Search tips
Search criteria 


Logo of plosonePLoS OneView this ArticleSubmit to PLoSGet E-mail AlertsContact UsPublic Library of Science (PLoS)
PLoS One. 2010; 5(3): e9922.
Published online 2010 March 29. doi:  10.1371/journal.pone.0009922
PMCID: PMC2847949

Microbial Community Structure of Leaf-Cutter Ant Fungus Gardens and Refuse Dumps

Ching-Hong Yang, Editor



Leaf-cutter ants use fresh plant material to grow a mutualistic fungus that serves as the ants' primary food source. Within fungus gardens, various plant compounds are metabolized and transformed into nutrients suitable for ant consumption. This symbiotic association produces a large amount of refuse consisting primarily of partly degraded plant material. A leaf-cutter ant colony is thus divided into two spatially and chemically distinct environments that together represent a plant biomass degradation gradient. Little is known about the microbial community structure in gardens and dumps or variation between lab and field colonies.

Methodology/Principal Findings

Using microbial membrane lipid analysis and a variety of community metrics, we assessed and compared the microbiota of fungus gardens and refuse dumps from both laboratory-maintained and field-collected colonies. We found that gardens contained a diverse and consistent community of microbes, dominated by Gram-negative bacteria, particularly γ-Proteobacteria and Bacteroidetes. These findings were consistent across lab and field gardens, as well as host ant taxa. In contrast, dumps were enriched for Gram-positive and anaerobic bacteria. Broad-scale clustering analyses revealed that community relatedness between samples reflected system component (gardens/dumps) rather than colony source (lab/field). At finer scales samples clustered according to colony source.


Here we report the first comparative analysis of the microbiota from leaf-cutter ant colonies. Our work reveals the presence of two distinct communities: one in the fungus garden and the other in the refuse dump. Though we find some effect of colony source on community structure, our data indicate the presence of consistently associated microbes within gardens and dumps. Substrate composition and system component appear to be the most important factor in structuring the microbial communities. These results thus suggest that resident communities are shaped by the plant degradation gradient created by ant behavior, specifically their fungiculture and waste management.


Leaf-cutter ants (Atta and Acromyrmex: Tribe Attini) are the most phylogenetically derived members of a monophyletic group of Neotropical ants that cultivate specialized fungal symbionts that serve as the colony's primary food source. Unlike other fungus-growing ants, leaf-cutters are unique in that they exclusively use fresh plant material to cultivate their fungal mutualist [1][4]. The ants protect their fungal cultivar from pathogens and parasites [3], [5][7], provide the fungus with a constant source of nutrients [3], [4], and aid in its growth and dispersal [3]. Fungiculture in attine ants originated once approximately 50 million years ago while the transition to fungal cultivation using fresh plant material appears to have arisen about 40 million years later [8]. Leaf-cutters are one of the most dominant herbivores in their native habitats [3], [4], with mature colonies of Atta potentially containing millions of workers living in an elaborate subterranean system of hundreds of interconnected fungal chambers [4]. Colonies can harvest a large amount of plant material, for example a study of Atta colombica colonies in Panama found that they collected an average of ~250 Kg (dry weight) of plant material (including leaves, fruit, nuts, and flowers) per colony during a one-year observation period [9], [10]. Though a single colony will utilize up to 50% of the available local plant species, the bulk of plant material is harvested from relatively few plant species [9], [11] (Figure 1A). Foraging workers harvest fresh plant material that is brought back to the nest and incorporated into the top of the garden. To promote the initial degradation of plant biomass this material is masticated, mixed with ant fecal droplets, and inoculated with fungal mycelia [3] (Figure 1B).

Figure 1
Representation of the leaf-cutter ant system.

The process of substrate addition and fungal planting is continuous: over time, the top layer of the garden is replaced by new plant material and gradually moves to the bottom of the garden matrix. Studies suggest that the more easily degraded plant compounds, such as starch and pectin, are extracted in the top of the garden [12]. In addition, degradation of hemicelluloses [12] and cellulose also occur throughout the fungus garden, while lignin appears to pass through undegraded [13]. Partially degraded plant material is removed from the bottom of the garden and deposited in refuse repositories (dumps) by workers. Most leaf-cutter ant species excavate specialized subterranean refuse chambers [3], [14], while a few deposit their waste in specialized and localized aboveground piles [3], [15] (Figure 1C). Mature colonies produce large amounts of refuse material [16], and considerable colony resources are devoted to waste management [15]. In addition, refuse dumps also provide important nutrient reservoirs for surrounding plant and animal communities [17][19].

Thus, the leaf-cutter ant garden and dump system represents a plant-based biomass degradation gradient, where the flow of plant-derived nutrients is mediated by fungiculture and waste management. These behaviors result in the concentration and separation of nutrients into two spatially distinct elements: the fungus garden and the refuse dump. Fungus gardens receive continual inputs of fresh plant material, while garden waste serves as the principal inocula for the dumps. Given the different degree of degradation between gardens and dumps, these environments differ markedly in the relative amounts of many plant polysaccharides and available nutrients. Such nutrient gradients may create spatial structuring for unique and specialized microbial communities. Although many studies have demonstrated that fungus gardens can harbor a variety of microorganisms [20][26], it remains unknown whether many of these microbes in gardens and dumps represents resident members (i.e. not transient) or how uniform community composition is among host ant species. Furthermore, little is known about the compositional changes in microbial community structure between gardens and dumps. In this study, we address whether microbial community structure changes significantly from gardens to dumps. Specifically, we assess if the relative abundance of microbial populations change significantly between these two colony components, and if these changes are similar across colonies.

To address this question, we examine the microbiota in the leaf-cutter ant system using a modified version of microbial membrane lipid analysis, a well-established method for examining broad-scale microbial community structure [27][30]. Specifically, using phospholipid fatty acid/fatty acid methyl ester analysis [31], [32] we investigate field-collected and laboratory-maintained colonies of leaf-cutter ants to evaluate the microbial community structure of fungus gardens and refuse dumps, and then use a variety of metrics to compare microbial community structure. Overall, our results indicate that community structure is more similar based on location/substrate (i.e. garden or dump) rather than environment/colony source (lab or field). Analysis of lab colonies indicates strict and consistent community segregation between gardens and dumps. Though our data shows some affect of the lab environment on overall community structure in gardens and dumps, we find this same pattern of segregation reflected in field colonies. Correlation of the detected lipids to their taxonomic designation reveals that in general, gardens contain predominately Gram-negative bacteria. Gram-negative taxa are also present in dumps but our results show a significant enrichment of Gram-positive and anaerobic bacteria. The apparent spatial differences in community structure between leaf-cutter ant gardens and dumps suggest that there is specific partitioning, likely due to substrate availability mediated by ant agricultural practices.

Materials and Methods

Sample collection and nomenclature

We sampled gardens and refuse dumps of three species of Atta (A. cephalotes, A. colombica, and A. sexdens) and four species of Acromyrmex (A. laticeps, A. octospinosus, A. hispidus fallax, A. niger) leaf-cutter ants. Colonies sampled included, nine laboratory-maintained and five field-collected colonies (Table S1). Lab colonies (L) represented species of both Atta and Acromyrmex collected in 2003 from sites in Argentina and Panama (Table S1). Lab colonies were maintained at the University of Wisconsin–Madison, housed in 55 L plastic containers consisting of two smaller containers, one for the fungus garden and the other for the refuse dump (Figure 1D). Colonies were fed three times per week using a mixture of oak and maple leaves. Laboratory colonies were first sampled for lipid analysis in 2006 and again in 2008. Field samples were collected from five mature colonies of Atta colombica in the Canal Zone of central Panama during a 12-day period from 30 May to 10 June, 2008 (N 09° 07′ W 079° 42′). Most species of leaf-cutter ants excavate deep, subterranean refuse chambers. These chambers are difficult to sample [14], and as such, we chose to focus exclusively on A. colombica because refuse material is considerably more accessible. Since mature field colonies forage on a wide variety of host plants, even within a single day [9], it was not possible to detail the type of plant material used by each colony at the time of collection. Two fungus gardens and several dump samples were collected from each colony. Field samples were aseptically collected, stored at −20°C, and subsequently transported on dry ice to our lab at the UW–Madison for lipid analysis. In total, our analysis consisted of 62 samples from four major groups: i) lab gardens (LG, sampled in 2006 and 2008), ii) lab dumps (LD, sampled in 2006 and 2008), iii) field gardens (FG, sampled in 2008), and iv) field dumps (FD, sampled in 2008) (Table S1).

Hybrid PLFA-FAME analysis

Samples of garden and dump were frozen at −20°C, lyophilized, and milled to #40 mesh size prior to phospholipid fatty acid (PLFA) extraction. Membrane lipids were extracted from 1 g lyophilized and milled material in a two-phase aqueous-organic extraction, as described by Bligh and Dyer [33]. Samples were extracted twice and the supernatants were combined after phase separation overnight. The organic phase was isolated and evaporated to dryness using a RapidVap (LabConco, Kansas City, MO). We performed fatty acid methyl ester (FAME) analysis as described by Microbial ID Inc. (MIDI) [34] on the dried organic residue. Lipids were saponified and then subjected to alkaline methanolysis. Lipids were isolated from the mixture in a hexane extraction.

Lipid methyl esters were analyzed in a 2 µL injection using a Hewlett-Packard 6890 Gas Chromatograph configured and maintained for lipid analysis according to the recommendations of MIDI [34]. Gas chromatogram parameters were specified and peaks were identified by the MIDI EUYKARY method (MIDI, Newark, DE). Fatty acid concentration was quantified by comparing peak areas of the samples compared to two internal standards, 9[ratio]0 (nonanoic methyl ester) and 19[ratio]0 (nonadeconoic methyl ester) (Sigma, St. Louis, MO) of known concentration. In all subsequent statistical analyses, we excluded fatty acids that were at an average abundance of <0.5 mol% or present in less than three of the 62 samples.

The following standard fatty acid nomenclature [35], briefly summed up by the following formula, is used throughout this article: A:XωB, where A is the number of carbon atoms in the lipid chain; X is the number of double bonds; B is the position (ω) of the double bond from the methyl end. The prefixes i and a represent iso and antiso branching, respectively. The prefix cy indicates cyclopropane fatty acid and the suffix 10Me refers to a methyl group on the tenth carbon. The suffixes c and t represent cis or trans configuration, respectively [35].

Community Analysis

Richness estimators

Lipid diversity indices were calculated from the complete dataset using EstimateS v.8.2.0 [36]. We used the Mao Tau function to calculate observed richness (accumulation curves) and the Chao2 function to calculate estimated richness. We used lower and upper Chao2 log-linear 95% confidence intervals (CI) to create a confidence limit (CL) envelope (Figure 2). Samples were randomized 1000 times without replacement using the biased corrected Chao2 formula.

Figure 2
Rarefaction and lipid accumulation curves.

Analysis of samples using major lipid classes

We grouped lipid biomarkers into seven chemically-related classes [37]: saturated fatty acids; hydroxyl fatty acids; monounsaturated fatty acids; cyclic fatty acids; methylated fatty acids; branched fatty acids; and polyunsaturated fatty acids. Certain lipids were too close in size to be separated by gas chromatography, and therefore, an exact identification was not possible. These lipids were grouped together as “Unclassified”. This approach is recognized as an effective means of comparison because major lipid classes correspond to particular groups of organisms [38]. Absolute lipid concentrations (mol%) were averaged by class across sample groups (lab gardens, lab dumps, field gardens, field dumps). Analysis of Variance (ANOVA) and Tukey-Kramer highly significant difference (HSD) tests were performed using the JMP software package v.8.0.1 [39] (Table 1). We also calculated the relative contribution (mol%) of major lipid classes for each sample group (Figure 3).

Figure 3
Contribution of major lipid classes to community structure.
Table 1
Average amount (mol%) of major lipid classes.

Principal component analysis (PCA)

For multivariate PCA, all lipids (mol%) were transformed using the square root of the arcsine to approximate normality [40] prior to analysis using the JMP software package v.8.0.1 [39]. We used PCA to explore the following comparisons between the major sample groups: a) lab gardens vs. field gardens; b) lab dumps vs. field dumps; c) lab gardens vs. lab dumps); and d) field gardens vs. field dumps (Figure 4).

Figure 4
Principal Component Analysis (PCA) of lipid markers recovered from the four sample types.

Cluster analysis of lipid biomarkers

Lipid concentrations (mol%) for all samples were compiled into two different profile matrices. In the first profile matrix (the sample matrix), each row corresponded to a sample and each column corresponded to a lipid. In total, we omitted six lipid markers that were common to all samples because they did not affect clustering (9[ratio]0 and 19[ratio]0, which were internal standards and 14[ratio]0, 16[ratio]0, and 18[ratio]0, which are abundant in most organisms [38]). Otherwise, the full lipid dataset was used for clustering analysis. In the profiles second matrix (the lipid matrix), each row corresponded to a lipid and each column corresponded to a sample. Next we used each profile matrix to construct a corresponding similarity matrix by calculating the correlation between all sample pairs and lipid pairs, respectively, using Spearman's rank correlation (with tie-correction), an approach that has been used to effectively cluster a variety of genetic datasets [41][43]. We then used each similarity matrix to generate corresponding dendrograms using the neighbor-joining program in PHYLIP [44] and visualized with the Phylodendron phylogenetic tree printing webserver (, accessed: 11/09/2009). Thus, we created a sample dendrogram (S) and a lipids dendrogram (L). We then reorganized the original sample profile matrix to reflect the topology of each dendrogram, samples on the horizontal axis and lipids on the vertical axis. In order to approximate normality, mol% concentrations were transformed using the square root of the arcsine [40] and each raw arcsine transformed data point was represented in a heat map using the Microsoft Excel Yabs Heat map macro (, accessed 11/09/2009) (Figure 5).

Figure 5
Heat map and dendrograms of lipid distribution.

Correlation of lipids to microbial taxonomy

To determine the set of common microbial taxa within each of the four major groups (LG, FG, LD, and FD) we designed a lipid-to-microbial map. This allowed us to identify lipids common to samples within groups and then correlate each lipid to known microbial classification. Using previously published reports [28], [32], [38], [45][54] we first created a lipid-microbe classification table to tabulate each sample's complement of microbial taxa. Only lipids corresponding to reported taxa were used in the analysis. In total, we identified 31 lipids from our full dataset that corresponded to reported taxa. The remaining lipid markers had no known identity and thus could not be considered in this particular analysis. For each lipid, we calculated average mol% concentration (±S.E.) within groups. Finally, we used ANOVA and Tukey's HSD to assess differences in microbial taxa between major groups (LG, FG, LD, and FD). In some cases, individual lipids corresponded to more than one group of organisms. In these instances, all taxonomic identifications were considered in the analysis, and the first one listed occurred most often in the literature (Table 2).

Table 2
Indicator lipids from lab (2008) and field colonies.


We determined the structure of microbial communities in leaf-cutter ant gardens and dumps. Our sampling of gardens (n = 17) and dumps (n = 17) from lab-maintained colonies and field-collected gardens (n = 9) and dumps (n = 19) from five mature colonies of Atta colombica resulted in a complete dataset that consisted of 156 lipid markers (Table S2).

Lipid richness and abundance

Lipid richness

We calculated lipid richness for lab gardens, lab dumps, field gardens, and field dumps (Figure 2). Richness analysis indicated that gardens were well sampled, and that richness was similar between lab and field fungus gardens. Lab dumps were also sufficiently sampled; however, field dumps were significantly richer than all other groups and expressed unsampled diversity.

Abundance of major lipid classes

We separated lipids into the following seven chemically-related classes and calculated the average mol percentage (mol%) contribution of each class: saturated fatty acids, monounsaturated fatty acids, hydroxy fatty acids, cyclic fatty acids, methylated fatty acids, branched fatty acids and polyunsaturated fatty acids. Lab and field gardens were composed almost entirely of saturated and monounsaturated fatty acids. Hydroxy fatty acids, though not as abundant, constituted the third largest lipid class detected in the gardens (Figure 3). Dumps contained these three classes, but were also enriched for cyclic fatty acids, methylated fatty acids, branched fatty acids, and polyunsaturated fatty acids (Figure 3). One-way ANOVA revealed that each of the seven major lipid classes differed significantly across the four sample groups: saturated fatty acids, F(3,58) = 14.33, p<0.001; monounsaturated fatty acids, F(3,58) = 5.84, p<0.001; hydroxy fatty acids, F (3,58) = 3.67, p = 0.0172; cyclic fatty acids, F (3,58) = 39.06, p<0.001; methylated fatty acids, F(3,58) = 13.74, p = 0.0015; branched fatty acids; F(3,58) = 91.27, p<0.001; polyunsaturated fatty acids, F(3,58) = 18.71, p<0.001.

Tukey-Kramer post-hoc comparisons revealed that gardens from lab and field colonies differed only in the amount of saturated fatty acids present (p<0.001) (Table 1), and there was no significant difference in the amount of other six lipid classes (Table 1). Lab dumps contained greater amounts of hydroxy fatty acids (p = 0.035) and branched fatty acids (p = 0.018) than that of field dumps, but lower amounts of methylated fatty acids (p = 0.036). Lab and field dumps did not differ in the amounts of saturated fatty acids, monounsaturated fatty acids, cyclic fatty acids, and polyunsaturated fatty acids (Table 1).

We found no difference in the amounts of saturated fatty acids and hydroxy fatty acids between lab gardens and lab dumps (Table 1). Lab dumps had significantly greater amounts of branched fatty acids (p<0.001), methylated fatty acids (p = 0.024), cyclic fatty acids (p<0.001), and polyunsaturated fatty acids (p<0.001) than lab gardens. Lab gardens contained significantly greater amounts of monounsaturated fatty acids compared to lab dumps (p = 0.0024). Field dumps contained significantly greater amounts of branched fatty acids (p<0.001), methylated fatty acids (p<0.001), cyclic fatty acids (p<0.001), and polyunsaturated fatty acids (p = 0.025) than that of field gardens (Table 1). Field gardens had significantly greater amounts of saturated fatty acids than dumps (p<0.001). We found no differences in the amount of monounsaturated fatty acids or hydroxy fatty acids levels (Table 1).

Principal component analysis

We used Principal Component Analysis (PCA) to assess similarity in microbial community structure across all samples (Figure 4). In total, we performed four different PCAs representing the four main comparisons used throughout this study. PCA of field samples revealed garden samples clustered to the exclusion of dump samples separated primarily along Principal Component 1 (PC1) (Figure 4A). Dump samples again exhibited considerable variation and no clustering based on host ant colony was observed (data not shown). PC1 accounted for 20.1% of the observed variation and PC2 accounted for 12.7% of the variation. Lab samples exhibited a similar pattern in that gardens and dumps formed distinct clusters (Figure 4B). Based on the separation of samples along PC1, clustering was related to the source of the sample (either garden or dump) and separation of samples along PC2 was related to year of sampling (2006 or 2008). PC1 accounted for 23.5% of the observed variation and PC2 accounted for 11.9% of the variation. PCA of garden samples from lab and field colonies indicated that clustering along PC1 was explained by sample source (lab or field) (Figure 4C). Lab samples clustered according to year sampled along PC2. PC1 accounted for 19.1% of the variation and PC2 accounted for 12.6% of the variation. Examination of field samples revealed considerable variation between samples and no clustering was evident based on nest or location origin (data not shown). Dump samples from lab colonies formed a single, tight cluster from both 2006 and 2008 (Figure 4D) likely due to the large variation of field samples. PC1 accounted for 17.8% of the variation and PC2 accounted for 9.2% of the variation.

Heat map and clustering analysis

We performed heat map and clustering analysis using a subset of named lipid markers in our dataset to help explain the observed patterns from our richness estimators, major lipid class analysis, and PCA. Lipids formed two main clusters within the lipid (L) analysis, groups L1 and L2, as shown in Figure 5. Furthermore, we found that gardens and dumps also formed individual clusters within the sample (S) dendrogram. Group S1 was composed entirely of garden samples (from lab and field colonies) and sample group S2 contained only lab and field dump samples. Thus, sample clustering was related to microhabitat type (gardens or dumps) and not sample source (lab or field). However, sub-clustering was apparent within each major group, with lab samples forming their own groups distinct from field samples.

Further analysis revealed that gardens were typically composed of group L2 lipids (specifically hydroxy fatty acids: 8[ratio]0-3OH; 10[ratio]0-3OH; 11[ratio]0-3OH; 18[ratio]0-2OH; and monounsaturated fatty acids: 16[ratio]1ω5c; 17[ratio]1ω8c; 18[ratio]1ω9c) and contained very few lipid markers from group L1. In contrast, dumps were enriched for group L1 lipids and clustered based primarily on branched (i14[ratio]0, a15[ratio]0, i15[ratio]0, i15[ratio]1, i16[ratio]0, i17[ratio]0, a17[ratio]0, i17[ratio]0-3OH) and cyclic (cy17[ratio]0 and cy19[ratio]0) fatty acids. Branched fatty acids were absent from garden samples (with the exception of a17[ratio]0, present in 18 of 26 samples). Cyclic fatty acids were present in less than half of the field garden samples and only one lab garden sample. Cluster analysis revealed no grouping of lab samples by ant genus (neither gardens nor dumps), or field samples according to source nest. There was a major shift from group L2 lipids in lab gardens to group L1 lipids in lab dumps. L2 lipids were very rare in lab dumps.

Correlation of lipid markers to microbial taxonomy

We analyzed 31 individual lipids across lab gardens, lab dumps, field gardens, and field dumps in order to determine which lipids were enriched between samples. Table 2 summarizes these results.

Analysis of taxa between garden samples

Ten of the 31 (32%) markers were not found in either lab or field gardens, while only 7 of 31 markers tested (23%) were found in significantly different amounts between lab and field gardens. Three were found to be present in greater amounts in lab gardens than in field gardens: 2 lipids corresponding to Gram-negative taxa (17[ratio]1ω8c and 18[ratio]1ω5c) and 1 actinomycete marker (19[ratio]0 10Me). Four lipids were found to be present in greater amounts in field gardens than in lab gardens, including a general Gram-negative marker (18[ratio]1-2OH), a marker specific to γ-Proteobacteria (Enterobacter/Vibrio) (10[ratio]0-3OH) and 2 Gram-positive markers, specifically corresponding to bacteria in the phylum Firmicutes (Bacillus, Lacobacillus, and Enterococcus; i19[ratio]0 and 20[ratio]2ω6c).

Analysis of taxa between dump samples

Comparison of dump samples from lab and field colonies revealed that 12 of 31 (39%) markers were detected in significantly different amounts. Of these, 8 (26%) markers were found in greater amounts in lab dumps than in field dumps: 3 Gram-negative markers; 10[ratio]0-3OH, i15[ratio]-3OH, and i17[ratio]0-3OH, corresponding to Enterobacter-Vibrio, Xanthomonas-Pseudomonas, and Cytophaga-Flexibacter-Bacteroides, respectively. Additionally, 3 markers were specific to the Bacillus-Clostridium group (18[ratio]1ω9c, i15[ratio]0, and i16[ratio]0), 1 to general Gram-positive (a15[ratio]0), and 1 to Gram-negative anaerobes (cy17[ratio]0). In total, 4 markers (13%) were found in greater amounts in field dumps than in lab dumps. These include 3 markers indicating the presence of Gram-positive taxa, specifically the Bacillus-Clostridium group (BC) (i11[ratio]0, a13[ratio]0, i13[ratio]0) and 1 Gram-negative marker (16[ratio]1ω5c). In general, lab dumps contained higher levels of markers indicative of both Gram-negative and Gram-positive bacteria.

Analysis of taxa within lab colonies

Of the 31 lipids analyzed, 18 (58%) were detected in significantly different (p<0.05) amounts in gardens compared to refuse dumps (Table 2). Of these, only 2 (11%) lipids were encountered in greater amounts in the gardens and 16 (89%) were present in greater amounts in the dumps. Garden samples contained greater amounts of 18[ratio]1ω5c and 19[ratio]0 10Me, which correspond to Gram-negative bacteria and Actinobacteria, respectively. We did not detect lipid 18[ratio]1ω5c in any dump samples. Of the 16 lipids present in greater amounts from dumps samples, 8 (50%) correspond to Gram-positive taxa and 2 (13%) lipids indicated the presence of Gram-negative anaerobes (cy17[ratio]0 and cy19[ratio]0). Lab dumps were also enriched with a different type of actinomycete marker (18[ratio]0 10Me) as well as 5 markers indicative of Gram-negative bacteria. Of the 16 markers detected in greater amount from dump samples, 12 (75%) were not detected in garden samples.

Analysis of taxa within field colonies

Analysis of field samples revealed similar patterns to lab colonies. Of the 31 lipids analyzed, 20 (65%) were found in significantly different amounts between field garden and dumps. Of these, 3 (15%) were present in greater amounts in fungus gardens and 17 (85%) were higher in refuse dumps (7 of which were not detected in garden samples). Similar to lab dumps, field dumps are enriched for Gram-positive bacteria. In fact, 9 of the markers present in greater amounts from dumps correspond to Gram-positive taxa including members of the Firmicutes (Bacillus-Clostridium). Again, as seen in lab dumps, field dumps were enriched for lipids indicating the presence of Gram-negative anaerobes (cy17[ratio]0 and cy19[ratio]0).


In this study, we present the first broad-scale analysis of microbial community structure within, and across, leaf-cutter ant colonies using microbial lipid analysis. Membrane lipids, essential components of living cells, provide a reliable and quantitative assessment of general microbial community structure [29], [30], [38]. Furthermore, because lipids exhibit structural diversity that can be linked to specific microbial taxa, community assemblages can be resolved, to some degree, at a broad phylogenetic scale [32], [54], [55]. Perhaps the biggest strength of a lipid-based approach, as compared to other microbial community assays, is that because phospholipid fatty acids are not found in storage molecules and degrade rapidly during cell death, it provides a census of the current living community (i.e., it is a proxy for living biomass) [28]. Many genetic-based approaches contain inherent biases such as unequal efficiency in DNA extraction, preferential amplification of certain taxa, and overrepresentation of particular groups in public databases [29], [56]. Further, genetic approaches reveal total diversity and say little about expressed diversity. Because diversity in many systems is sizable, genetic approaches may fail to identify fine-scale treatment effects [28]. Thus, lipid membrane analysis provides a logical first step for detecting shifts in community structure and establishes a framework for more intensive analyses of community composition.

Using microbial lipid analysis and various community metrics, we demonstrate that the fungus gardens and refuse dumps of leaf-cutter ants are composed of a diverse, yet distinct, community of microbes. In all colonies surveyed, the same pattern of broad-scale community change and segregation was apparent. In general, gardens contained a high abundance of Gram-negative bacteria, specifically markers indicative of Proteobacteria and Bacteroidetes. Though there was some affect of colony source (i.e. lab vs. field) on sample relatedness, we found striking similarities among community structure of lab-maintained and field-collected gardens, indicating the possibility that spatiotemporally stable communities are consistently maintained over time. Thus, our findings represent important support of other recent work [26] specifically indicating that the fungus gardens of leaf-cutter ants contain additional microbial symbionts to the fungal cultivar.

In contrast to gardens, dumps were found to be more diverse and enriched for Gram-positive and anaerobic bacteria. In addition, field dumps contained considerably richer communities than lab dumps and our results indicated the presence of a commonly-associated community across all samples, even though there was some affect of colony source on sample relatedness. This study is among the first to assess the microbial composition of leaf-cutter ant refuse dumps.

Community structure of leaf-cutter ant colonies

Our findings that microbial communities differ markedly between gardens and dumps are consistent across colonies (in both lab and field), host genera (Atta and Acromyrmex; lab only), and time (2006 vs. 2008; lab only). Our analysis of field colonies indicates significant shifts in community structure between gardens and dumps (Figure 2, ,3,3, ,4A,4A, and and5,5, Table 1 and and2).2). Given that refuse dumps of Atta colombica are exposed to numerous environmental inputs while field gardens remain relatively sheltered from the environment in subterranean chambers, these results are perhaps not too surprising. However, our data reveal similar patterns in lab colonies. Here too, large-scale and consistent community changes are evident between gardens and dumps (Figure 3, ,4B,4B, and and5,5, Table 1 and and2).2). Many of the environmental variables inherent to field colonies are likely greatly reduced, or completely absent, in the laboratory. Moreover, gardens and dumps in the lab are housed in close proximity within the same container (Figure 1D), further minimizing any potential effects of spatial separation.

Specifically, we find that microbial communities of fungus gardens, whether from lab or field colonies, are more similar to each other than to their corresponding refuse dumps. Diversity indices reveal no difference in taxa richness between lab and field gardens (Figure 2). Relatedness among samples reflects the source of the host colony (lab or field; Figure 4C and and5);5); however, our finding that Gram-negative bacteria dominate gardens is consistent across lab and field colonies. Direct analysis of lipids common between gardens reveal the presence a number of groups including those belonging to γ-Proteobacteria and Bacteroidetes (Table 2). Overall, the similarities between lab and field gardens are unexpected given that they are exposed to dramatically different environmental conditions (and hence potential microbial colonizers). For example, lab gardens are maintained in relatively sterile containers and are fed a very limited diet. In contrast, field gardens reside in close quarters to the surrounding soil and colonies forage on a wide variety of plant species [9]. A strict lab diet would likely contribute similar microbial inputs whereas a varied diet potentially contains a much greater diversity of microorganisms [57] and thus represents a larger source of possible colonizers. This finding provides evidence for strong substrate control over bacterial community structure, in contrast to the more commonly assumed environmental control.

We also demonstrated that lab dumps share many community members with field dumps. This is in spite of the fact that a) taxa diversity appeared much more limited in the lab environment and b) field dumps exhibited considerable variation between samples (Figure 2 and and4D).4D). As is in the case of fungus gardens, relatedness of microbial communities in refuse dumps generally reflected sample source (Figure 4D and and5).5). In the lab, refuse dumps are maintained in relatively sterile conditions and have no exposure to external environmental variables. Conversely, field dumps are exposed to continual environmental inputs from rain, soil, falling debris, and animal and plant incursions [15]. It is possible that many members of field dump microbial communities are recruited from outside the ant system via some combination of these factors. Alternatively, it is plausible that the dump community is specialized, and that refuse dumps of other colonies act as a source of microbial colonizers via a yet unknown mechanism. Regardless, while lab dump diversity levels seemed to be hindered by limited environmental inputs, our data indicated that lab and field dumps share many community characteristics. For example, phylogenetic analysis of select markers in these dumps revealed that the majority of taxa (61%) were present in equal concentrations (Table 2). Unlike fungus gardens, both lab and field dumps were enriched for Gram-positive (specifically in the Bacillus-Clostridium group) and anaerobic bacteria (Table 2). In fact, many taxa present in refuse dumps were not detected in fungus gardens (Figure 3 and and5,5, Table 2). Since garden waste material is the main source of microbial inoculum for dumps (especially in lab colonies), this indicated that certain organisms pass through the fungus garden and flourished in the dump.

Although we did not test the effect of host ant taxa on microbial community structure using field samples, our analyses of lab colonies revealed no effect of host ant genus (Atta vs. Acromyrmex) on microbial community structure in either gardens or dumps (Figure 4B and and5).5). Though Atta and Acromyrmex are known to cultivate genetically similar fungal cultivars [58], [59], many aspects of their natural history are fundamentally different (e.g. colony size, degree of worker polymorphism, habitat preference, forage material, etc) [3]. Additionally, although our data showed some effect of sample year on community structure in lab colonies (Figure 4B and and5),5), the same pattern of partitioning was evident (Figure 5). Overall, we found that gardens and dumps retained their community structure across ant genera and time (2006 and 2008), with Gram-negative bacteria dominating in the gardens and Gram-positive bacteria abundant in the dumps. This pattern strongly suggests that these assemblages can be strictly maintained.

Leaf-cutter Ant Colonies as Natural Composts

In general, our data suggest that gardens are dominated by Gram-negative bacteria and dumps by Gram-positive bacteria. Given that community structure seems largely independent of external environmental variables (temperature, humidity, rainfall, etc.), host ant taxa, and spatiotemporal factors, we propose an alternative explanation for this pattern of community partitioning. As we suggest above, leaf-cutter ant colonies are a two-stage plant biomass degradation system. This is analogous to human composting activities, a process whereby a complex community of microorganisms under controlled conditions degrades organic material [60]. Microbial community succession in composting systems is coupled with the degradation of plant cell wall polysaccharides (hemicellulose, cellulose, and lignin) [61]. Using an agricultural waste compositing system, Yu et al. [61] found that early stages of composting were marked by hemicellulose and cellulose degradation and dominated by Gram-negative bacteria, specifically Proteobacteria [61]. Later stages of composting were characterized by increased lignin degradation coupled with shifts in microbial communities to more Gram-positive taxa such as Actinobacteria. In a study of the fungus garden microbiome of Atta colombica leaf-cutter ants, Suen et al. [13] calculated the amount of plant polysaccharides in the top (input) and bottom (output) of fungus gardens. The authors determined that hemicellulose and cellulose degradation occurred through the fungus garden while lignin content remained relatively unchanged. Thus, garden waste (input for the refuse dump) had higher proportions of lignin and lower proportions of hemicelluloses and cellulose, relative to fungus garden plant biomass. Additionally, they demonstrated the abundance and dominance of Gram-negative bacteria in the fungus garden, specifically those belonging to the phyla Proteobacteria and Bacteroidetes, which confirms results presented in this study.

The microbial composition data presented in this study, together with previous reports of plant polysaccharide degradation through these colonies, provides strong evidence that leaf-cutter ant colonies may play a role as natural composts in tropical ecosystems. The fungus garden may thus serve as the initial site for early-stage composting while the dump serves as a concentrated area of late-stage composting. These processes are strictly maintained by the fungicultural behavior of the ants: plant biomass degraded by the fungal garden microbiota are removed and deposited into refuse dumps. This segregating behavior could thus act as a mechanism for the observed microbial structure in the fungus gardens and dumps of leaf-cutter ant colonies. By partitioning substrate in this manner, these ants likely promote the concentration and proliferation of specific microbial members throughout the different components of their colonies. This segregating behavior is potentially advantageous for the ants, as the strict maintenance of a specific microbial community in the garden may promote the stable maintenance of the fungus garden, while increasing the growth of their mutualistic fungus, thereby increasing food production. While in the dumps, greater diversity may increase the efficiency of waste degradation, and require less behavioral intervention by the ants. Our study of the microbial communities associated with leaf-cutter ant colonies provides insight into how host behavior can potentially influence the composition of microbial assemblages, and further illustrates the important role of symbiotic associations in shaping the dynamics of these communities.

Supporting Information

Table S1

Collection information.

(0.13 MB DOC)

Table S2

Lipid quantites (mol%) found in each sample.

(0.20 MB PDF)


We thank the Autoridad Nacional del Ambiente (ANAM) for permission to collect material in the Republic of Panama and the Smithsonian Tropical Research Institute (STRI) in Panama for logistical support during field collection. We thank U. Mueller and S. Price for colony collections in 2003, and J. Kao-Kniffin for help with initial lipid extractions. We also thank Harry Reed for data aquisition and curation, and E. Caldera for assistance with fieldwork. J. Hite and J. Hulcr for advice and assistance with statistical and community analysis, respectively. Finally, we thank M. Poulsen, J. Hulcr, E. Caldera, K. Grubbs, S. Marsh, for helpful suggestions on earlier versions of this manuscript, and T. Schultz and one anonymous reviewer for constructive comments.


Competing Interests: The authors have declared that no competing interests exist.

Funding: This work was funded by the National Science Foundation [grants DEB-0747002 (to CRC), MCB-0702025 (to CRC and TCB), and MCB-0731822 (to CRC)] and the DOE Great Lakes Bioenergy Research Center (DOE BER Office of Science DE-FC02-07ER64494, to (CRC)). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


1. Belt T. London: J. Murray.; 1874. The Naturalist in Nicaragua. p. 403 p.
2. Weber NA. Fungus-growing ants. Science. 1966;153:587–604. [PubMed]
3. Weber NA. Philadelphia: American Philosophical Society.; 1972. Gardening Ants: The Attines. p. 146 p.
4. Hölldobler B, Wilson EO. Cambridge, Mass: Belknap Press of Harvard University Press.; 1990. The Ants. p. 752 p.
5. Currie CR, Stuart AE. Weeding and grooming of pathogens in agriculture by ants. Proc Roy Soc B–Biol Sci. 2001;268:1033–1039. [PMC free article] [PubMed]
6. Little AEF, Murakami T, Mueller UG, Currie CR. The infrabuccal pellet piles of fungus-growing ants. Naturwissenschaften. 2003;90:558–562. [PubMed]
7. Poulsen M, Bot AN, Boomsma JJ. The effect of metapleural gland secretion on the growth of a mutualistic bacterium on the cuticle of leaf-cutting ants. Naturwissenschaften. 2003;90:406–409. [PubMed]
8. Schultz TR, Brady SG. Major evolutionary transitions in ant agriculture. Proc Nat Acad Sci USA. 2008;105:5435–5440. [PubMed]
9. Wirth R, Herz H, Ryel RJ, Beyschlag W, Hölldobler B. Berlin Heidelberg New York: Springer-Verlag; 2003. Herbivory of leaf-cutting ants: A case study on Atta colombica in the tropical rainforest of Panama.
10. Herz H, Beyschlag W, Hölldobler B. Herbivory rate of leaf-cutting ants in a tropical moist forest in Panama at the population and ecosystem scales. Biotropica. 2007;39:482–488.
11. Rockwood LL. Plant selection and foraging patterns in two species of leaf-cutting ants (Atta). Ecology. 1976;57:48–61.
12. Schiøtt M, De Fine Licht HH, Lange L, Boomsma JJ. Towards a molecular understanding of symbiont function: identification of a fungal gene for the degradation of xylan in the fungus gardens of leaf-cutting ants. BMC Microbiol. 2008;8:40. [PMC free article] [PubMed]
13. Suen G, Scott JJ, Aylward FO, Adams SM, Tringe SG, et al. An insect herbivore microbiome with high plant biomass-degrading capacity. 2010. manuscript submitted. [PMC free article] [PubMed]
14. Moser JC. Contents and structure of Atta texana nest in summer. Ann Entomol Soc Am. 1963;56:286–291.
15. Hart AG, Ratnieks FLW. Waste management in the leaf-cutting ant Atta colombica. Behav Ecol. 2002;13:224–231.
16. Herz H, Beyschlag W, Hölldobler B. Assessing herbivory rates of leaf-cutting ant (Atta colombica) colonies through short-term refuse deposition counts. Biotropica. 2007;39:476–481.
17. Farji-Brener AG, Medina CA. The importance of where to dump the refuse: seed banks and fine roots in nests of the leaf-cutting ants Atta cephalotes and A. colombica. Biotropica. 2000;32:120–126.
18. da SternbergLS, Pinzon MC, Moreira MZ, Moutinho P, Rojas EI, et al. Plants use macronutrients accumulated in leaf-cutting ant nests. Proc Roy Soc B–Biol Sci. 2007;274:315–321. [PMC free article] [PubMed]
19. Hudson TM, Turner BL, Herz H, Robinson JS. Temporal patterns of nutrient availability around nests of leaf-cutting ants (Atta colombica) in secondary moist tropical forest. Soil Biol Biochem. 2009;41:1088–1093.
20. Craven SE, Dix MW, Michaels GE. Attine fungus gardens contain yeasts. Science. 1970;169:184–186. [PubMed]
21. Bacci MB, Ribeiro SB, Casarotto MEF, Pagnocca FC. Biopolymer-degrading bacteria from nests of the leaf-cutting ant Atta sexdens rubropilosa. Braz J Med Biol Res. 1995;28:79–82.
22. Carreiro SC, Pagnocca FC, Bueno OC, Bacci MB, Hebling MJA, et al. Yeasts associated with nests of the leaf-cutting ant Atta sexdens rubropilosa Forel, 1908. Antonie Leeuwenhoek. 1997;71:243–248. [PubMed]
23. Currie CR, Scott JA, Summerbell RC, Malloch D. Fungus-growing ants use antibiotic-producing bacteria to control garden parasites. Nature. 1999;398:701–704.
24. Currie CR, Mueller UG, Malloch D. The agricultural pathology of ant fungus gardens. Proc Nat Acad Sci USA. 1999;96:7998–8002. [PubMed]
25. Little AEF, Currie CR. Black yeast symbionts compromise the efficiency of antibiotic defenses in fungus-growing ants. Ecology. 2008;89:1216–1222. [PubMed]
26. Pinto-Tomás AA, Anderson MA, Suen G, Stevenson DM, Chu FST, et al. Symbiotic nitrogen fixation in the fungus gardens of leaf-cutter ants. Science. 2009;326:1120–1123. [PubMed]
27. Frostegård A, Petersen SO, Bååth E, Nielsen TH. Dynamics of a microbial community associated with manure hot spots as revealed by phospholipid fatty acid analyses. Appl Environ Microb. 1997;63:2224–2231. [PMC free article] [PubMed]
28. Kaur A, Chaudhary A, Choudhary R, Kaushik R. Phospholipid fatty acid - A bioindicator of environment monitoring and assessment in soil ecosystem. Curr Sci. 2005;89:1103–1112.
29. Ramsey PW, Rillig MC, Feris KP, Holben WE, Gannon JE. Choice of methods for soil microbial community analysis: PLFA maximizes power compared to CLPP and PCR-based approaches. Pedobiologia. 2006;50:275–280.
30. Hallmann C, Schwark L, Grice K. Community dynamics of anaerobic bacteria in deep petroleum reservoirs. Nature Geosci. 2008;1:588–591.
31. Smithwick EAH, Turner MG, Metzger KL, Balser TC. Variation in NH4+ mineralization and microbial communities with stand age in lodgepole pine (Pinus contorta) forests, Yellowstone National Park (USA). Soil Biol Biochem. 2005;37:1546–1559.
32. Mentzer JL, Goodman RM, Balser TC. Microbial response over time to hydrologic and fertilization treatments in a simulated wet prairie. Plant Soil. 2006;284:85–100.
33. Bligh EG, Dyer WJ. A rapid method for total lipid extraction and purification. Can J Biochem Phys. 1959;37:911–917. [PubMed]
34. Midi I. Sherlock Microbial Identification System. Version 4.0. 2001.
35. Bååth E. The use of neutral lipid fatty acids to indicate the physiological conditions of soil fungi. Microb Ecol. 2003;45:373–383. [PubMed]
36. Colwell RK. EstimateS: Statistical estimation of species richness and shared species from samples. Version 8. 2006. User's Guide and application published at:
37. Hendrick D, Peacock A, White D. Margesin R, Schinner F, editors. Interpretation of Fatty Acid Profiles of Soil Microorganisms. 2005. pp. 251–259. Manual for Soil Analysis: Monitoring and Assessing Soil Bioremediation: Springer.
38. Zelles L. Fatty acid patterns of phospholipids and lipopolysaccharides in the characterisation of microbial communities in soil: a review. Biol Fert Soils. 1999;29:111–129.
39. SAS-Institute. Cary, N.C.: SAS Institute Inc; 2007. JMP. 7 ed.
40. Zar JH. New Jersey: Prentice Hall.; 1998. Biostatistical Analysis 4th Edition. p. 929 p.
41. Srinivasan BS, Caberoy NB, Suen G, Taylor RG, Shah R, et al. Functional genome annotation through phylogenomic mapping. Nat Biotechnol. 2005;23:691–698. [PubMed]
42. Tringe SG, von Mering C, Kobayashi A, Salamov AA, Chen K, et al. Comparative metagenomics of microbial communities. Science. 2005;308:554–557. [PubMed]
43. Suen G, Jakobsen JS, Goldman BS, Singer M, Garza AG, et al. Bacterial postgenomics: the promise and peril of systems biology. J Bacteriol. 2006;188:7999–8004. [PMC free article] [PubMed]
44. Felsenstein J. PHYLIP–Phylogeny Inference Package (Version 3.2). Cladistics. 1989;5:164–166.
45. Vestal JR, White DC. Lipid analysis in microbial ecology - quantitative approaches to the study of microbial communities. Bioscience. 1989;39:535–541. [PubMed]
46. Zelles L, Bai QY, Beck T, Beese F. Signature fatty-acids in phospholipids and lipopolysaccharides as indicators of microbial biomass and community structure in agricultural soils. Soil Biol Biochem. 1992;24:317–323.
47. Bardgett RD, Hobbs PJ, Frostegård A. Changes in the structure of soil microbial communities following reductions in the intensity of management of an upland grassland. Biol Fert Soils. 1996;22:261–264.
48. Bull ID, Parekh NR, Hall GH, Ineson P, Evershed RP. Detection and classification of atmospheric methane oxidizing bacteria in soil. Nature. 2000;405:175–178. [PubMed]
49. Fritze H, Pietikäinen J, Pennanen T. Distribution of microbial biomass and phospholipid fatty acids in Podzol profiles under coniferous forest. Eur J Soil Sci. 2000;51:565–573.
50. Waldrop MP, Balser TC, Firestone MK. Linking microbial community composition to function in a tropical soil. Soil Biol Biochem. 2000;32:1837–1846.
51. Keith-Roach MJ, Bryan ND, Bardgett RD, Livens FR. Seasonal changes in the microbial community of a salt marsh, measured by phospholipid fatty acid. Biogeochemistry. 2002;60:77–96.
52. Petersen SO, Frohne PS, Kennedy AC. Dynamics of a soil microbial community under spring wheat. Soil Sci Soc Am J. 2002;66:826–833.
53. Zak DR, Holmes WE, White DC, Peacock AD, Tilman D. Plant diversity, soil microbial communities, and ecosystem function: are there any links? Ecology. 2003;84:2042–2050.
54. Oravecz O, Elhottová D, Krištůfek V, Šustr V, Frouz J, et al. Application of ARDRA and PLFA analysis. Folia Microbiol. 2004;49:83–93. [PubMed]
55. Balser TC, Firestone MK. Linking microbial community composition and soil processes in a California annual grassland and mixed-conifer forest. Biogeochemistry. 2005;73:395–415.
56. Tringe SG, Rubin EM. Metagenomics: DNA sequencing of environmental samples. Nat Rev Genet. 2005;6:805–814. [PubMed]
57. Lambais MR, Crowley DE, Cury JC, Büll RC, Rodrigues RR. Bacterial diversity in tree canopies of the Atlantic forest. Science. 2006;312:1917. [PubMed]
58. Mikheyev AS, Mueller UG, Boomsma JJ. Population genetic signatures of diffuse co-evolution between leaf-cutting ants and their cultivar fungi. Mol Ecol. 2007;16:209–216. [PubMed]
59. Mikheyev AS, Vo T, Mueller UG. Phylogeography of post-Pleistocene population expansion in a fungus-gardening ant and its microbial mutualists. Mol Ecol. 2008;17:4480–4488. [PubMed]
60. Epstein E. 1997. p. 504 p. The Science of Composting: CRC Press.
61. Yu H, Zeng G, Huang H, Xi X, Wang R, et al. Microbial community succession and lignocellulose degradation during agricultural waste composting. Biodegradation. 2007;18:793–802. [PubMed]

Articles from PLoS ONE are provided here courtesy of Public Library of Science