentrating conidia in sterile water, and applying them as a spray onto cassava leaves. Control plants had sterile water sprayed on their leaves (as in[14]). For C. sativus, forest inoculations involved moving a subset of potted cucumber plants from the greenhouse to the forest during the night time only. The plants obtained the natural complement of endophyte spore fall by night, but the daytime conditions in the growth chamber remained the same for inoculated and control plants (as in[17]) (for details on plant inoculations see Additional file1).
Throughout our study, we isolated endophytes from 124 experimental leaves of C. sativus and 12 experimental leaves of M. esculenta to assess the abundance and diversity of endophytes and thereby check the validity of our treatments. Using standard endophyte isolation methods[20], we cut 2 mm2 leaf segments, surface sterilized them, and plated them on to 2% malt extract agar (MEA), a standard mycological medium. The plates were sealed and incubated at room temperature until we observed mycelia extending into the medium. From the C. sativus leaf pieces, a subset of the fungi (207 individual strains) were isolated into pure culture. After two weeks, plates were sorted by morphospecies and one or two individuals of the most common 5 morphotypes were identified using molecular genetic methods (n = 8 fungal strains, detailed below).
We harvested some of our treated cucumber leaves throughout the experiment for measurements of leaf mass per unit leaf area (LMA) and of nutritional content. These leaves were inoculated using forest inoculations, and were from the same plants that we used to provide forage material to the ants. For LMA leaves, we scanned the leaves to measure leaf area and dried them in a drying oven at 70°C for at least 5 days. Then we measured the mass of each dried leaf to estimate the LMA on a per leaf basis (n = 60 leaves). An additional subset of cucumber forage leaves were dried and sent for analysis of nutritional content (P, K, Ca, Mg, Zn, Mn, Cu, Fe) at the University of Florida Analytical Research Laboratory (Gainesville, FL).
Feeding experiment
We collected ~ 450 incipient A. colombica queens during their nuptial flight in May 2010. A. colombica queens carry a small piece of their natal fungal garden for establishment after mating. We kept the queens in small plastic containers with a small amount of soil collected from areas where they were found digging. Queens naturally established their fungal gardens, produced eggs and survived this period without external nutrition provided by our experiment. After five weeks, approximately 40% of the queens survived and we photographed their growing fungal garden to assess fungal area. To minimize disturbance, we did not remove queens during these photographs. When the first workers emerged (~6 weeks after the queens’ flight) we selected 120 colonies to include in our fungal garden development experiment. Colonies were haphazardly assigned to a low endophyte diet (hereafter, Elow, n=60) or a high endophyte diet (Ehigh, n=60) group. These colonies were placed in a 15 × 30 cm open plastic container with fluon (Teflon® PTFE 30) surrounding the edges to prevent the foraging ants’ escape. This plastic container also served as the foraging arena.
Over the course of the feeding experiment, colonies were provided with (1) cucumber leaves with high or low densities of endophytes, using leaf material from forest inoculations (1 seedling /day, 5 days/week); (2) cassava leaves with high or low density endophytes, using laboratory inoculations (~1-2 leaves/month); and (3) an open petri plate with 1 g of mixed corn meal and oats (for supplementary feeding over the weekend, 2 days/week). We did not remove uncut plants or leaves from prior feedings until they were replaced with fresh plants in subsequent feedings. Our intent was to constantly provide forage material (with some diversity) to the colonies, although the natural diversity of Atta colombica forage material is much higher[9].
We haphazardly chose colonies in each treatment group to destructively sample after 2, 4, 10, 16 and 20 weeks of foraging. The sample sizes for E- and E+ colonies from each period were; 15 and 15 colonies in week 2, 15 and 13 colonies in week 4, 9 and 8 colonies in week 10, 9 and 9 colonies in week 16, and 7 and 9 colonies in week 20. The effort needed to produce the forage material necessitated having fewer fungal gardens as the colonies grew. At each sampling time, we placed the colonies in the freezer. We then measured the fresh weight of the entire colony (including brood, garden and all ants except for the queen), the queens’ fresh mass, and we photographed each colony. We then separated the workers from the fungal garden and counted the pupae and the worker ants, grouping them into 2 size categories; small < 3 mm, and large > 3 mm. We lyophilized the fungal garden and measured the dry mass of the fungal garden only.
Laboratory vs. field comparisons
We wanted to assess the impact of our laboratory conditions and feeding regime on general microbial communities within the ants’ fungal garden. We sampled from 18 laboratory fungal gardens for culturable bacteria and fungi during week 16 of the feeding experiment. During the three previous weeks, we sampled from 19 A. colombica fungal gardens in the field around Gamboa in order to compare the microbial communities in field and laboratory conditions. Each sampling consisted of plating 18 1.5 mm2 pieces of the fungal garden per colony on 3 MEA plates (6 per plate). We scored the plates after 7 days to assess whether the garden isolate grew the ants’ symbiont fungi only, non-symbiont fungi, or bacteria. In some cases we saw that the fungal garden isolate produced both symbiont and non-symbiont fungi, in which case we scored them for both. We calculated percentages of fungal garden isolates with each type of growth, with overall percentages reaching over 100% in some cases where several types of fungi or bacteria grew out of the fungal garden section. Within the non-symbiont fungi we did a visual estimate of how many morphospecies per isolate were present on the plate, based on colony characteristics. Some garden pieces grew more than one non-symbiont or bacteria morphotype, in which case we counted each as an individual.
We also documented incipient ant queen and colony mortality in field versus laboratory conditions. In 2011, we measured the survival of A. colombica colonies in the field during 3 months after queens flew. This was measured in 4 plots that were ~ 25 m2 each, at least 0.5 km apart, and were all in open areas where queens dug holes in high abundance. Our first measurement was taken on the morning after the mating flight, when the queens’ entrance holes were clearly visible. The second measurement was 8 weeks later, when the young incipient colonies had just emerged. The third measurement was 4 weeks later, to emulate the 4-week sample in our laboratory feeding experiment. Across the plots we measured the fate of 375 queens.
Molecular analyses
We selected 8 representative strains from our 5 most common morphospecies isolated from cucumber leaves as endophytes, and 13 strains isolated from the fungal gardens as non-symbiont fungi, to identify the most common morphotypes present in the leaves and gardens. To determine the taxonomic affinities of these strains, primers ITS5 and ITS4[21] were used to amplify the approximately 540bp ITS, followed by sequencing with the same primers.
Statistical analyses
We used general linear models to estimate the effects of endophyte treatment (nested within sampling time) on fresh colony mass, number of workers, queen mass, dry garden mass, fresh colony mass per worker (× 1000), and dry garden mass per worker (× 1000). We followed this with Mann–Whitney U tests for the individual sample times. For various reasons, such as colony mortality, queen mortality, and problems with fungal garden masses, our final sample size for most general linear models was approximately 105–108 colonies. We used standard t tests to compare LMA, nutrient content, endophyte abundance, and garden isolates between the treatments or sampling areas. These tests were all performed using Systat[22]. We further used the species diversity estimator Chao 2[23] to compare the endophyte morphospecies diversity in our forage material.