The Evolution of Agriculture in Ants

See allHide authors and affiliations

Science  25 Sep 1998:
Vol. 281, Issue 5385, pp. 2034-2038
DOI: 10.1126/science.281.5385.2034


Cultivation of fungi for food by fungus-growing ants (Attini: Formicidae) originated about 50 million years ago. The subsequent evolutionary history of this agricultural symbiosis was inferred from phylogenetic and population-genetic patterns of 553 cultivars isolated from gardens of “primitive” fungus-growing ants. These patterns indicate that fungus-growing ants succeeded at domesticating multiple cultivars, that the ants are capable of switching to novel cultivars, that single ant species farm a diversity of cultivars, and that cultivars are shared occasionally between distantly related ant species, probably by lateral transfer between ant colonies.

Fungus farming by ants of the tribe Attini originated in the early Tertiary (1, 2) and thus predates human agriculture by about 50 million years (3). During its extensive evolutionary history, this symbiosis between “attine” ant farmers and their fungal cultivars has acquired an astonishing complexity, involving secretion of antibiotic “herbicides” to control weed molds and elaborate manuring regimes that maximize fungal harvests (4, 5). Of the over 200 known extant attine species, all are obligate fungus farmers. Cultivars are propagated vegetatively (as asexual clones) within nests, and between parent and offspring nests. In the few studied cases, the foundress queen carries in her mouth a pellet of fungus from the natal nest that she uses to start her own garden (2). This mode of propagation suggested the long-standing hypothesis that attine fungi are ancient clones that have strictly coevolved with their hosts (2).

Most attine cultivars are propagated as a mycelium (multicellular phase), but those cultivated by a group in the ant genusCyphomyrmex are maintained as masses of yeast (unicellular phase) (2). The great majority of these fungi, including the yeasts, are members of the tribe Leucocoprini (Basidiomycotina: Agaricales: Lepiotaceae) (6, 7), a large, poorly known group of predominantly tropical mushrooms (8). The cultivation of a non-lepiotaceous fungus by some ants in the genusApterostigma is the only exception to this ancestral association with lepiotaceous fungi and indicates a historically unique switch to a cultivar outside the Lepiotaceae (7).

Domestication of novel cultivars and switching between cultivars may be ancient themes in attine agriculture, which presumably began with an ancestral ant that facultatively associated with fungi (1, 2, 5). Indeed, repeated and possibly recent domestication events were suggested by phylogenetic patterns (7). First, some attine cultivars appeared to be more closely related to free-living Leucocoprini than to other cultivars; and second, ant and cultivar phylogenies were topologically incongruent. Both patterns are inconsistent with a scenario of a single domestication event followed by strict clonal propagation. Support for these conclusions was weak, however, and other processes can generate topological incongruence, including cultivar transfers between ant lineages.

To distinguish between these diverse scenarios of ant-fungus evolution, we surveyed Neotropical free-living and ant-cultivated fungi and sequenced representative lineages for phylogenetic analysis. The survey focused on the seven genera of “lower” [phylogenetically basal (9)] attine ants, those most likely to have retained the least modified forms of the ancestral farming behavior (10). We surveyed sympatric Panamanian communities of lower attine cultivars (n = 337) and free-living Leucocoprini collected as fruiting bodies (mushrooms) (n = 309 collections) (11), and an additional 216 cultivars along an axis intersecting the United States (n = 96), Costa Rica (n = 13), Trinidad (n = 65), Guyana (n= 39), and Brazil (n = 3) (12). Collections were screened for restriction fragment length polymorphisms (RFLPs) (13), and representative RFLP types (25 free-living and 57 ant-cultivated fungi) were sequenced for two nuclear ribosomal DNA (rDNA) gene regions [internal transcribed spacer (ITS) and large subunit (25S) (14)]. Phylogenetic analyses using parsimony and maximum likelihood criteria produced results identical in the features emphasized below (Fig. 1) (15).

Figure 1

Phylogeny for 57 fungal cultivars of the seven genera of “lower” fungus-growing ants and 36 free-living fungi in the family Lepiotaceae. All cultivars belong to a group grown only by “lower attines,” ant lineages that are phylogenetically “primitive” and thus most likely to have retained ancestral farming behaviors that existed early in the history of this 50-million-year-old symbiosis. Two free-living counterparts of ant-cultivated fungi (A and B) reveal two recent domestications of cultivars by ants. Statistically distinct cultivar lineages, represented by colored branches, provide evidence for a minimum of three additional domestications (18). Cultivated fungi are indicated in red by their respective ant host species. Free-living fungi are indicated in black; some of these are undescribed species and are denoted by their collection IDs. Almost all ant species cultivate several phylogenetically distant cultivars. The tree shown is the strict consensus of five equally parsimonious trees obtained by parsimony analysis and successive approximations weighting of a 1422–base pair sequence of two nuclear rDNA regions (ITS and 25S), and resembles in all important details the tree found using maximum likelihood (15). Numbers on branches are bootstrap values from 500 pseudoreplicates; where multiple bootstrap values are indicated, the second value is derived from analyses of two phylogenetically restricted subsets of taxa in which additional characters, excluded from the global analysis because of alignment ambiguities, could be unambiguously aligned. Asterisks (*) indicate branches not present in the strict consensus of 180 equally parsimonious trees found in the analysis before successive weighting; bullets (•) indicate branches above which a single representative taxon was included in the analyses (15).

DNA sequences of one field-collected, free-living leucoprinoid mushroom matched exactly (PA-234: 100% sequence identity) (Fig. 1B) and a second matched nearly exactly (PA-302: differing in one base pair) (Fig. 1A) the sequences of two separate, ant-cultivated fungi. This establishes the existence of free-living counterparts of two attine cultivars. During our year-long survey, four individuals of one counterpart (PA-302) were collected within a 2-week period, and the other counterpart (PA-234) was collected only once. This rarity suggests that free-living analogs may also exist for other ant cultivars, but were simply not encountered. The discovery of two free-living counterparts, each matching a cultivated fungus in a fast-evolving gene (ITS), suggests that these cultivars were domesticated recently. The alternative explanation, that the free-living counterparts represent “escapes” from a monolithic, clonal lineage that originated with a single domestication event 50 million years ago, is inconsistent with observed levels of allele sequence divergence (see below) and theory predicting genetic degradation of unexpressed fruiting (mushroom) structures under long-term asexuality (16). These objections to sexual forms arising from ancient clones do not, however, preclude repeated instances of domestication and escape that occur over time spans too short to permit degradation of underlying fruiting genes.

The majority of lower-attine cultivars fall into two monophyletic groups, but two unique cultivars (Mycocepurus smithi S60 andMyrmicocrypta infuscata G11) (Fig. 1) fall outside these clades. This pattern suggests four independent domestications and contradicts the hypothesis of a single domestication followed by long-term clonal propagation (2). Statistical tests in which cultivars were constrained to form less than four independent lineages resulted in highly significant support for a minimum of three independent domestications (17). Perhaps more compelling, the statistical tests conservatively excluded the two free-living counterparts of two cultivars (Fig. 1, A and B) that, as argued above, represent two additional cases of recent domestication. This increases to five the minimum number of domestications of fungi by ants (18).

Only one ant cultivar, Myco. smithi S60 (Fig. 1), confidently falls outside of the two main cultivar clades, suggesting that lower attine agriculture is somewhat specialized in that domestication of fungi from outside of the two clades occurs infrequently, or that such associations are relatively short-lived, or both. This may indicate that cultivars derived from the two main clades are more suitable for cultivation, nutritionally or otherwise, than other leucocoprinoid fungi.

Whereas any single attine nest contains only a single cultivar, different nests of the same ant species may contain distantly related cultivars (Fig. 1). In an extreme case, Myco. smithi was found to cultivate four distinct fungal lineages in central Panama and four additional lineages in Trinidad (Fig. 1); even adjacent nests of this ant, sometimes separated by only a few centimeters, were found to contain distantly related cultivars. At the other extreme, all 34 nests of Myrm. sp. 1 contained the same fungal lineage. Most species cultivate at least two distinct fungal lineages (19), sometimes drawn from both of the two main cultivar clades, such that most lower attine ants are generalists within or even across the two clades. The only exceptions are the yeast-cultivatingCyphomyrmex species, which specialize on a derived monophyletic group of cultivars (Fig. 1). Within this yeast group, however, the same ant species may cultivate a diverse set of fungi, a pattern consistent with both cultivar acquisition from free-living stocks [note the free-living PA-302 in the yeast clade (Fig. 1)] and with cultivar exchange between yeast-cultivating species (see below). Thus, although the yeast-cultivating Cyphomyrmex species are specialized on a narrowly defined fungal clade, there is no apparent specialization within that clade.

The same or very closely related fungi may be cultivated by distantly related ants. For example, in central Panama, Myrmicocrypta ednaella, Mycocepurus tardus, and Myco. smithi cultivate fungi with identical rDNA sequences; similarly, in Guyana, Cyphomyrmex faunulus and Myrmicocrypta cf. buenzlii share identical cultivars. Overall, cultivar sharing was observed seven times between two species from the same genus and four times between two or more species of different genera (Fig. 1). Such sharing obviously generates topological incongruence across ant and fungal phylogenies and indicates that different ants acquired cultivars from the same free-living stock, or that they acquired cultivars from each other, or both.

To determine whether some instances of this cultivar sharing were due specifically to cultivar transfers between ant species, we generated amplified fragment length polymorphism (AFLP) fingerprints for cultivars that were identical in rDNA sequence and that were isolated from gardens of different ant species (20). Fingerprints revealed that some cultivars grown by different ant species are genetically identical and that they are asexual descendants of the same clone (Fig. 2). Cases of genetically identical, shared cultivars were detected in two species ofCyphomyrmex (minutus and rimosus from Florida), in two species of Mycocepurus (smithiand tardus from Panama), and in two species in widely separated genera, Cyphomyrmex longiscapus andApterostigma auriculatum [from Panama (Fig. 2)]. Within each species pair, nests were collected over areas spanning hundreds of square kilometers (Fig. 2). Given this geographic scale, it is implausible that the ants acquired their cultivars from the same free-living fungal clone (21); instead, cultivars appear to be occasionally transferred across ant species and subsequently disperse with the ants. Such lateral transfer may occur fortuitously (for example, gardens from several nests are simultaneously disturbed and mixed) or after accidental cultivar loss that forces ants to obtain a replacement from a neighboring colony. The case of shared cultivars in C. minutus and C. rimosus from Florida is particularly instructive, because C. rimosus was introduced to Florida apparently during this century (22). The fact that C. rimosus presently shares a genetically identical cultivar with the indigenous C. minutus indicates that cultivar replacement may occur on an ecological time scale.

Figure 2

AFLP fingerprints of three cultivar clones shared between different ant species. These three cultivar clones were found, respectively, in gardens of three species pairs:Cyphomyrmex longiscapusApterostigma auriculatum, C. rimosusC. minutus, andMycocepurus tardusMyco. smithi (20). Three cultivars are shown for each clone: two isolated from geographically proximate nests of different species (shown as the two leftmost in each triplet; with distances of 9 km, 100 m, and 11 km, respectively, for the three species pairs), one from a distant nest of one of the two species (85, 110, and 360 km, respectively). AFLP fingerprints are shown for 4 of 12 different primer systems tested (20). Ten systems (only two are shown in the two top panels) failed to reveal any genetic variation within each cultivar clone, and only two (two bottom panels) revealed minor band differences (circles). The largest detected difference (2 of more than 100 bands total) involved two isolates from the geographically most distant nests (360 km; Panama–Costa Rica). Identity of banding patterns (amplifying more than 100 AFLP fragments that are dispersed randomly throughout the genome) indicates that the respective cultivars are genetically identical and therefore are asexual descendants of the same clone. The genetic identity of cultivars found in gardens of different ant species or genera, some of them separated by large geographic distances, suggests that these cultivars were transferred between ant nests and were subsequently dispersed by the ants to distant locations. Panama: Pan 1 = Pipeline Road, Pan 2 = Gamboa, Pan 3 = Nusagandi; Florida: Flor 1 = Archbold Biological Station, Flor 2 = Orlando; CR = Costa Rica.

Two lines of evidence support the clonal nature of attine cultivars: the existence of widespread AFLP-fingerprint cultivar types (Fig. 2) (23), and behavioral observations of clonal cultivar transfer by foundress queens from parent to daughter nests (2, 5). The absence of significant allele sequence divergence (ASD) (24) at the two sequenced regions indicates that these clones are not ancient. Specifically, ASD levels across the two rDNA genes are almost identical between sequenced cultivars and free-living Leucocoprini [0.00146 ± 0.00148 and 0.00109 ± 0.00163 heterozygosity per site (mean ± SD), respectively], as is expected if cultivars were acquired recently from sexually outcrossing stocks (24). Observed rDNA ASD levels therefore are inconsistent with the long-standing model of ancient cultivar clonality spanning millions of years (2).

Cultivar collections did not reveal phylogeographic patterns of cultivar usage along the transect Costa Rica–Panama–Trinidad–Guyana (Fig. 1). Cultivars are diverse at each location and largely shared across the entire transect. Thus cultivar lineages do not cluster by location. This suggests that ants disperse widely and mix cultivar lineages across locations, or that the free-living stocks from which cultivars are drawn are widespread and have overlapping ranges, or both.

In summary, phylogenetic patterns demonstrate that ants, like humans, succeeded at domesticating multiple cultivars during the history of their agricultural symbiosis, that the acquisition of novel cultivars is an ongoing process occurring in at least some extant ant species, and that cultivars are shared occasionally between ant lineages, probably by transfer between ant species. Domestication of free-living stocks therefore continued after the origin of fungus-growing and, along with cultivar exchange between ant species, may have persisted for 50 million years of ant evolution as a means to replace cultivars after accidental or pathogen-driven loss of gardens, to respond to environmental changes by acquiring cultivars with novel features, or to capitalize on strains that were improved while associated with other ant lineages. If exchanges occur frequently among ants, as population-genetic patterns suggest, the history of ant agriculture may have been characterized by the same rapid lateral spread of cultivars that has shaped the history of human agriculture (3). Given these parallels between ant and human agriculture, promising avenues for future attine research include the testing for ecological zoning (correlating cultivar types with ecological conditions) and, given the recent and repeated domestications of novel cultivars, testing for the gradual modification of cultivars by ants through a sensory bias–driven analog of artificial selection.

  • * To whom correspondence should be addressed at the University of Maryland (e-mail: um3{at} or the Smithsonian Institution (e-mail: schultz{at}


View Abstract

Navigate This Article