Introduction

Animals use a variety of signaling modalities (e.g., visual, mechanical, chemical) to communicate vital information to nestmates, conspecifics, and heterospecifics (Endler 1992; Wyatt 2014). Chemical signaling is widespread in nature, in part, because pheromones and other chemical messages often are relatively inexpensive to produce, and can provide important contextual physiological and ecological information over long distances (Bossert and Wilson 1963; Symonds and Elgar 2008; Wyatt 2014). For insects, and especially social insects such as ants (Hymenoptera: Formicidae), chemical signaling is a ubiquitous part of life. Ants use a variety of compounds to organize individual behaviors and maintain colony structure (Hölldobler and Wilson 1990; Roitberg and Isman 1992). Ants also leverage volatile chemical signals as alarm pheromones in part because they are detectable over longer distances than are non-volatile chemicals (Blum 1969; Wyatt 2014; Leonhardt et al. 2016).

Alarm pheromones in some ant species occur as relatively conspicuous plumes of volatile compounds (Blum 1969; Attygalle and Morgan 1984; Keeling et al. 2004). These volatile chemical signals can be problematic because the strength and distribution of the message cannot be controlled once it is emitted (Bossert and Wilson 1963; Wyatt 2014). Such signals become freely available "public" information that is a reliable cue for exploitation by unintended receivers (i.e., eavesdroppers; Peake 2005) including ant-associated eavesdroppers, especially parasitoids and predators (Feener et al. 1996; Mathis et al. 2011; Cárdenas et al. 2012; reviewed by Adams et al. 2020). Indeed, chemical eavesdropping is common in diverse groups of animals and is well-studied in insects (Stowe et al. 1995; Wyatt 2014). However, to our knowledge, no studies have explored the possibility that co-occurring ants eavesdrop on heterospecific ant alarm pheromones.

Interspecific eavesdropping on alarm pheromones is likely to be advantageous for co-occurring competitive or antagonistic species. Given that ant workers are valuable to their colonies directly as biomass (Wilson 1968) and indirectly via foraging and defensive behaviors (Carroll and Janzen 1973), selection should favor eavesdropping on any warning signals that consistently prevent worker loss by inducing avoidance behaviors in submissive species. Eavesdropping on heterospecific alarm pheromones could provide such a mechanism because alarm pheromones are reliable indicators of nearby dangers, including negative interspecific interactions among competing ants, the presence of predators and parasitoids, or any generalized threat or disturbance (Blum 1969).

Azteca chartifex/trigona (Dolichoderinae) is a common Neotropical canopy ant species complex that has conspicuous nests (Fig. S1), large polydomous colonies, distinct alarm pheromones, and aggressive workers (Wheeler 1986; Adams 1994; Longino 2007). Hereafter, we refer to this species complex as A. trigona for simplicity. In some forests, A. trigona influences local ant community structure via territorial behaviors creating a mosaic of species distributions (Adams 1994). Alarm pheromones coordinate A. trigona defensive behaviors (Adams 1994), but the effects of A. trigona alarm pheromones on other ant species are unknown.

Chemical components of alarm pheromones frequently are conserved within genera or subfamilies (Blum 1969; Wheeler et al. 1975; Norman et al. 2017), thus eavesdropping species are likely to be closely related to the emitting species. Additionally, some behaviorally dominant ants competitively exclude other ant species that are ecologically or morphologically similar (Hölldobler and Wilson 1990; Andersen and Patel 1994; Adams 2016). If ants respond to the alarm pheromones of closely related heterospecific species, and if A. trigona similarly influences local species assemblages via competitive exclusion, we expect eavesdropping on A. trigona pheromones to be best developed in phylogenetically and behaviorally similar subordinate or codominant species. Our personal observations in the forests of Panama (e.g., Yanoviak and Kaspari 2000, Adams et al. 2017) indicate that most canopy ants avoid direct interactions with A. trigona workers while foraging. We also frequently observe strong negative behavioral responses (i.e., fleeing) by some arboreal species when presented with forceps contaminated with A. trigona alarm pheromones. These observations suggest that chemical cue recognition and eavesdropping on A. trigona are common among canopy ants.

The principal objective of this study was to determine if tropical canopy ants eavesdrop on the alarm pheromones of A. trigona ants. We asked if canopy ants change their behaviors when exposed to A. trigona alarm pheromones, how their behaviors change, and whether behavioral responses are more common in certain subfamilies. We predicted that differences in behavioral responses would be associated with ant subfamily identity, and species that are more phylogenetically similar to A. trigona would exhibit the greatest frequency of responses. We experimentally tested these predictions in Panama with freshly captured worker ants of 16 species.

Methods

This study was conducted on Barro Colorado Island (BCI) in Panama (9.15 °N, 79.85 °W) between July 2016 and October 2018. BCI is a seasonally moist lowland forest with a wet season spanning May to December. All data for this study were collected during wet season months. More details about the site are provided elsewhere (Croat 1978; Leigh et al. 1996).

All worker ants used in experiments were collected by hand or with forceps from tree trunks and branches on BCI. Workers of a given species were collected from multiple colonies, and ants collected from a given colony (usually < 5 workers) were housed together in a vial until the start of an experiment. All ants were used in experiments within 48 h of collection and were provided with water and honey during the holding period. The 16 focal ant species (Table 1) were chosen to maximize phylogenetic diversity and to include species that commonly co-occur in trees with Azteca trigona at the study site (Adams et al. 2017).

Table 1 Summary of tests conducted on workers of 16 recipient ant species exposed to Azteca trigona alarm pheromones and ambient air (Control)
Full size table

Alarm pheromone trials

To determine which ant species responded to A. trigona alarm pheromones and how they responded, we placed a single ant (hereafter recipient ant) in a small glass arena (25.4 × 76.2 × 76.2 mm; Figs. S2, S3) and allowed it to acclimate for 3–5 min before each trial. At the beginning of each trial, A. trigona volatile odors (hereafter alarm pheromones) were obtained by aggravating ca. 30 workers inside a plastic vial (i.e., by continuously shaking the vial for 5 s) and then drawing air from the vial into a 20 mL polypropylene syringe (Norm-Ject, Henke Sass Wolf GmbH, Germany). Twelve milliliters of air containing A. trigona alarm pheromones were then injected into the chamber via the syringe. Control treatments followed the same protocol using a new conspecific recipient ant worker for each trial and a clean syringe filled only with ambient air. All trials were conducted at temperatures ranging 26–30 °C.

We recorded recipient ant behavior for multiple individual workers of 16 different species (Table 1). In each case, ant behavior was recorded 5 s before and 5 s immediately after exposure to A. trigona alarm pheromones using the video function of a compact digital camera (Canon-PowerShot ELPH 180, Canon Inc., Japan). We observed the videos blindly (i.e., odor treatments were unknown to the viewer) and noted the following conspicuous changes in behavior of recipient ants: increased or decreased running speed, increased or decreased frequency of antennal movement, mandible flaring, and gaster flagging/tucking (i.e., positioning the gaster approximately orthogonal to the body axis either dorsally or ventrally; Curtis 1985; Obin and van der Meer 1985; Fig. 1). Changes in behavior were visually estimated. If individuals exhibited two or more of the behavior changes listed above, we recorded a behavioral change in response to the odor source. We similarly quantified the behavioral responses of recipient ants in control treatments.

Fig. 1
figure 1

The stereotypical ant responses to odors observed in this study: running (a), antennating (b), mandible flaring (c), and gaster flagging or tucking (d, e)

Full size image

Analyses

Analyses were performed in the R statistical environment (R Core Team 2019). We used Fisher’s Exact tests to determine if the frequency of behavioral responses to A. trigona alarm pheromones differed from controls. A separate test was conducted for each species. We used a mixed-effect generalized linear model to compare subfamily responses to A. trigona and control odor treatments (glmer, package lme4). Responses of A. trigona workers to A. trigona and control odors were excluded from this analysis. The response term was a binomial variable (behavioral response or no response), subfamily and odor treatment were fixed effects, and species was a random effect. Stepwise model reduction with likelihood ratio tests removed the non-significant interaction between subfamily and odor source. To compare subfamily responses to A. trigona alarm pheromones and control odors separately, we used a mixed-effect generalized linear model comparing subfamily responses using only recipient ants exposed to A. trigona alarm pheromones and a separate model with only recipient ants exposed to control odors. The response term was again a binomial variable, subfamily was the only fixed effect, and species was a random effect. Subfamily-level differences in behavioral responses were determined with post-hoc Tukey HSD tests.

Results

Seven of the 16 focal ant species responded more frequently to A. trigona odors than to ambient air (Fig. 2). Those seven species represented four distinct subfamilies; three were dolichoderines. Ants in all five focal subfamilies showed at least some behavioral response to A. trigona pheromones. Specifically, when worker responses were pooled at the subfamily level, response frequencies for A. trigona odors were consistently higher than for controls (χ2 = 68.82, df = 1, p < 0.001) and differed among subfamilies (χ2 = 13.61, df = 4, p = 0.009). However, only dolichoderines responded to A. trigona pheromones at a higher frequency than myrmicines and pseudomyrmecines (Table S1, Fig. S4). Response frequencies to the ambient air control also differed among subfamilies (χ2 = 15.53, df = 4, p = 0.004), with dolichoderines and formicines responding to control odors at a higher frequency than myrmicines and pseudomyrmecines (Table S1, Fig. S4).

Fig. 2
figure 2

The proportion (with ± 95% Clopper-Pearson CI) of trials in which workers of 16 recipient ant species exhibited behavioral responses when exposed to the alarm odors of Azteca trigona (black bars) and ambient air controls (gray bars). Ecta. = Ectatomminae. Significant differences between responses to A. trigona alarm pheromones and control ambient air are indicated as *p < 0.05, **p < 0.01, and ***p < 0.001

Full size image

In all, 334 individual ants (out of 833) changed their behaviors when exposed to pheromones or ambient air. Half (53%) of the 444 ants that were exposed to A. trigona pheromones showed a behavioral response, whereas only 25% of the 389 ants exposed to control odors responded. Most (75%) of the responding ants exhibited a combination of altered running speed and increased antennating frequency (Table 2, A + R). The least common behavioral change was gaster flagging, although dolichoderines exhibited combined behaviors that included gaster flagging in 26% of responses (Table S2).

Table 2 The different combinations of qualitative behavioral responses (Behavior) observed in this study
Full size table

Discussion

Here we show that multiple Neotropical canopy ant species across a broad distribution of subfamilies change their behaviors when exposed to Azteca trigona alarm pheromones. Four of the five subfamilies tested included species that were both responsive and non-responsive to A. trigona pheromones (except for Ectatomminae, in which only one species was tested), suggesting that eavesdropping is a selective phenomenon among coexisting ant species that is not based on phylogenetic relatedness. Eavesdropping species occur across all domains of life (Stowe et al. 1995; Joint et al. 2007; Frost et al. 2008) and eavesdropping specifically on heterospecific alarm signals in animals is likely to be advantageous in many circumstances (Stowe et al. 1995; Adams et al. 2020). In this study, aggressive responses to A. trigona odors were uncommon (77% of responders exhibited what appeared to be flight behaviors—more frequent antennation and faster running speeds—which was similar to the 72% response frequency to control air including these behaviors). This suggests that avoidance of a potentially threatening species is the basis for eavesdropping behaviors among the focal ants. Such an "ecology of fear" occurs among many animal taxa (Pfeiffer 1962; Apfelbach et al. 2005; Goodale and Nieh 2012).

The results of this study do not support the prediction that species responding to A. trigona alarm pheromones would be closely related (i.e., dolichoderines). The major components of alarm pheromones that elicit worker responses are often specific within ant subfamilies or genera (Blum 1969; Norman et al. 2017). For example, among ants, cyclopentanoid monoterpenes and sulcatone (6-methyl-5-hepten-2-one) apparently occur only in Azteca alarm pheromones (Wheeler et al. 1975; McCann et al. 2013). These compounds elicit sustained alarm responses in Azteca workers (Blum 1969; Wheeler et al. 1975; McCann et al. 2013). Ants often respond to pheromones only with a specific compound ratio (Blum 1969; Pokorny et al. 2020), suggesting that species that do not respond more to A. trigona odors either do not detect the odors of A. trigona, or ignore them. Additional research is needed to isolate the specific compound or compounds within the A. trigona alarm pheromone that elicit behavioral responses from heterospecifics.

Given the lack of strong phylogenetic signal, it is likely that the observed ant responses to A. trigona alarm pheromones reflect ecological pressures. It is potentially advantageous for foraging workers (regardless of their identity) to recognize the presence of potential heterospecific competitors, or to detect the existence of a nearby disturbance that is provoking an alarm response by one or more species. Indeed, in a related pilot study, we found that a subset of the species used in this study also responded to alarm pheromones from ants other than A. trigona, although with different frequencies. Thus, we hypothesize that the species responding to A. trigona alarm pheromones are those that are more likely to have negative interactions with A. trigona workers (i.e., competing species). For example, A. trigona tend to have non-overlapping foraging territories with the responding species Atta colombica and Ectatomma tuberculatum (Jutsum et al. 1981; Armbrecht et al. 2001). We also cannot exclude the possibility that co-occurring species eavesdrop on other A. trigona cues, such as trail pheromones or chemical footprints, similar to the trail parasites Cephalotes maculatus and Camponotus beebei (Wilson 1965; Adams 1990; Wüst and Menzel 2017). Ultimately, additional field-based studies and natural history observations are needed to clarify links between pheromone eavesdropping and foraging decisions among potentially competing ants (Adams et al. 2020).

The results of this study are consistent with observations of eavesdropping in a variety of non-ant systems (Pfeiffer 1962; Goodale and Nieh 2012). However, the specific compounds that elicit responses, and the ecological consequences of responding to A. trigona alarm pheromones (e.g., potential loss of access to food resources) remain unknown. Understanding such patterns will clarify the role of chemical eavesdropping on species interactions, foraging behavior, and community structure in arboreal ants.