Explaining the Abundance of Ants in Lowland Tropical Rainforest Canopies

See allHide authors and affiliations

Science  09 May 2003:
Vol. 300, Issue 5621, pp. 969-972
DOI: 10.1126/science.1082074


The extraordinary abundance of ants in tropical rainforest canopies has led to speculation that numerous arboreal ant taxa feed principally as “herbivores” of plant and insect exudates. Based on nitrogen (N) isotope ratios of plants, known herbivores, arthropod predators, and ants from Amazonia and Borneo, we find that many arboreal ant species obtain little N through predation and scavenging. Microsymbionts of ants and their hemipteran trophobionts might play key roles in the nutrition of taxa specializing on N-poor exudates. For plants, the combined costs of biotic defenses and herbivory by ants and tended Hemiptera are substantial, and forest losses to insect herbivores vastly exceed current estimates.

Ants (Hymenoptera, Formicidae) comprise up to 94% of arthropods in insecticidal fogging samples from tropical rainforest (TRF) canopies and 86% of the biomass of those samples (13). Although commonly viewed as procuring much of their nitrogen (N) through predation and scavenging (4), ants are typically more abundant in canopy samples than are known arthropod herbivores. This observation is consistent with at least four untested and not mutually exclusive hypotheses. First, certain insect herbivores (such as immobile Coccoidea, which are often tended by ants) might be far more abundant than currently estimated, because of undersampling by canopy fogging (5) and the failure of leaf-area removal measures to quantify their impact (6). Second, high turnover of arthropod herbivores might sustain a high standing biomass of ants and produce inverted pyramids of numbers and biomass (2). Third, large populations of long-lived workers might be supported principally by abundant dietary carbohydrates and demand little protein. Fourth, ecologically dominant ant taxa (7) might themselves feed as herbivores, deriving both carbohydrates and N from plant and insect exudates (13, 8).

We focus here on the last hypothesis. Many tropical arboreal ants forage extensively for plant and insect secretions. Approximately one-third of tropical woody dicots and herbaceous vines produce extrafloral nectar (EFN) and/or lipid-rich pearl bodies as “biotic defenses” to attract ants that defend vegetative and reproductive structures against herbivores (9). Additionally, many arboreal ants attend sap-feeding Hemiptera (10) in suborders Auchenorrhyncha (especially Membracoidea) and Sternorrhyncha (especially Coccoidea and Aphidae). These “trophobionts” pass carbohydrate-rich “honeydew” to attending ants while concentrating N from N-poor plant sap. Protected by ants from predators and parasitoids, they can reach extremely high densities and transmit plant pathogens. (Some ants also harvest trophobionts as prey, but this phenomenon is poorly studied.) Although many abundant canopy ants unquestionably consume exudates (8), colony growth and reproduction require substantial N as amino acids and proteins (1, 3), and exudates are N-poor (11, 12). Consisting principally of carbohydrates and water, plant and insect secretions might simply fuel a search for protein (as the limiting factor for colony growth and reproduction) in the form of hunted or scavenged prey. Additionally, exudate feeders may process large volumes of liquid foods (3, 13) and/or extract or recycle N with help from microsymbionts (1417). The identification of N sources of abundant canopy ants is key to understanding the biology of these taxa (3), the structure of canopy arthropod communities, and ant effects on plant fitness.

The ratio of 15N/14N, formulated as δ15N [per mil (‰)], provides a tool for comparing N sources across ant species. The greater frequency at which the lighter isotope is lost in catabolic reactions leads to progressive enrichment of 15N relative to 14N with increasing trophic level (1821). We analyzed whole-body δ15N for a broad spectrum of ants from lowland TRFs of Peru and Brunei and compared δ15N ratios both across ant taxa and with data for coexisting plants, sap-feeding trophobionts, chewing herbivores, and arthropod predators (22). Additionally, based on activity frequencies in systematic surveys (23), arboreal foragers were categorized as mainly trophobiont tenders or as “leaf foragers,” defined as species searching leaf laminae continuously for dispersed foods [such as EFN, leaf diffusates, cast-off honeydew, isolated Coccoidea, plant wound secretions, prey, vertebrate feces, or pollen and microbes (8, 22, 23)]. Observations of predation and scavenging were also recorded for species in both categories.

Means and ranges for plants, sap-feeding trophobionts (Membracoidea and Coccoidea), chewing herbivores, and predatory arthropods calibrate the scales for ants (Figs. 1 and 2). Although higher δ15N values in Peru than in Brunei at all trophic levels remain unexplained (supporting online text), both data sets show the expected pattern of δ15N increase with trophic level. In Peru, means for chewing herbivores, but not for trophobionts, significantly exceeded those of plants (difference ∼1.8‰, F1,47 = 20.43, P < 0.0001, versus ∼0.25‰, F1,60 = 0.49, P = 0.48, respectively). In Brunei, means for both herbivore classes differed significantly from those of plants, but more so for chewing herbivores (∼ 2.6‰, F1,27 = 25.14, P = 0.0001) than for sap feeders (∼ 1.0‰, F1,27 = 5.02, P = 0.0335). In both forests, the mean ratios of predators exceeded those of chewing herbivores by ∼3.6‰.

Fig. 1.

Mean δ15N values (‰, by subfamily) for numbered Peruvian species (identified with sample sizes in table S1A). Predominant foraging modes are shown as follows (21): italics, leaf foraging; bold, trophobiont tending; underlined, predation; none of the above, poorly studied. Superscripts: A, ant-garden resident; F, attine fungus cultivators; M, specialized associate of myrmecophytes; P, brood parasite; U, unicolonial species. Calibration: Pl, 36 nonmyrmecophytic plant samples; Tr, 26 sap-feeding trophobionts, not from myrmecophytes; C, 13 chewing herbivores (Coleoptera, Orthoptera, and Diptera); P, 7 arthropod predators (Acari, Araneida, Hemiptera, and Pseudoscorpionida).

Fig. 2.

Mean δ15N values (‰, by subfamily) for numbered Bornean ant species (identified with sample sizes in table S1B). Predominant foraging modes are shown as in Fig. 1. Calibration: Pl, 20 nonmyrmecophytic plant samples; Tr, 9 sap-feeding trophobionts, not from myrmecophytes; C, 9 chewing herbivores (Coleoptera, Myriapoda, Orthoptera, and Gastropoda); P, 9 arthropod predators (Acari, Araneida, Phalangida, Pseudoscorpionida, Chilopoda, and Coleoptera).

At both sites, extensive overlap in ratios for sap feeders and plants is consistent with trophobiont acquisition of plant free amino acids, rather than breakdown of leaf tissue (24). Furthermore, members of both Sternorrhyncha (25) and Auchenorrhyncha (26) carry obligate intracellular bacterial symbionts that, at least in Aphidae, contribute to N nutrition of hosts. Housed in bacteriocytes lining the aphid hemocoel, microsymbionts convert glutamate, which is abundant in plant sap and is synthesized from other nonessential amino acids (27, 28), to a variety of isotopically lighter (18, 19) essential amino acids (2830).

Among ants (identified in table S1; see also supporting online text), δ15N values ranged over 9.5 and 9.0‰, in Peru and Brunei, respectively (Figs. 1 and 2). Calibration data show ratios below ∼5.5‰ in Peru, or 2.5‰ in Brunei, as indicating some N acquisition from sources other than predation and scavenging; numerous ant taxa fall below these thresholds. The highest δ15N values in both forests were for generalized and specialized predators [Ponerinae, Ecitoninae, and Aenictinae (13) (table S1)] and closely resembled ratios for other carnivorous arthropods. The lowest δ15N values for Peruvian ponerines and dolichoderines, and for Bruneian formicines and myrmicines, represent specialized inhabitants of myrmecophytes [“ant plants” (table S1)] that directly provision resident colonies with N (31, 32). Although these taxa fed as herbivores, other “plant ants,” especially certain Azteca species (neotropical Dolichoderinae, species 16 to 18), acquired considerable N through carnivory.

N sources were initially most uncertain for opportunistic exudate foragers (3, 8), including formicines, dolichoderines, and pseudomyrmecines, as well as many myrmicines (Figs. 1 and 2). High trophic diversity in the Myrmicinae (13) is evident from the broad ranges of δ15N ratios in this subfamily in both forests. Among Pheidolini and Pheidologetini (13), all but Pheidole minutula (no. 15), an inhabitant of New World myrmecophytes (table S1), and even P. biconstricta (no. 25), the most arboreal and exudate-dependent of our free-living Pheidole, are mainly or wholly carnivorous (in Peru, nos. 22, 25, 27, 28, 30, and 34; in Brunei, nos. 21, 25, 26, and 28 to 30). Of four trophobiont-tending Bruneian Myrmicaria, two species fed as carnivores (myrmicines 22 and 23), and two apparently garnered some N from other sources (myrmicines 12 and 17). Most free-living Crematogaster, which are almost all trophobiont tenders, fell marginally below thresholds for strict carnivory [nos. 8, 11 to 14, 17, and 18 (Peru); nos. 5, 7, 9 to 11, and 14 (Brunei) (supporting online text)].

Evidence for herbivory was strongest among leaf-foraging Cephalotini [Myrmicinae 1 to 7 (Fig. 1)] and Pseudomyrmecinae, trophobiont-tending dolichoderines, and formicines with various foraging modes (Figs. 1 and 2). Nevertheless, trials with mosquitoes showed that most species in all these taxa take weak or dead prey (or hemolymph) opportunistically (4). It is therefore remarkable that many free-living members of these groups exhibited δ15N ratios overlapping those of plants and trophobionts, whose modified mouthparts preclude predation and scavenging altogether. Moreover, despite the absence of folivory in ants (13), values for many more ant species overlapped those of chewing herbivores. We conclude that the δ15N ratios of exudate foragers in these taxa reflect a combination of infrequent-to-frequent carnivory and consumption of other low-ratio resources that reduce mean values to levels typical of trophobionts and other herbivores.

What could such resources be? First, some leaf foragers glean leaf laminae for epiphylls, including hyphae and spores [Camponotus, Polyrhachis, and Echinopla (Formicinae) (23)] or pollen and epiphylls [pseudomyrmecines and Cephalotes (13)]. Both fungivory and pollen feeding could produce δ15N ratios similar to those of chewing herbivores. (For example, see the value for Atta columbica, Peruvian myrmicine 10, whose larvae consume products of fungi cultivated on fresh leaves.) Second, all dolichoderines in our study were trophobiont tenders, and their larvae, with vestigial mouthparts, feed principally on worker-regurgitated liquids (33). Honeydew is slightly 15N-enriched relative to trophobionts (34), perhaps because nonessential plant amino acids exceed 15N-depleted essential amino acids manufactured by microsymbionts (12). Thus, several dolichoderines had ratios just slightly exceeding those of trophobionts. Harvesting of trophobiont hemolymph perhaps also contributed to slightly higher δ15N for tending ants, especially in Peru, where ratios of tending dolichoderines differed more substantially from those of trophobionts (Fig. 1).

Finally, microsymbionts of ants may contribute usable N to low-ratio formicines, myrmicines, and pseudomyrmicines, most of which are leaf foragers, rather than trophobiont tenders (23). The majority of formicines represent genera (Camponotus, Polyrhachis, and perhaps Echinopla) housing maternally transmitted γ Proteobacteria in mycetocytes lining the midgut lumen (14). Nesting phylogenetically among the microsymbionts of Auchenorrhyncha and Sternorrhyncha (25), these bacteria are thought to contribute to the colony's N budget (14), perhaps even by glutamate conversion (35). δ15N ratios are exceptionally low for some Camponotus, Polyrhachis, and Echinopla (Peru 1 to 9 and Brunei 3 to 11) and are intermediate for epiphyll-feeding congeners (at least 15, 17 to 19, 22, and 27 in Brunei) and the few Camponotus known for trophobiont tending (Peru 13 to 17). Values are still higher for carnivorous leaf foragers or trophobiont tenders (Peru 18 to 22 and Brunei 28, 29, and 32) and are highest of all for several predatory formicines from which microsymbionts are unknown: Oecophylla smaragdina (Brunei 36), a territorial predator and trophobiont tender (13, 36); two Paratrechina spp. (Peru 23 and Brunei 35); and Myrmelachista sp. (Peru 24). Neotropical Cephalotes (myrmicines 1 to 7) and paleotropical Cataulacus (myrmicines 2, 4, and 15) and Tetraponera (pseudomyrmecine 1) also carry microsymbionts hypothesized to contribute N to colony nutrition (1517), and these taxa exhibited δ15N ratios similar to those of the least predacious formicines in their respective forests (slightly higher in Tetraponera). A caecum at the junction of the mid-and hindguts in Tetraponera attenuata houses members of four genera of N-fixing bacteria (15) and is well supplied with tracheae and Malpighian tubules; in situ bacterial N fixation or N recycling are plausible scenarios. (For further discussion, see supporting online text.)

Coupled with observations of foraging, isotopic ratios lend insight into the probable effects of ants on TRF plants (supporting online text). Taxa obtaining substantial N from prey are most likely to benefit plants by reducing herbivore loads. However, the most abundant arboreal TRF taxa are typically trophobiont tenders (2, 3, 8) fueled by rapid growth of hemipteran populations (10, 36). Of genera appearing as superabundant in canopy-fogging samples [Camponotus, Oecophylla, Dolichoderus, Azteca, Technomyrmex, Crematogaster, Myrmicaria, and Paraponera (1, 3)], all but Paraponera and a few Camponotus species tend large populations of trophobionts (23), not only parasitizing plant N, carbohydrates, and water, but potentially transmitting plant pathogens. Among trophobiont tenders that are well represented in the mid-story or high canopy, only neotropical Azteca (dolichoderines 14 to 21), paleotropical Oecophylla (formicine 36), and the four unicolonial species included in our study (Figs. 1 and 2) obtain most of their N from prey. If the proportion of N acquired through carnivory correlates with the quality of protection afforded to plants by ants, this may help to explain why so many neotropical myrmecophytes have evolved relationships with Azteca species (37) and why Oecophylla has proven useful for 1700 years in suppressing herbivores of Chinese citrus crops (13) (supporting online text).

Together with the disproportionate representation of ants in canopy samples (13) and the probable undersampling of their trophobionts by standard methods (5, 6), large contributions of nonprey N to the diets of trophobiont-tending and other arboreal ants evidence the high combined costs of biotic antiherbivore defenses and herbivory by ants and their associates. Plant resource losses to insects, the dominant herbivores of lowland TRFs (6), greatly exceed the current estimate of 0.8 tons/ha/year [versus only 0.3 tons/ha/year for vertebrates (6)].

Supporting Online Material

Materials and Methods

SOM Text

Table S1

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article