Ant Foraging Behavior

In May 2011 we located 9–11 foraging trails each for five of the six ant species (Table 1), always <30 m from the nest for Atta and <5 m from the nest for Acromyrmex. The sixth species, Acromyrmex volcanus, was so rare that only one trail was found. We observed Acromyrmex trails for 15 to 30 min and Atta trails for 2 min (or 4 times 0.5 min when trails were very busy) to obtain comparable data when counting ants that passed an imaginary line perpendicular to the trail. We replicated observations by sampling either trails of different colonies or multiple trails of the same colony going in different directions so they could be considered as independent samples of foraging habitat (Table S1). Diversity of forage material was classified in six categories: (parts of) flowers, (pieces of) fruit, herbaceous leaves, tree-leaves, other material (always rare), and ants carrying nothing on their way back to the colony. When in doubt, we verified the origin of forage particles by backtracking the trail to the source. Observations were repeated across parts of the day (morning 9 AM–12 PM, afternoon 12 PM–5 PM, evening in the dark 10 PM–11 PM) and compared statistically to see whether this made any difference. The Smithsonian Tropical Research Institute (STRI), Panama, and the Autoridad Nacional del Ambiente y el Mar (ANAM) provided research permits to sample ant colonies, logistic help and facilities to work in Gamboa.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Differences in foraging rate between loaded and unloaded foragers.

https://doi.org/10.1371/journal.pone.0094284.t001

Based on earlier field surveys at this mosaic landscape of secondary growth forest and suburban areas for a period of two decades the following generalizations of habitat differentiation [35] appear to apply in Gamboa: Atta cephalotes and Acromyrmex volcanus are forest canopy foragers, whereas Acromyrmex octospinosus forages on the forest-floor. Atta colombica occurs both in the forest (usually at lower elevations) and in moist open grassland habitats, while Atta sexdens and Acromyrmex echinatior prefer open and sunlit nesting habitats for foraging. The latter two species extend their distributions towards the Pacific coast where annual rainfall is less than in Gamboa and natural habitat resembles savannas rather than a mosaic of forest patches [12], [28], matching their preference for open habitat in Gamboa. We thus had some a-priori qualitative notions to work with and designed the present study to quantify them.

To assess rates of foraging, we expressed our records in numbers of workers on trails per hour, counting both workers carrying material back to the nest and those without. Data were log-transformed to approximately equalize variances and analyzed with R [36], using Linear Mixed-Effects Models (“lme”) [37] and generating p-values with General Linear Hypotheses (“glht”) in the package “multcomp” [38]. Proportional distributions of forage types (tree-leaves, herbaceous leaves, flowers, fruit, other) were analyzed with the same tests, with separate trails being considered as a random factor in both analyses. To test for heterogeneity across trails within species we created a sub dataset consisting of four Atta trails from two colonies (one Atta colombica and one Atta cephalotes) and compared the foraging category scores between trails and species (with trails nested within species, and 0.5 min replicate observations for each trail). This was only possible for these Atta species, as we did not have replicate samples within single trails for Acromyrmex (Table S1).

Results for the proportional distribution of forage types were visualized using “heatmap.2” in the R package “gplots” [39]. Dendrograms were calculated with the “pvclust” package [40] using 1000000 bootstrap iterations. Final clustering plots were based on the overall similarities in mean proportions (p) between species and supplemented by estimates of the inverse Simpson Diversity index (D = 1/[Σp_i²]) to allow an explicit analysis of the degree of evenness (high D) between the different forage or expressed enzyme categories across species and genera of leaf-cutting ants. The denominator of the index decreases when more categories (p) enter the equation, but when the number of categories is constant (as in our analyses) more even distributions will give lower sums in the denominator and thus higher values of D. For the purpose of our study, D therefore functions as an index of generalist foraging or equal enzyme expression, so that high values (low Σp_i²) indicate that all categories are important and low values (high Σp_i²) indicate specialization either on a subset of forage categories or on a subset of expressed enzymes that were most active.

AZCL Enzyme Activity Assays

For each of the six ant species the garden enzyme profiles were analyzed for five different colonies, with the exception of Acromyrmex volcanus for which only one colony was available, but where we could add data for another colony obtained in the previous year by H. H. De Fine Licht (pers. com.). For each colony, fungus gardens were dug up, and about equal size fragments (ca 80 mg) from top, middle and bottom layers of fungus gardens were collected and immediately homogenized together to obtain representative average enzyme activity measures per colony. These measurements were performed using previously published methods, which are easily applicable in the field and give repeatable results [21], [34], [41]. In short, fungus garden material (ca. 240 mg) was crushed with a pestle in a 1.5 ml eppendorf-tube containing 1000 μl 0.05 M TRIS-HCl buffer (pH 7.0), vortexed immediately and then centrifuged for 15 minutes (15000 g) after which the supernatant was removed and applied immediately to each of 12 different assay-plates containing 0.1 g/L of the Azurine-Crosslinked (AZCL) substrates: amylose, arabinoxylan, barley β-glucan, casein, collagen, debranched arabinan, galactan, galactomannan, HE-cellulose, rhamnogalacturonan, xylan and xyloglucan that were chosen because they yielded positive enzyme activities in an earlier study [21].

The assay-plates of 6 cm diameter were prepared separately for each substrate using an agarose medium (1% agarose, 23 mM phosphoric acid, 23 mM acetic acid, 23 mM boric acid), and pH adjusted according to the manufacturer’s description (Megazyme, Bray, Ireland). After the medium had solidified, round wells (area of ca. 0.1 cm²) were made in each plate with a cut-off pipette tip and 12 μl of the supernatant was applied to each well in triplicate. After 22 hours of incubation at 25°C the plates were photographed and the area of the blue halo surrounding each well (a quantitative measure for the absolute amount of enzyme activity [21], [34]) was measured using the software program ImageJ ver. 1.43u for Macintosh. Enzyme activity measurements were grouped into categories based on which plant cell wall component is the main target of the enzymes ([42] and Megazyme, Bray, Ireland): amylases (measured with amylose), cellulases (measured with barley β-glucan and HE-cellulose), hemicellulases (measured with arabinoxylan, galactomannan, xylan and xyloglucan), pectinases (measured with debranched arabinan, galactan and rhamnogalacturonan), and proteases (measured with casein and collagen). We present these data grouped for five categories of enzymes, implying that each of these categories had data from 2–3 enzymes, except for amylases that were represented by only a single enzyme amylose.

The enzyme activity scores were analyzed using a General Linear Model in SAS, with “colony” nested within “species” and “species” nested within “genus”. Colony was then treated as having a random enzyme class activity, composed of specific enzyme activities nested within enzyme activity classes. This procedure implied that we had to omit amylase, because we only had a single substrate (amylose = starch) for testing activity and because starch is not a primary challenge in the degradation of plant material [34]. We also left out rare Acromyrmex volcanus and thus report original mean values for enzyme activity for all six species and ten enzyme classes, whereas statistics given in the text refer to the reduced data set of four enzyme classes and the gardens of five ant species. In our final analyses we combined the enzyme and foraging datasets and visualized patterns of association with the plot.PCA function after Principal Component Analysis (“PCA”) using the “FactoMineR” package [43].

Genus-level Niche Segregation between Atta and Acromyrmex

Our results confirm that the two genera of leaf-cutting ants operate at different scales and show that their foraging niches are systematically different and that the enzymatic processing activities of fungus gardens appear to reflect these differences. The differences in foraging preferences quantified our intuitive expectations based on two decades of fieldwork in Gamboa, but the enzymatic activity differences were more substantial than we expected, because the two genera rear fungus-garden symbionts that belong to a single species L. gongylophorus [33]. This suggests that studies of phenotypic plasticity in enzyme gene expression will be worthwhile to enhance our understanding of the versatility of the leaf-cutting ant symbiosis. We will return to this in more detail below.

Our finding that average genus-specific foraging rates show a 42 fold difference in loaded-worker return rates and a 63 fold difference in total worker return rates per hour, seems to match the ca. two order of magnitude difference in colony size between Atta and Acromyrmex. The fact that the differences do not quite reach 100 fold [14], [18], [26], [46] may be due to our primary focus on the largest Acromyrmex colonies (smaller colonies have too little foraging activity making the type of sampling that we did less feasible), whereas our selection of Atta colonies mostly contained medium size colonies. It is also conceivable that the Atta workers that returned to their nests without carrying plant material may have had their crops filled with plant sap as suggested by Littledyke & Cherrett [47], Quinlan & Cherret [48] and Hölldobler & Wilson [19], but the present setup did not allow any measurements on this. This suggests that considering only loaded workers may underestimate foraging effort, and that larger scale comparative studies should include sampling of liquid food in the crops of returning foragers. In spite of these limitations, we will also return to tentative inferences on species- and genus-level niche segregation that our snapshot data for Gamboa appeared to allow.

Garden Enzyme Activity and Forage Material – is there a Connection?

It has long been known that the fungus is a major producer of enzymes for the decomposition of plant material that leaf-cutting ant foragers provide, and recent work has shown that these decomposition services are supplemented by several other microorganisms that live in attine gardens [49], [50]. Other recent studies have emphasized that the expression of enzymes can be remarkably plastic and substrate dependent [21], [34]. This is consistent with earlier notions that there are active feedback loops between forager supply and symbiont demand, such that foragers may discard some forage material under specific conditions where its excess processing would not be optimal [51].

Our present results quantify the notion that Atta and Acromyrmex represent ecologically distinct ant genera, both with regard to forage acquisition/diversity and garden enzyme activity/diversity. We acknowledge that this may not necessarily apply in other Latin American regions and that our results thus make no predictions about the extent to which, for example, grass-cutting Atta and Acromyrmex should partition their foraging habitats on the Argentinean pampas. However, the sampling schemes and analyses reported here can also be applied in other habitats, so that any hypothesis suggesting that similar niche partitioning rules could apply also there can be tested and if need be rejected. The genus-level distinction in garden enzyme activity/diversity appears consistent with recent other results showing that proteomes differ between sympatric fungus gardens of Acromyrmex echinatior and Atta cephalotes [52]. Such functional differences between Atta and Acromyrmex cultivars could possibly be due to species and/or genera rearing different lineages of the same fungal species, consistent with Atta and Acromyrmex in Gamboa rearing non-overlapping subclades when using rapidly evolving microsatellite or AFLP markers to characterize them [53].

It is remarkable that our results indicate that Atta gardens generally produced lower amounts of enzymes, even though these ants forage mostly on tree-leaves (Figure 3), which one would expect to be more demanding to decompose. It also appeared that Atta gardens tended to overproduce two classes of enzymes, cellulases and pectinases, in addition to amylases (Figure 2), whereas Acromyrmex gardens produced higher amounts of all enzyme categories. This suggests that Atta gardens may somehow extract necessary nutrients more efficiently, but further work will be needed to understand the details of these processes. An additional factor to consider in this context is that Atta colonies produce conspicuous waste heaps or underground compost chambers, whereas this is rare for Acromyrmex (J.J. Boomsma & P.W. Kooij pers. obs.). This is consistent with Panamanian Atta discarding a larger fraction of not fully degraded older fungus garden biomass than Acromyrmex [54], [55], perhaps because average enzyme activity per unit of fungus garden mass is lower and fresh tree leaves are more abundantly available than flower parts.

An earlier comparative study [21] has hypothesized that Atta species focus on the rapid degradation of starch and proteins, but discard fungus garden material before most of the cellulose and hemicellulose is degraded. This is consistent with other recent studies showing that high amounts of cellulose and hemicellulose are still present in the bottom layer of fungus gardens [50], [56], [57] and that only cellulases from L. gongylophorus remain highly active in this bottom layer [52], [58]. The larger scale and more wasteful substrate processing practiced by Atta may thus leave more substantial niches for additional bacterial and/or yeast [49] decomposition, similar to the domestication of specialized gut bacteria in large ungulates [59] that rely on residues of leaf-material that were hard to digest even for ruminants. Focused comparative transcriptomics to investigate conditional gene expression in fungus gardens of the two leaf-cutting ant genera could shed further light on possible differences of this kind and metagenome sequencing could identify the microbial communities involved, similar to an earlier yeast study on the fungus gardens of Acromyrmex and Atta [49].

Niche Partitioning in Panamanian Atta and Acromyrmex

The data provided in our study are a snapshot of year-round foraging, which is known to vary across the seasons [25]. This implies that we cannot be sure that sampling in other seasons or at other sites would have yielded similar results. However, we note that the five forage categories that we distinguished are very general and likely to be available throughout the year and that medium-size colonies of Atta and large colonies of Acromyrmex are unlikely to move over substantial distances (but see [60]), so their central place for foraging would tend to cover the same (sub) habitat over time. In spite of these caveats, our study shows that genus- and species-level differences across leaf-cutting ants can be quantified with the statistical tools we developed during this study.

The results of our study suggest that direct competition for forage material between the two genera of leaf-cutting ants is likely to remain limited, because Atta and Acromyrmex species target rather different types of forage, in spite of some overlap consistent with earlier reports that mostly report allopatrically collected data [18], [22], [24], [28]–[30], [32]. The correlations between garden enzyme activity and genus-level difference in forage use that we uncovered for the Gamboa community of leaf-cutting ants may be reinforced or supplemented by differences in salivary gland secretions between the two ant genera [61], a variable we were unable to measure. However, comparisons at the species level suggested that both Atta and Acromyrmex species tend to have habitats that are largely mutually exclusive, with Acromyrmex volcanus and Atta cephalotes foraging in the canopy, Acromyrmex octospinosus and (somewhat less specifically) Atta colombica foraging on the forest floor, and Acromyrmex echinatior and Atta sexdens foraging in the open landscape. Although it is possible that these differences are less pronounced in other seasons or sites, these results seem consistent with ecological theory predicting that interspecific competition is more pronounced when species are more similar, so that habitat partitioning may evolve [4]–[6].

The only case in which habitat segregation was somewhat less pronounced was between Atta sexdens and Atta colombica, which often overlapped in park-like and man-made habitats. Although there is a clear gradient across the isthmus of Panama, Atta sexdens is the dominant Atta species along the Pacific coast and becomes less abundant towards Gamboa in central Panama, whereas the pattern is opposite for Atta colombica [28]. It is interesting that these are the only two species for which we once observed active avoidance behavior on neighboring trails, i.e. trails of a colony stopping ca. one meter from the trail of another colony (P.W. Kooij, pers. obs.), behavior expected for all Atta spp. when foraging territories overlap. For the two common Panamanian Acromyrmex species, of which our research group has dug up ca. 500 colonies over the last two decades, habitat segregation (forest for A. octospinosus and open grassland areas for A. echinatior) is so pronounced that they will rarely encounter each other, similar to what is seen in Costa Rica [62]. In this way mature colonies of these species are unlikely to compete for the same type of plant forage. As far as we are aware distributions of incipient (founding) colonies are similar to those of mature colonies in Gamboa, but this is harder to quantify.