OUP user menu

Ecology of microfungal communities in gardens of fungus-growing ants (Hymenoptera: Formicidae): a year-long survey of three species of attine ants in Central Texas

Andre Rodrigues, Ulrich G. Mueller, Heather D. Ishak, Maurício Bacci Jr, Fernando C. Pagnocca
DOI: http://dx.doi.org/10.1111/j.1574-6941.2011.01152.x 244-255 First published online: 1 November 2011


We profiled the microfungal communities in gardens of fungus-growing ants to evaluate possible species-specific ant–microfungal associations and to assess the potential dependencies of microfungal diversity on ant foraging behavior. In a 1-year survey, we isolated microfungi from nests ofCyphomyrmex wheeleri, Trachymyrmex septentrionalis andAtta texana in Central Texas. Microfungal prevalence was higher in gardens ofC. wheeleri (57%) than in the gardens ofT. septentrionalis (46%) andA. texana (35%). Culture-dependent methods coupled with a polyphasic approach of species identification revealed diverse and changing microfungal communities in all the sampling periods. Diversity analyses showed no obvious correlations between the number of observed microfungal species, ant species, or the ants' changing foraging behavior across the seasons. However, both correspondence analysis and 5.8S-rRNA geneunifrac analyses suggested structuring of microfungal communities by ant host. These host-specific differences may reflect in part the three different environments where ants were collected. Most interestingly, the specialized fungal parasiteEscovopsis was not isolated from any attine garden in this study near the northernmost limit of the range of attine ants, contrasting with previous studies that indicated a significant incidence of this parasite in ant gardens from Central and South America. The observed differences of microfungal communities in attine gardens suggest that the ants are continuously in contact with a diverse microfungal species assemblage.

  • attini
  • diversity
  • fungi
  • symbiosis
  • Escovopsis


Mutualisms are generally viewed as interspecies interactions with reciprocal fitness benefits for both partners. The understanding of mutualisms as dual associations was transformed in the past decade by the realization that the nature of dual interactions can be changed categorically in the presence of additional interactors (Stanton, 2003; Little et al., 2008). Such mutualisms actually need to be understood as an intricate interaction network among diverse hosts and symbionts. For example, insect hosts progress through their life cycle within a changing network composed of a myriad of microorganisms inhabiting their nest environment and their bodies (Ganter, 2006). The task to elucidate the contributions of specific microorganisms to the host life cycle is challenging, and yet such studies of insect-associated microbiota are needed to go beyond the traditional studies of dual mutualistic interactions.

As an example of insect–fungus interactions that involves both specific microbial symbionts as well as non-specialized microorganisms is the fungus-growing ant–microbe mutualism. Ants in the tribe Attini (Hymenoptera: Formicidae) are social insects that have lived in an obligate symbiosis with fungi for the past 50 million years (Schultz & Brady, 2008). These insects cultivate basidiomycetous fungi as their main source of food and digestive enzymes (Mueller et al., 1998; Silva et al., 2003, 2006a; De Fine Licht et al., 2010; Schiøtt et al., 2010). In order to nourish their symbiotic cultivar, workers forage on different substrate types depending on the ant species. Primitive attine ant species (the ‘lower attines’) forage for seeds, insect frass, and dry plant debris (e.g. withered flower parts). In contrast, the derived attines (the ‘higher attines’) use mostly dry plant substrate in addition to bits of fresh plant material (De Fine Licht & Boomsma, 2010). A subset of species within this group, the well-known leafcutter ants, uses fresh leaves and flower parts to manure their fungal gardens (Weber, 1972). Workers of all attine genera process substrate before they incorporate it into the fungus garden, which the ants build as a three-dimensional structure composed of fungal mycelium and the processed substrate. Substrate processing before incorporation into the garden is thought to help remove unwanted alien microorganisms that could compete with or harm the symbiotic cultivar fungus (Fernández-Marín et al., 2006, 2009).

The understanding of the attine symbiosis has changed dramatically in the past decade (Bacci et al., 1995a, b, 2009; Mueller (2002); Mueller et al., 1998, 2005; Schultz & Brady, 2008; Hölldobler & Wilson, 2010; Mikheyev et al., 2010). Leaf-cutting ants (generaAtta andAcromyrmex) across the Neotropics cultivate a single, phylogenetically narrow fungal lineage (Silva-Pinhati et al., 2004; Mikheyev et al., 2006, 2010), and all fungus gardens propagated by a single colony of ants appear to be a monoculture of a single cultivar clone (Poulsen & Boomsma, 2005; Mueller et al., 2010). Organisms with reduced genetic diversity can be threatened by fast-evolving pathogens and parasites that readily overcome the defenses of a genetically depauperate host population (Hamilton et al., 1990). This seems to be the case of fungus gardens of attine ants, which are subject to significant parasite pressures (Currie et al., 1999a; Currie, 2001; Currie et al., 2003; Rodrigues et al., 2005, 2008).

Fungal parasites in the genusEscovopsis (Ascomycota, Hypocreales) were previously found in association with attine nests (Currie et al., 1999a; Currie et al., 2003). So far, this parasite has only been found in attine gardens and garden dumps, and it is thought to derive nutrients directly from the fungal cultivar (Reynolds & Currie, 2004), thus negatively impacting garden and colony growth (Currie, 2001). In order to protect gardens from the attack of such parasites, ants try to physically weed outEscovopsis-infected garden parts (Currie & Stuart, 2001). Some attine ants also appear to use antibiotic-secreting bacteria (in the genusPseudonocardia orStreptomyces) found on their bodies to help suppressEscovopsis sp. (Currie et al., 1999b; Cafaro & Currie, 2005; Haeder et al., 2009; Barke et al., 2010; Poulsen & Currie, 2010), but application of actinomycete-derived antibiotics to gardens may also inhibit cultivar growthin vitro (Sen et al., 2009; Barke et al., 2010). The cultivars, in turn, can inhibitPseudonocardia growth (Poulsen & Currie, 2010). Despite a decade of research (Currie et al., 1999b), the hypothesized primary importance ofPseudonocardia symbionts inEscovopsis suppression remains puzzling (Mueller et al., 2008; Boomsma & Aanen, 2009; Poulsen et al., 2009; Sen et al., 2009; Barke et al., 2010; Poulsen & Currie, 2010). It is possible that both coevolved and environmentally acquired actinomycetes help ants to defend their gardens against the parasiteEscovopsis sp. (Barke et al., 2010 and references within).

Attine fungus gardens can be viewed as an interacting community of microorganisms (Bacci et al., 1995b; Santos et al., 2004; Mueller et al., 2005). The fungal cultivar is thought to contribute the main biomass to this community, but several other microorganisms such as bacteria (Bacci et al., 1995b), filamentous fungi (Rodrigues et al., 2005), and yeasts (Craven et al., 1970; Carreiro et al., 1997) are also found in this microbiome, including yeast species only associated with attine gardens, as is the case ofCryptococcus haglerorum (Middelhoven et al., 2003) andTrichosporon chiarellii (Pagnocca et al., 2010). Most noncultivar microorganisms probably enter the garden communities with the plant substrate or when shed from the ant cuticle, but some noncultivar microorganisms are already present in the inoculum carried by queens to start new gardens (Mueller et al., 2008; Pagnocca et al., 2008). The roles of such secondary microorganisms in the attine ant symbiosis are largely unknown.Santos (2004) andRodrigues (2009) demonstrated that bacteria and yeasts, respectively, isolated from leafcutter gardens inhibitEscovopsis sp., suggesting that diverse microorganisms in the garden matrix may contribute to the protection of the ant garden against diseases.

Although some microorganisms in attine nests may confer benefits to the ants, others may represent antagonists (pathogens) or competitors for nutrients of the cultivated fungus (Bacci et al., 1995a; Silva et al., 2006b). For example, soil microfungi in the nest walls and endophytic fungi in the gardening substrate may thrive in garden chambers that are optimized by the ants for fungal growth (Fisher et al., 1996; Rodrigues et al., 2008; van Bael et al., 2009). Despite the physical and chemical barriers that ants impose to suppress such unwanted microbial invaders, microfungi persist in surprisingly high abundance in gardens of different attine species (Currie et al., 1999a; Rodrigues et al., 2008). For example, the microfungal generaCunninghamella, Fusarium, andTrichoderma are frequently found in field gardens ofAcromyrmex andAtta leafcutter ants (Rodrigues et al., 2005, 2008). However, little is known regarding how these fungi interact with ants and whether they are in fact parasites or transient commensals. Unlike the specializedEscovopsis sp. (Poulsen & Currie, 2006; Rodrigues et al., 2008), frequent invaders such asCunninghamella sp.,Fusarium sp., andTrichoderma sp. appear to be nonspecialized and generalized antagonists (‘garden weeds’) of the fungal cultivar (Rodrigues et al., 2005, 2008).

To further characterize the associations between attine ants and garden microfungi, and to assess any species-specific associations between microfungi and ants, we conducted a 1-year study of the phenological changes in microfungal diversity in gardens of three fungus-growing ants in Central Texas. In addition, we investigated whether the diversity of microfungal communities correlates in part with the ant foraging behavior (Weber, 1956, 1972; Mintzer, 1979; Waller, 1986). Our results indicate that the attine species studied are confronted with a diverse assemblage of microfungi that appears to be transient and nonspecific associates in attine gardens.

Materials and methods

Ant species and sampling design

We surveyed the microfungal community of three attine ant species,Cyphomyrmex wheeleri (a lower-attine ant),Trachymyrmex septentrionalis (a higher-attine ant), andAtta texana (a leafcutter ant). These ants represent three of the five main fungicultural systems in the tribe Attini (Schultz & Brady, 2008). Fungus gardens were collected from these species at four sampling periods: (1) winter (December 2005 and January 2006); (2) spring (March–May 2006); (3) summer (July 2006); and (4) fall (September and October 2006).

At each sampling period, four colonies ofC. wheeleri were excavated at Bull Creek Park (Austin, Travis County, TX – GPS: 30°22′07″N, 97°47′04″W) and four colonies ofT. septentrionalis at Stengl ‘Lost Pines’ Biology Station (Smithville, Bastrop County, TX – GPS: 30°05′04″N, 97°10′23″W), totaling 16 nests for each of these two ant species.Cyphomyrmex wheeleri nested in the open areas of a cedar-brake forest with dry clay-sand alluvial soils and sparse ground vegetation (e.g. grasses). In such an environment,C. wheeleri workers forage for grass seeds, catkins, diverse fibrous plant material, and arthropod frass (U.G. Mueller, pers. commun.). Nests ofT. septentrionalis nested in the sandy soil of a pine forest, and workers foraged for dry leaves, seeds, and arthropod frass (Seal & Tschinkel, 2008; A. Mikheyev, pers. commun.; J. Seal, pers. commun.). Both ant species build nests with generally two to four garden chambers. In our study, samples were generally collected from the top-most garden accessed through careful excavation.

Three nests ofA. texana were sampled at Hornsby Bend Environmental Research Center (Austin, TX – GPS: 30°13′06″N, 97°38′21″W) and a fourth nest was sampled along Park Road 1C connecting Bastrop State Park and Buescher State Park (Bastrop County, TX – GPS: 30°05′49″N, 97°13′29″W). Both sites have sandy soils, but the forest at Hornsby Bend is mostly riverine pecan and oak trees, whereas the forest at Park Road 1C is dominated by pines (comparable to Stengl Biological Station, where we studiedT. septentrionalis). Because mature nests ofA. texana have hundreds of garden chambers, we could repeatedly collect from the same nests at 3-month intervals. Most of the fungus garden sampled fromA. texana had a white-brown color, which indicates that ants did not recently add much plant substrate. However, gardens for nests UGM060121-01 and UGM060121-02 collected in spring 2006 had a dark brown-green appearance, indicating that workers recently added fresh leaf substrate. In fall 2006, the latter type of garden was found in nest UGM060121-02, but not in the other nests sampled at that time.

All leafcutter nests were excavated according toRodrigues (2009) in the center of the mound, where most fresh excavate had accumulated. A deep hole was dug next to the targeted fungus gardens, and then the soil was carefully shaved off at the side of the hole until a garden chamber was exposed. Gardens were extracted intact, and fragments from the center of the fungus garden were collected with some attending ants into sterile containers without disturbing the natural architecture of the garden, then maintained at environmental and room temperatures until processing in the laboratory (i.e. usually within 4–8 h of collection). Fungus gardens were not collected if any soil accidentally came into contact with a garden during excavation. In such cases, we continued our excavation until we could access a garden without soil contamination. The excavated holes were filled after each collection to prevent colony movements that could be triggered by disturbances. In fall 2006, fungus gardens were found in only three of the fourA. texana nests; in nest AR060123-01, no fungus garden could be found in 4 days of excavation, but many empty chambers and several chambers filled with decayed garden substrate similar in consistency to attine dumps.

Microfungi isolation

We used culture-dependent methods to study the microfungal diversity in attine gardens. Before microfungal isolation, workers, pupae, and larvae were removed from fungus garden samples using a sterile forceps. Fungus garden fragments of about 3 mm3 were inoculated in three culture media supplemented with 150 μg mL−1 of chloramphenicol (Sigma): (1) 2% malt agar (MA2% in g L−1: 20 malt extract and 15 agar; Difco), (2) potato dextrose agar (EMBL – Difco), and (3) synthetic nutrient agar (in g L−1: 1 KH2PO4; 1 KNO3; 0.5 MgSO4·7H2O; 0.5 KCl; 0.2 glucose and 0.2 sucrose). For each sampling period, 20 fungus garden fragments ofC. wheeleri andT. septentrionalis, and 25 garden fragments ofA. texana were inoculated per medium. BecauseT. septentrionalis gardens shrink during the winter (Weber, 1956), the little garden that we were able to collect in winter could be separated into only 10 fragments for microfungal isolations.

All plates were incubated at 25 °C for 10 days with a 12 h/12 h light–dark regime. Isolation plates were observed for 4 weeks for any noncultivar fungi sprouting out from the garden fragments. Once a growing fungus was observed; it was subcultured to a new MA2% plate to obtain a pure culture. Isolates were stored live in 10% glycerol at −80 °C for fungal identification.

Microfungi identification

We followed a polyphasic approach for fungal identification, using both morphological and molecular markers. First, all fungal isolates were grown in oatmeal agar and MA2% for initial morphotyping, and then grouped to the lowest taxonomic level using classical taxonomical keys (Ellis, 1971, 1976; Subramanian, 1971; Domsch et al., 1980; Samson et al., 2000). Following this preliminary morphotyping, isolates were inoculated in taxon-specific media to further study the macroscopic and microscopic characteristics that would allow a more precise taxonomic identification. We used specific taxonomic keys for fungi in the following groups:Aspergillus sp. (Klich, 2002),Cunninghamella sp. (Liu et al., 2001),Fusarium sp. (Nelson et al., 1983),Penicillium sp. (Pitt, 1979),Phoma sp. (Boerema et al., 2004),Mucorales (Zycha, 1935), and insect pathogenic fungi (Samson et al., 1988). Fungal isolates that failed to sporulate in oatmeal agar and MA2% were inoculated in potato carrot agar (Hi-Media) and incubated at 25 °C for several months. These isolates were regularly inspected for any reproductive structures that may allow fungal identification.

Third, for molecular identification of the isolated fungi, the internal transcribed spacer region (ITS1-5.8S-ITS2) was sequenced for representative isolates (these fungi are marked with ‘S’ in Supporting Information, Tables S2–S4). DNA extraction followed the methods ofMikheyev (2006); a small quantity of fresh fungal mycelium was collected from culture plates (grown on MA2% for 7 days at 25 °C), heated to 60 °C for 1.5 h, and then boiled for 10 min in the presence of 100 μL of 5% Chelex resin solution (Sigma). PCR amplifications were carried out using a final volume of 10 μL containing: 10 × KCl buffer, 0.8 μM of each primer (ITS4 and ITS5), 1 mM of each dNTPs, 2.5 mM MgCl2, 1 U of Taq polymerase, and 1 μL of DNA extractions. The PCR profile included an initial denaturation at 94 °C for 3 min, followed by 35 cycles of 94 °C for 1 min, 52 °C for 1 min, and 72 °C for 2 min. Sequences were generated on an ABI 3100 Automated Sequencer (Applied Biosystems).

Sequence analysis and taxon definition

Forward and reverse sequences were assembled inbioedit v. (Hall, 1999) and used forblastn at the NCBI-GenBank database (http://www.ncbi.nlm.nih.gov). Sequences found in the database with the highest nucleotide similarities were considered as the closest relative of the query sequence (Tables S2–S4 for GenBank accession numbers of the closest relatives found).

Pairwise-distance comparisons of ITS sequences were computed inmega v.4 (Tamura et al., 2007) using default settings. Sequences with <3% sequence variability were treated as the same taxon (seeUnterseher & Schnittler, 2010). Tables S2–S4 summarize all taxa that were identified by the combination of morphological and molecular markers. Sequences of representative isolates were deposited at NCBI-GenBank under the accession numbers HQ607791HQ607830, HQ607832HQ607892, and HQ607894HQ608158.

Analysis of microfungal communities

To establish the prevalence of microfungi in the sampled fungus gardens, we calculated, separately for each isolation medium and each season, the mean proportion of garden fragments in which microfungi were present. Nonparametric Kruskal–Wallis and Dunn tests implemented inbioestat v.5.0 (Ayres et al., 2007) were used to test at the 0.05 level for statistically significant differences among treatments.

We assessed the actual microfungal species richness through individual-based rarefaction curves computed inecosim v.7.72 (Gotelli & Entsminger, 2011). Such analyses used the respective datasets presented in Tables S2–S4 to evaluate differences in the observed species richness among seasons. Moreover, individual-based rarefaction curves were computed for each ant species using a comprehensive dataset pooled across seasons for each ant species. Chao 1 species richness estimators and diversity indices (Fisher's alpha, Shannon, and Simpson) were calculated for each seasonal dataset and the pooled datasets usingestimates v.8.2 (Colwell, 2009).

To evaluate differences in microfungal community composition between attine gardens, we used the information summarized in Tables S2–S4 for two additional analyses. In the first analysis, we performed a correspondence analysis on the complete dataset in order to investigate the nonapparent relationships among microfungal communities and the three ant species. Such calculations were carried out injmp® v.4.0.3 (SAS Institute Inc.). In the second analysis, we used just the conserved 5.8S portion of the ITS region for a phylogenetic-based clustering analysis using the programunifrac (Hamady et al., 2010). A sequence alignment was generated usingclustal x (Larkin et al., 2007) and further aligned inmacclade 4.08 (Maddison & Maddison, 2005). An approximate maximum likelihood tree was generated using the GTR model infasttree (Price et al., 2009). We performed aunifrac principal coordinate analysis (PCoA) to evaluate the similarities between microfungal communities associated with the different ant species, between collection location, and between seasons.


Prevalence of microfungi in fungus gardens of attine ants

A total of 1435 microfungal isolates were recovered from fungus garden samples of all three ant species studied in our 1-year survey (Table S1). A higher percentage of garden fragments yielded non-cultivar microfungal isolates fromC. wheeleri (57%) than fromT. septentrionalis (46%) andA. texana (34%); however, these percentages were not significantly different (Kruskal–Wallis,H=0.00,P=1.00,Fig. 1). On the other hand, 43% of the garden fragments ofC. wheeleri, 54% ofT. septentrionalis, and 66% ofA. texana showed growth of the resident cultivar.


Proportion of garden fragments fromCyphomyrmex wheeleri, Trachymyrmex septentrionalis, andAtta texana that showed non-cultivar microfungal growth (bars indicate SEs). Sixteen gardens ofC. wheeleri andT. septentrionalis were sampled from individual nests in a year-long survey. ForA. texana, a total of 15 gardens were sampled from four nests over the same study period.

There were no significant differences in the proportion of garden fragments with microfungi across all the three culture media used when comparing within ant species:C. wheeleri (Kruskal–Wallis,H=0.22,P=0.89),T. septentrionalis (Kruskal–Wallis,H=0.65,P=0.72), andA. texana (Kruskal–WallisH=1.13,P=0.56). However, the proportion of fragments with microfungi significantly differed between the four seasons (Fig. S1) forC. wheeleri (Kruskal–Wallis,H=12.26,P=0.006),T. septentrionalis (Kruskal–Wallis,H=11.45,P=0.009), andA. texana (Kruskal–Wallis,H=7.91,P=0.04). InC. wheeleri, there were proportionally fewer fragments with microfungi in spring than in winter, summer, and fall (Dunn test,z=2.82,P<0.05). ForT. septentrionalis, proportionally fewer microfungi were recovered in winter than in summer fungus garden samples (Dunn test,z=3.04,P<0.05). We observed a tendency to recover less microfungal isolates in summer and fall onA. texana garden samples; however, the observed differences were not significant (Dunn test,z=3.04,P>0.05). In addition, there was no significant difference in the proportion of garden fragments with microfungi between freshly growing gardens and mature gardens fromA. texana (Mann–Whitney,U=149,P=0.35).

Microfungal diversity and composition in attine gardens

In the 1-year survey of garden-associated microfungi fromC. wheeleri, a total of 524 isolates were recovered (Table S1), belonging to 111 observed taxa. These taxa comprise 52 genera, 72 species, and 13 unidentified ascomycetes as determined by our polyphasic approach of fungal identification (Table S2). According to rarefaction analysis of the pooled dataset across seasons (Fig. 2a, Table S5), the Chao 1 richness estimator predicted a total of 292±74.7 taxa (mean±SD) in gardens ofC. wheeleri, which indicates that more sampling effort is necessary to cover the full diversity of microfungi in fungus gardens ofC. wheeleri.


Individual-based rarefaction curves (dark solid lines) and Chao 1 richness estimators (dashed lines) of microfungal communities in gardens of three attine ants from Central Texas: (a)Cyphomyrmex wheeleri, (b)Trachymyrmex septentrionalis, (c)Atta texana. Curves were generated by pooling samples across seasons for each ant species. Light solid lines represent Chao 1 95% confidence intervals.

Microfungi that were consistently present in all seasons inC. wheeleri gardens wereCladosporium cladosporioides, Cochliobolus sp. 1,Fusarium solani, andNigrospora sphaerica. However, we detected significant differences in the observed taxon richness across seasons forC. wheeleri (Fig. S2a). Microfungi richness was higher during spring and winter when compared with the richness found in summer and fall (Fig. S2a). The prevalent microfungal taxa in each season were (relative to the total number of isolates in each season)Alysidium sp. 1 (13%),Mucorales sp. 1 (8%), andN. sphaerica (8%) in winter;Verticillium sp. 1 (12%),C. cladosporioides (11%), andAspergillus oryzae (9%) in spring; an unidentified ascomycetes strain (30%),Cochliobolus sp. (14%), and an unidentifiedDothideales sp. 1 (11%) in summer; andN. sphaerica (23%), an unidentifiedDothideales sp. 1 (8%), andTrichoderma koningii (7%) in fall (Table S2).

Fungus garden fragments ofT. septentrionalis rendered a total of 386 isolates (Table S1) belonging to 117 taxa. The observed taxa richness comprises 47 genera, 66 species, 30 unidentified ascomycetes, and one unidentified basidiomycetes fungus (Table S3). The Chao 1 richness estimator predicted a total of 218.16±36.05 taxa (mean±SD) in gardens ofT. septentrionalis, which also indicates that more sampling effort is necessary to cover the full diversity of microfungal taxa in the fungus gardens of such ant species (Fig. 2b, Table S5). InT. septentrionalis gardens, microfungi occurring in all seasons wereMortierella sp. 1,Penicillium verruculosum, Penicillium sp. 3, andTrichoderma harzianum. When analyzed by season, the observed species richness was not significantly different among the spring, summer, and fall seasons (Fig. S2b). In the winter season, fewer taxa were recovered from gardens ofT. septentrionalis, but this was not significantly different from the observed richness of the spring season. The prevalent microfungal taxa in each season werePenicillium restrictum (22%) andCeratocystis fimbriata (8%) in winter;Trichoderma hamatum (11%) andPenicillium sp. 1 (9%) in spring;Gloeotinia sp. anamorph (19%),P. verruculosum (12.5%), andPenicillium sp. 3 (11%) in summer; andVerticillium sp. complex (8%) andPenicillium decumbens (7%) in fall (Table S3).

Fungus garden fragments ofA. texana yielded a total of 525 isolates (Table S1) with an observed richness of 153 taxa. A total of 53 genera, 106 species, and 23 unidentified ascomycetes fungi were observed onA. texana samples (Tables S4). The Chao 1 richness estimator predicted a total of 308±44.53 taxa (mean±SD) in the gardens ofA. texana, which again indicated that more sampling effort is necessary to cover the full microfungal diversity in the gardens ofA. texana (Fig. S2c, Table S5). Microfungi occurring in all seasons wereAlternaria alternata, Aspergillus ustus, Beauveria sp. 1, and one unidentified ascomycetes. Comparisons across seasons revealed that the observed species richness was significantly higher in spring and summer than for the other seasons (Fig. S2c). The most prevalent microfungal species in such gardens were:C. cladosporioides (16%),A. ustus (11%), andAspergillus ochraceus (10.8%) in winter;C. cladosporioides (8%) andA. alternata (7.4%) in spring; an unidentifiedChaetothyriales sp. 1 (7.6%) andPenicillium sp. (7.6%) in summer; andCladosporium sp. (14%),Cochliobolus australiensis (10%), andCochliobolus lunatus (9%) in fall (Table S4).

Overall, Fisher's alpha, Shannon, and Simpson diversity indices suggest thatA. texana andT. septentrionalis have similar species richness when compared with fungus gardens ofC. wheeleri (Table S5). Fungus gardens ofC. wheeleri harbored a more diverse community in winter and spring than summer and fall (Table S5). In the fungus gardens ofT. septentrionalis, the microfungal community appeared to be more diverse in spring and fall. In addition, the diversity indices indicate that microfungal communities inA. texana gardens were diverse in spring and summer. Interestingly, the specialized parasiteEscovopsis sp. was not isolated from any of the nests sampled in this study over the 1-year long period.

Garden microfungal community structure

Cluster analyses suggest structuring of microfungal communities in fungus gardens of the three studied ants. Correspondence analysis using taxa identified by a combination of morphological and molecular markers (Tables S2–S4) revealed that microfungal communities clustered by ant genera (Fig. 3). The two main components in the correspondence analysis explain 76.5% of the total variation observed. The analysis also suggests that grouping by ant genera is further resolved by collecting sites as microfungal communities ofA. texana were sampled in two distinct sites: Bastrop and Hornsby Bend (Fig. 3).


Correspondence analysis of microfungal community diversity from the three attine ant speciesCyphomyrmex wheeleri, Trachymyrmex septentrionalis, andAtta texana. Black squares represent microfungal observations from any of the three ant host (recorded as presence–absence data in a particular ant host). Microfungal communities separate by ant species. The biplot suggests further separation by collection site, as theA. texana samples collected at different sites separate in two groups: Hornsby Bend Environmental Research Center (H) and Buescher State Park (B).

Such findings were also corroborated when the microfungal community was analyzed with just the sequenced fungal isolates (Tables S2-S4). Using a phylogenetic approach, theunifrac PCoA indicates that fungal isolates of each community clustered by ant genera (Fig. 4, the two principal coordinate axes explained 34.5% of the total variation). On the other hand, no clear clustering by season was observed (data not shown). Similar to the correspondence analysis,unifrac grouping revealed that clustering by ant genera is further explained by collection sites becauseA. texana samples collected in Bastrop and Hornsby Bend formed two distinct communities.


PCoA of the unweightedunifrac distance matrix based on the 5.8S rRNA gene-region of microfungi associated with fungus gardens of attine ants. Microfungal communities cluster by genus of ant host. TheAtta texana samples separate out into two distinct clusters, each collected at a different location: Hornsby Bend Environmental Research Center and Buescher State Park.


Fungus gardens of attine ants are a miniature ecosystem of diverse microbial consortia (Bacci et al., 1995b; Mueller et al., 2005; Gerardo et al., 2006; Youngsteadt, 2008). Microfungi are usually thought to be largely transient components of the microbiome in an attine nest (Poulsen & Currie, 2006). This view is supported by the observation that most microfungi associated with attine nests are ubiquitous species that ants may accidentally introduce into their nests. On the other hand, microfungi are usually found in high frequencies in gardens of attine ants in the tropics (e.g. 40.5% and 54%, in field and laboratory nests, respectively;Rodrigues et al., 2005, 2008). These high frequencies suggest a predictable rather than an accidental presence in attine gardens. Our study showed that fungus gardens of three North American attine species harbor as a diverse community of microfungi as gardens of tropical attine ants.

Fungus gardens ofC. wheeleri had proportionally more microfungi than gardens ofT. septentrionalis andA. texana (Fig. 1).Currie (1999a), working with attine ants from Panama, observed a similar pattern of microfungal prevalence across attine genera where gardens of lower-attine ants, such asCyphomyrmex sp., exhibited a higher prevalence of microfungi than gardens of higher-attine ants. In fact,Currie (2004) observed that, even in the presence of garden-tending workers, laboratory nests of lower-attine ants were occasionally overgrown by fungi other thanEscovopsis sp.; such severe garden infection by generalist microfungi was assumed to be rare in laboratory gardens of higher-attine ants (Currie, 2004), but additional data are needed to support this view.

The observed proportions of microfungi in attine gardens are likely associated with (1) the mycobiota found in soils next to ant nests (Hughes et al., 2004), (2) the type of plant substrate collected by the ants, or (3) both. Soil adjacent to ant nests may be a source of microfungi (Hughes et al., 2004) and ants at different sites may be in contact with different soil microfungal communities (see the discussion below). Another explanation is that lower- and higher-attine ants may treat microfungi in different ways. For example, lower-attine ants engage in a less elaborative substrate preparation process when compared with higher-attine ants (Mangone & Currie, 2007), and lower-attine ants utilize on average more diversified gardening substrate than higher-attine ants (De Fine Licht & Boomsma, 2010). Specifically, lower-attine ants remove alien contaminants after the substrate becomes incorporated into the fungus gardens. In contrast, higher-attines engage in a more elaborative substrate preparation by removing contaminants from plant substrate before incorporation into the fungus garden (Magone & Currie, 2007). Additionally, higher-attine ants have special behaviors and adaptations like weeding and grooming of fungus gardens in order to removeEscovopsis sp. (Currie & Stuart, 2001). Perhaps, these adaptations may also increase the removal of generalist microfungi in the garden matrix, contributing to the observed low proportions of microfungi inA. texana gardens in comparison with other attine species. In support of this view,Rodrigues (2004) observed that workers ofAtta sexdens rubropilosa intensively removed fungus garden fragments after inoculation with spores of microfungi (e.g.Syncephalastrum racemosum) other thanEscovopsis sp.

Our culture-dependent approach revealed that the most abundant microfungal species from gardens of the three target ant species wereA. ustus (2.4%, relative to the total number of isolates),A. ochraceus (1.8%),C. cladosporioides (6%),Cochliobolus sp. 1 (2.7%), andN. sphaerica (4.2%), well-known cosmopolitan representatives that are common in soil and plant substrates (Tables S2–S4).Rodrigues (2008) likewise found soil microfungi such asCunninghamella sp.,Fusarium sp., andTrichoderma sp. prevalent in gardens of leaf-cutting ants in south Brazil. These microfungal genera were also recovered in the present study, but they were represented by only a few isolates (Tables S2–S4). Although most microfungal species were not consistently present throughout all four sampling periods, a few species were persistent. For example, microfungal taxa such asC. cladosporioides, Cochliobolus sp. 1,F. solani, andN. sphaerica were found in all seasons inC. wheeleri gardens (Table S2). Again, these are ubiquitous taxa found in diverse environments such as soil, plant material, or even air samples (Samson et al., 2000; Crous et al., 2009).

Because no particular set of microfungi was associated with only one of the three ant species, our survey revealed no clear species-specific associations between microfungal species and the three ant species studied. Similarly,Rodrigues (2008) found a diffuse association of microfungal species and severalAcromyrmex species from southern Brazil. If more subtle species-specific interactions between microfungi and attine ants exist, these were not detected in the present study. A survey of a large number of gardens will be necessary to evaluate such subtle specificities.

Despite the apparent lack of specific ant–microfungus associations in our study, correspondence analysis revealed distinct microfungal communities between the three ant species studied (Fig. 3). These differences could be because ants foraged on different substrates, which may have contained different microfungal species. A second factor explaining the structuring of microfungal communities is habitat (collection location). For example, theA. texana samples collected in different sites cluster in two separate groups (samples collected at Horsnby Bend and at Buescher State Park) in the correspondence andunifrac analyses (Figs 3 and 4). Thus, ants collected in different sites may be exposed to different microfungal communities present in the soil and in the gardening substrates utilized, which could account for the observed differences. Indeed, the soil mycobiota next to ant colonies may be a source of microfungi that may potentially infects the fungus gardens (Hughes et al., 2004).

In contrast to our expectations, differences in microfungal diversity across seasons were not correlated with ant foraging behavior. For example, in the northern hemisphere, attine ants decrease their foraging activity during the colder periods of the year (Mintzer, 1979), and in extreme cases, such asT. septentrionalis, workers completely cease their activities during the winter (Weber, 1956; Seal & Tschinkel, 2008). Thus, if fungal diversity in gardens is mostly driven by the foraging behavior of ants, we expected that, compared with warmer seasons, we would find reduced diversity of microfungal species in months with low temperatures and little addition of substrate (Weber, 1972). However, this was not the case; rarefaction curves and diversity indices predicted the same number of species in gardens ofC. wheeleri andT. septentrionalis for the winter and summer seasons as for the winter and fall (Fig. S2, Table S5). Moreover, microfungi prevalence in garden fragments was not significantly different between winter and summer forC. wheeleri as for spring and fall forT. septentrionalis (Fig. S1). On the other hand, the results fromA. texana gardens seem to support our expectation because the diversity of microfungi species in spring and summer was higher than that for winter (Fig. S2). In fact, Chao 1 estimated microfungal species richness forA. texana gardens as 123 (spring) and 271 (summer), respectively, compared with 88 (winter) and 85 (fall) (Table S5). Our results indicate that additional factors other than the ant foraging behavior (see the above discussion on microfungal community in soils next to ant nests) contributed to the observed differences in microfungi prevalence (Fig. S1) and species richness (Fig. S2) over the seasons.

A surprising finding was the complete absence of the parasiteEscovopsis sp. in the gardens of all three surveyed attine ant species. Using the same isolation methods described above, additional sampling of gardens of 12 youngA. texana nests (∼6 months old) from east Texas revealed noEscovopsis sp. (A. Rodrigues, unpublished data). Finally, a nest ofA. texana from which hundreds of garden pieces were sampled by C.R. Currie from dozens of gardens in February 2001 did not yield anyEscovopsis isolate (U.G. Mueller & C.R. Currie, unpublished data). According toCurrie (1999a),Escovopsis sp. is found at high frequencies in gardens of attine ants from Central America, with as many as 50% of the colonies infected withEscovopsis, depending on the ant genus.Gerardo (2006) found that ants in the genusApterostigma (a lower-attine ant) from Central America may have up to 52% of gardens infected byEscovopsis sp. In south Brazil (State of Rio Grande do Sul),Escovopsis sp. prevalence in gardens was estimated at 27% forAcromyrmex gardens (Rodrigues et al., 2008). In contrast to these tropical populations, our study indicates thatEscovopsis appears to be rare or absent in attine gardens in Central Texas.

The observation thatEscovopsis sp. may be locally infrequent should stimulate further studies on the geographic variation ofEscovopsis sp. prevalence in attine gardens. Such studies would benefit from intensive and systematic sampling from different localities, especially in other northern temperate areas as well as southern temperate regions and high-altitude habitats. DeterminingEscovopsis sp. incidence in these localities might reveal additional aspects of the attine ant–microorganism symbiosis as observed in the cold-tolerant fungal symbionts cultivated byA. texana in the northern range of occurrence of this ant species (Mueller et al., 2011).

Using culture-dependent methods, the present study shows that the attine garden microbiome harbors diverse microfungal species. The absence of clear species-specific associations of garden microfungi or a clear correlation with ant foraging preferences suggests that most garden microfungi are transient components in such an environment. Although microfungi may act as transient antagonists of the ant cultivar (Silva et al., 2006b); this contrasts with the roles reported forBurkholderia sp. bacteria (Santos et al., 2004), actinobacteria (Sen et al., 2009), nitrogen-fixing bacteria (Pinto-Tomás et al., 2009), and plant biomass-degrading bacteria (Bacci et al., 1995b; Suen et al., 2010), which appear to be resident symbionts in attine gardens.

Supporting Information

Fig. S1. Seasonal changes in the proportion of garden fragments fromCyphomyrmex wheeleri,Trachymyrmex septentrionalis andAtta texana that yielded some microfungal growth.

Fig. S2. Individual-based rarefaction curves of microfungal communities associated with fungus-growing ants from a year-long survey in Central Texas.

Table S1. Number of microfungal isolates recovered from three attine ant species during a year-long survey in Central Texas (2005–2006).

Table S2. Microfungi associated with fungus gardens ofCyphomyrmex wheeleri surveyed inWinter, Spring, Summer, and Fall 2005/2006 in Central Texas.

Table S3. Microfungi associated with fungus gardens ofTrachymyrmex septentrionalis surveyed in Winter, Spring, Summer, and Fall 2005/2006 in Central Texas.

Table S4. Microfungi associated with fungus gardens ofAtta texana surveyed in Winter, Spring, Summer, and Fall 2005/2006 in Central Texas.

Table S5. Observed species richness, richness estimators, and diversity indices of microfungal communities associated with fungus gardens of attine ants in Central Texas.

Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.


We thank K. Anderson for permission to work at the Hornsby Bend Environmental Research Center and P. Schappert for permission to work at Stengl Lost Pines Biology Station. We are also grateful to M. Leibold for advice on community analyses and to three anonymous reviewers for helpful comments on this manuscript. This work was supported by a scholarship to A.R. from the International Cooperation Program Capes-UT Austin (02/05), and by NSF award DEB-0110073 to U.G.M. We thank the Brazilian agencies CNPq and FAPESP for financial support.


View Abstract