Seed odor mediates an obligate ant-plant mutualism in Amazonian rainforests

Abstract

Seed dispersal mutualisms are essential for the survival of diverse plant species and communities worldwide. Among invertebrates, only ants have a major role in seed dispersal, and thousands of plant species produce seeds specialized for ant dispersal in "diffuse" multispecies interactions. An outstanding but poorly understood ant-seed mutualism occurs in the Amazonian rainforest, where arboreal ants collect seeds of several epiphyte species and cultivate them in nutrient-rich nests, forming abundant and conspicuous hanging gardens known as ant-gardens (AGs). AG ants and plants are dominant members of lowland Amazonian ecosystems, and their interaction is both specific and obligate, but the means by which ants locate, recognize, and accept their mutualist seeds while rejecting other seeds is unknown. Here we address the chemical and behavioral basis of the AG interaction. We show that workers of the AG ant Camponotus femoratus are attracted to odorants emanating from seeds of the AG plant Peperomia macrostachya, and that chemical cues also elicit seed-carrying behavior. We identify five compounds from P. macrostachya seeds that, as a blend, attract C. femoratus workers. This report of attractive odorants from ant-dispersed seeds illustrates the intimacy and complexity of the AG mutualism and begins to illuminate the chemical basis of this important and enigmatic interaction.

Fig. 1.

Arboreal AG involving the ants Camponotus femoratus and Crematogaster cf. limata parabiotica and the epiphytic plants Anthurium gracile and Peperomia macrostachya. Gardens are established when ants collect seeds of their mutualist epiphytes and embed them in the nest walls, where they grow.

Fig. 2.

Response of C. femoratus ants to different doses of P. macrostachya extract in seed-carrying assays. Hexane extracts of P. macrostachya seeds were applied to Piper laevigatum seeds, which ants typically ignore. Extract-treated seeds were presented within 5 cm of foraging trails of C. femoratus ants. n indicates the number of 20-min trials with six ant colonies.

Fig. 3.

Behavioral tests of olfactory attractants from AG seeds. (a) Two-choice glass Y-olfactometer used to assay attraction in the field. The apparatus was placed near foraging trails of C. femoratus; ants entered one arm of the olfactometer and chose between two odorant sources by turning right or left in air flow generated by an air pump (top). (b) Ant responses (mean ± SEM) to odorant choices presented in the Y-tube olfactometer. Trials were conducted in pairs in which the orientation of the odorant sources was reversed to control for spatial effects. A single trial consisted of 30 decisions by different ants, and n indicates the number of trials conducted. Asterisks indicate departures from the null hypothesis of no preference, represented by the dashed line (one-sample t tests, P ≤ 0.001).

Fig. 4.

Representative GC-EAD trace of Camponotus antenna response to a 0.25 seed equivalent of the behaviorally active fraction of P. macrostachya extract, separated on a nonpolar column. FID, flame ionization detector. Highlighted GC peaks were active in 19 analyses on polar and nonpolar columns at doses as low as 0.04 seed equivalent. The active compounds are as follows: 1, 3,5-dimethoxytoluene; 2, 6-MMS; 3, methyl o-anisate; 4, methyl 3,5-dimethoxybenzoate; 5, unknown sesquiterpene alcohol; 6, (2E,4Z)-2,4-tetradecadienal; 7, (2E,4E)-2,4-tetradecadienal; 8, coeluting compounds whose activity could not be consistently correlated to a specific compound in analyses on different columns; 9, geranyl linalool.

Fig. 5.

Behavioral confirmation that a blend of five electrophysiologically active compounds from AG seeds attracts C. femoratus ants. Filter papers were treated with a synthetic five-component blend at 1× concentration (60 seed equivalents, to mimic the concentration emitted by a fallen P. macrostachya seed spike) or 10× concentration (600 seed equivalents) and tested in the field in a two-choice Y-olfactometer against either solvent-treated (blank) filter papers or only geranyl linalool, the major component in the active fraction (compound 9 in Fig. 3). P < 0.05 for 1× concentration, and P < 0.01 for 10× concentration vs. blank and vs. geranyl linalool alone.