Oecologia (2005) 142: 221–231 DOI 10.1007/s00442-004-1722-0
P L AN T A N IM A L I NT E R AC TI O NS
Diane W. Davidson
Ecological stoichiometry of ants in a New World rain forest
Received: 7 May 2004 / Accepted: 31 August 2004 / Published online: 21 October 2004 Springer-Verlag 2004
Abstract C:N stoichiometry was investigated in relation to diet (d15N), N-deprivation, and worker body size for a diverse assemblage of tropical Amazonian ants. Relative nitrogen (N) deprivation was assayed for 54 species as an exchange ratio (ER), deﬁned as SUCmin/AAmin, or the minimum sucrose concentration, divided by the minimum amino acid concentration, accepted as food by ‡50% of tested workers. On average, N-deprivation (ER) was almost ﬁvefold greater for N-omnivorous and N-herbivorous (N-OH) taxa than for N-carnivores. In two-way ANOVAs at three taxonomic levels (species and species groups, genera, and tribes), higher ER was associated with small body size and (marginally) with less carnivorous diets. ERs did not diﬀer systematically between trophobiont-tending and ‘‘leaf-foraging’’ functional groups, but specialized wound-feeders and coccid-tenders were prominent among high ER taxa. Paradoxically, some high ER taxa were among the most predatory members of their genera or subfamilies. Biomass % N was lower in N-OH taxa than in carnivores and varied inversely with N-deprivation (log ER) in the former taxa only. In an expanded data set, N-content increased allometrically in N-OHs, N-carnivores, and all ants combined, and with carnivory in large-bodied ants only. Exceptional taxa included small-bodied and predaceous Wasmannia, with high % N despite high ER, and Linepithema, with the lowest % N despite high d15N. Patterns in C:N stoichiometry are explained largely at the genus level and above by elemental composition of alarm/defensive/oﬀensive chemical weaponry and, perhaps in some cases, by reduced N investment in cuticle in taxa with high surface:volume ratios. Several Electronic supplementary material Supplementary material is available for this article at http://dx.doi.org/10.1007/s00442-0041722-0 D. W. Davidson Department of Biology, University of Utah, 257 South, 1400 East, Salt Lake City, UT 84112-0840, USA E-mail: [email protected]
consequences of C:N stoichiometry identify Azteca, and possibly Crematogaster, as taxa preadapted for their roles as prominent associates of myrmecophytes. C:N stoichiometry of ants should be incorporated into models of strategic colony design and examined in a phylogenetic context as opportunities permit. Keywords Allometry Æ Desiccation resistance Æ Exocrine product Æ Insect Æ Nitrogen limitation
Introduction Much ecological research on trophic interactions has focused on either energetics or consumer-resource interactions and their coevolution. However, there is mounting interest in the ﬁeld of ecological stoichiometry, deﬁned as ‘‘the study of the balance of multiple chemical elements in ecological interactions’’ (Elser et al. 2000b). Faced with spatial and temporal variation in the relative availabilities of essential nutrients, organisms are increasingly recognized to have adapted to variation in resource ratios by deploying these elements so as to increase reproductive ﬁtness, though under various historical constraints (reviewed in Elser et al. 2000b). Early studies in plants (e.g., Reichardt et al. 1991), though without direct reference to ‘‘stoichiometry,’’ were followed by work on aquatic organisms, but related ideas have been slower to appear in the literature on terrestrial animals. Despite similar C:N:P stoichiometry in freshwater and terrestrial herbivores, terrestrial food webs are built on a much more nutrient-poor autotrophic base (Elser et al. 2000a). Therefore, the eﬀect of ecological stoichiometry on the composition and structure of terrestrial communities is likely to be at least as great as that in aquatic communities, and it is important to develop more terrestrial models for such studies. Better understanding of elemental and ecological stoichiometry should contribute to our knowledge of the evolutionary ecology of focal taxa, the nature of their interactions
with other species, their roles in natural ecosystems, and their sensitivities to environmental stressors. Within species and developmental stages, and disregarding nutrient storage, multicellular animals appear to regulate elemental composition more stringently than do autotrophs (Elser et al. 2000b). Nevertheless, taxonomic diﬀerences in nutrient composition and use can be substantial, even at relatively low taxonomic levels. For example, biomass % nitrogen (N) of wild-caught, desertdwelling Drosophila fruit ﬂies ranged from 6.5 to 9.7% (factor of 1.5), correlating with N availability in the diet (Markow et al. 1999). The range may be broader still in ants (Hymenoptera: Formicidae), which include sapfeeders, scavengers and specialized predators, and oﬀer abundant opportunity to assess eﬀects of dietary resource ratios on insect ecology and evolution (e.g., Davidson 1997, 1998). Based on recent isotopic evidence, many tropical arboreal ant species appear to derive signiﬁcant fractions of their nitrogen (N) from herbivory [Davidson et al. 2003; Blu¨thgen et al. 2003, but see Gil et al. 2003; S. Cook and D. Davidson (unpublished work), for discussion of N recycling]. Direct activity censuses have placed these taxa in one or another of two functional groups of exudate-foragers (Davidson et al. 2003, 2004). Some species are principally dedicated to tending sapfeeding trophobionts [mainly Hemiptera in suborders Auchenorrhyncha (especially Membracoidea) and Sternorrhyncha (especially Coccoidea and Aphidae)]. Operationally, many others are ‘‘leaf-foragers,’’ scouring leaf laminae for dispersed resources in the form of extraﬂoral nectar (EFN), plant wound secretions, discarded honeydew from untended sap-feeders, fungal secretions, particulate matter such as pollen and fungal spores, and even animal feces (Wheeler and Bailey 1920; Buller 1950; Tobin 1994; Baroni Urbani and De Andrade 1997; Blu¨thgen and Fiedler 2002; Davidson et al. 2003, 2004). Considerable unresolved ecological variation surely exists within functional groups, but to some extent, membership in these groups follows taxonomic lines. Thus, Dolichoderinae specialize in trophobiont-tending, and Formicinae (with important exceptions) and Pseudomyrmecinae, in leaf-foraging (Davidson et al. 2004). Myrmicinae are more variable across species, genera, and tribes. Compared to animal material, many of the liquid resources used by herbivorous ants are both N-poor and unbalanced in amino acid (AA) composition (Baker et al. 1978, for EFN; Sandstrom and Moran 2001, for honeydew; Blu¨thgen and Fiedler 2004, for these and other exudates). How can exudate-feeding ants make their living on such resources? In mainly leaf-foraging camponotines (a tribe of Formicinae), endosymbiotic bacteria may contribute to the colony’s N-economy via AA upgrading and/or N-recycling (e.g., Gil et al. 2003). Some myrmicine and pseudomyrmecine leaf-foragers (cephalotine tribe group and Tetraponera) are also regularly associated with bacteria, these residing in or near the hindgut and presumed to recycle waste N (Caetano
and Cruz-Landim 1985; Van Borm et al. 2002). Although, upgrading of nonessential to essential AAs by hemipteran microsymbionts may not translate into more balanced AA proﬁles in honeydew (Sandstrom and Moran 2001), at least some hemipteran-tending ants harvest trophobionts or their hemolymph (e.g., Ho and Khoo 1997; Delabie 2001), and could be less N-deprived than are leaf-foragers. Despite these potential contributions to N nutrition, diets of N-poor and energy-rich honeydew and EFN may leave exudate-foraging ants with carbohydrates (CHOs) in excess of amounts that can be combined with N and utilized for colony growth and reproduction (Davidson 1997). It has even been suggested that EFN plants produce CHO-rich rewards to keep ants N-starved and avidly searching for N-rich prey (Carroll and Janzen 1973). However, in all exudate-feeders, lower N investment in various structures and functions, coupled with direction of excess CHOs to functions promoting N acquisition, could reduce C:N imbalance. Such functions may include ‘‘high tempo activity’’, leading to higher ‘‘dynamic densities’’ and faster protein discovery rates (Oster and Wilson 1978), defense of spatial territories with continuously renewing protein sources, and N-free or mainly carbon (C)-based alarm/ defensive (or oﬀensive) chemicals (hereafter ‘‘chemical weaponry’’), all of which may enhance ability to compete for protein (Davidson 1997, 1998). In view of these potential alternative uses for CHOs, it is possible that herbivorous ants might not be as N-starved as has been postulated (but see Kay 2002). No systematic tests have yet compared the extent of N-deprivation across major taxonomic and functional groups of ants. However, such comparisons are relevant to predicting which ant taxa are most likely to pursue prey, and thus to protect plants from insect herbivores. They may also explain why certain taxa are represented disproportionately among associates of myrmecophytic plants. This study presents the results of behavioral assays quantifying the degree of N-deprivation in 54 neotropical ant species ranging from highly herbivorous or omnivorous to mainly carnivorous. (Similar data for paleotropical ants currently lack replication for predatory species and will be augmented and published elsewhere.) Measures of N-deprivation are then related to stable isotope ratios of N (d15N), as indicators of trophic level (Davidson et al. 2003; Blu¨thgen et al. 2003) and to data on whole body biomass % N. Greater reliance on plant resources was hypothesized to leave ants more N-starved and to lead to reductions in biomass % N. Because N content scales allometrically in predatory insects generally (Fagan et al. 2002), worker body size was included as a factor in the analyses. The exact mechanism(s) linking % N to size remain to be elucidated. However, the possibility of assaying ants directly for N-deprivation precludes the need to rely on isotopic ratios as proxies for dietary resource imbalances, and allows direct assessment of variation in N-limitation with body size. Such possibilities identify ants as a
particularly useful system for studying elemental and ecological stoichiometry of insects (Kay 2002, 2004).
Materials and methods Studies were conducted during four dry season to wet season transition periods (September–November, 1998– 2001) at the Estacio´n Biolo´gica Cocha Cashu in tropical moist forest along the Rio Manu ﬂood plain in the Parque Nacional Manu, Department of Madre de Dios, Peru (1154¢S 7122¢W, 350 m elevation). Precipitation here averages 2 m/year and is strongly seasonal. The degree of protein deprivation across ant taxa was determined using the technique of exchange ratios (hereafter ERs), developed by economists to evaluate diﬀerent resources (apples and oranges, or CHO and N) in a single currency. (See Kay 2002 for application of this technique to ants, and demonstration that the assay measures net rates at which colonies expect to acquire individual resources elsewhere in their environments.) Ants were oﬀered either AAs or sucrose in a log2 range of weight per volume ratios, starting with low concentrations and administering successively higher concentrations until ‡50% of independently tested workers fed from drops of the solution. The next most concentrated solutions were also tested to conﬁrm that willingness to feed continued increasing at higher concentrations. Alternate trials rotated sucrose ﬁrst versus AAs ﬁrst, and drops were removed at the end of each worker trial. Six workers were sampled per colony if four or more ants either accepted or rejected the solution, but if observations were equally distributed between categories, sampling continued until observations in one category exceeded those in the other category by two. ER was then calculated as SUCmin/AAmin, or the ratio of the minimum sucrose (SUC) concentration to the minimum AA concentration for which half or more of the workers would forage. Higher ER values indicate greater N limitation, i.e., ants accepting sucrose only at high concentration, but AAs at low concentration. Because the two measurements were made at approximately the same times and places, dimensionless ratios can legitimately be compared across colonies and species. In contrast, neither SUCmin nor AAmin alone can be compared because of possible disparities in proximity to nests, perceived risk while foraging, and other variables inﬂuencing feeding selectivity (Davidson 1978; Nonacs and Dill 1991). Sucrose concentrations ranged from 0.031 to 16% wt/vol in the study as a whole. AA solutions were made by mixing all contents of an L-AA kit (Sigma), homogenizing the dry powder for 15 min, and then weighing out known quantities of the mixture for transport to the ﬁeld. Concentrations of AA solutions ranged from 0.031 to approximately 4% wt/vol. AAs did not dissolve completely at 0.4 g/10 ml, so dilutions from that most concentrated state were made from a wellmixed suspension. Just a few species refused to feed on
solutions of £ 2% AAs; ﬁve of these were listed as AAmin=8, because they also refused the 4% AA suspension. All solutions were prepared in the ﬁeld using distilled water and pre-weighed aliquots of sucrose or AAs, and new sets of solutions were constituted at least every 6 days. Drops were applied to epiphyll-free leaf laminae with micropipettes. Colonies never rejected concentrations greater than SUCmin or AAmin. Given the large number of species included in the study, substantial replication within all species was not possible, though replicate species were frequently sampled within genera (electronic supplementary material). Moreover, though SUCmin and AAmin often diﬀered across colonies of the same species (above), replicate ER values for many species tended to be close or even identical, possibly the result of the log scale of solution concentrations. Substantial replication was undertaken only for a subset of species. Certain ant taxa tended to accept lower concentrations of both SUC and AA solutions during dry periods, so testing was suspended after several days with no rain, and on any occasion when tested ants were found to take distilled water. Trials were also abandoned if ants were disturbed by other factors, e.g., falling debris, arriving nestmates, or the presence of competitors. To test the prediction that omnivores and herbivores are more N-deprived than are predatory/scavenging taxa, lognormally distributed ER values were compared between N-herbivores and N-omnivores (N-OHs) and ‘‘N-carnivores,’’ i.e., taxa with d15N £ 5.5& versus d15N>5.5&, the threshold value evidencing some N acquisition through herbivory (Davidson et al. 2003). Although most N-carnivores also take exudates when available, the low amounts of N in exudates do not contribute detectably to colony nutrition. Additionally, log ER was related to whole body d15N ratios and log worker body lengths (for outgoing, unladen workers, Davidson et al. 2004) in two-way ANOVAs at three taxonomic levels. [See Davidson et al. (2003) for methods of determining d15N, and below, for % N.] Isotopic ratios were species means (or individual values, where N=1) corresponding to those in Davidson et al. (2003), and therefore often represented broader groups of colonies than did behavioral assays. Log ER was predicted to scale inversely with both d15N, which increases with trophic level, and body size, based on the pattern for predatory insects generally (Fagan et al. 2002). Biomass % N Reports of isotopic analyses included measures of biomass % N, used to test independently for predicted inverse relationships with N-deprivation (log ER), diet (d15N), and log body size. Analyses included all taxa for which all pertinent measurements were available, except that specialized inhabitants of myrmecophytes were excluded. Sample size was smaller for models based on ER values, which were lacking for some taxa or impossible
to acquire (e.g., for army ants). Although results of whole body analyses might have been inﬂuenced by food in the gut, this eﬀect was deemed likely to be minor for unladen workers, relative to N as a component of worker exoskeleton, musculature, membranes and chemical weaponry (see Discussion). Phylogenetic considerations Inadequate knowledge of phylogenetic relationships among ant taxa precluded taking phylogeny into account explicitly in regression analyses. Instead, an attempt was made to reduce pseudoreplication by using means for informal species groups, genera, and tribes, in analyses performed at successively higher taxonomic levels. Bolton (1995) does not formally recognize species groups, but informal groupings are widely used, especially for highly speciose genera like Camponotus. With certain exceptions, nomenclature follows Bolton (1995). Although he designates no tribes within the diverse Dolichoderinae, smaller-bodied dolichoderines once assigned to the tribe ‘‘Tapinomini’’ (Ho¨lldobler and Wilson 1990) are separated here from representatives of the genus Dolichoderus due to anatomical and functional diﬀerences in the digestive system (Eisner 1957; Davidson et al. 2003), and from one another (Azteca vs Linepithema in % N analyses), based on such diﬀerences and the lack of close relationship in a recent phylogenetic treatment of the Dolichoderinae (Chiotis et al. 2000). All analyses were completed in JMP (SAS Institute 2001).
Results Feeding trials Forty-two of 54 Amazonian taxa were retained in ER data sets substituting means for taxa in informal species groups, and these taxa represented 17 genera and 16 tribes (electronic supplementary material). N-OH taxa exhibited signiﬁcantly greater ER values on average than did N-carnivores for both genera (Fig. 1, F1,15=7.46, P<0.02, r2=0.33) and tribes (F1,13=6.45, P=0.02, r2=0.32). Back-transformed least squares means were 8.6 and 1.8, respectively, for the genera and 8.3 and 1.7 for tribes. N-OH taxa all displayed relatively high ER values, but N-carnivores ranged from being somewhat CHO-limited (ERAzteca was the most Nlimited genus in the assay. In a two-way ANOVA investigating eﬀects of (uncorrelated) d15N and log body size on log ER at the level of species and species groups, a model lacking the non-signiﬁcant interaction term was signiﬁcant (F2,39=3.88, P=0.03, r2=0.17) and documented a signiﬁcant eﬀect of diet, but not size (P=0.01, vs P=0.10). The same model remained signiﬁcant after removal of an outlier, Solenopsis parabiotica, a tiny
Fig. 1 Distribution of mean ER values for N-OH and Ncarnivorous ant genera; note ordinate’s log scale. Subfamilies as in Fig. 4 legend. Vertical lines are ranges
brood parasite with unusually good access to protein for an ant of its body size (F2,38=4.30, P=0.02, r2=0.18). In that analysis, log ER varied inversely with both d15N (P=0.02) and log body size (P=0.03). With Solenopsis again removed, similar models were signiﬁcant at higher taxonomic levels (F2,13=5.31, P=0.02, r2=0.45 for genera, and F2,12=4.98, P=0.03, r2=0.45, for tribes). Eﬀects of body size were highly signiﬁcant in each case (P=0.01), but those of diet were just marginally so (P=0.09 for genera, and P=0.10 for tribes). Across individual species from genera rated as N-OH taxa, ER values diﬀered by a factor of 58.7, after deletion of an extreme outlier (Azteca sp. CC01-73, ER=131). Lowest values were for Pseudomyrmex tenuis and Cephalotes oculatus (both with ER=1), followed by two species in Camponotus subgenus Tanaemyrmex (atriceps and femoratus, mean ER=3.5). The ﬁrst two species are leaf-foragers, likely scouring leaf surfaces for microscopic rewards such as pollen and fungal spores (Wheeler and Bailey 1920; Baroni Urbani and De Andrade 1997), whereas both Camponotus species are dedicated trophobiont-tenders and voracious predators (author’s observation). Highest values were for the enigmatic Crematogaster acuta (ER=58.7), and several taxa regularly tending Coccidae: Azteca chartifex (=42), Azteca sp. CC00-102 (=32), and large-bodied Dolichoderus decollatus (=23.3). Also high were two largebodied plant-wound feeders, Camponotus depressus (32) and Camponotus sericeiventris (=22) (see Discussion). Among all species from N-carnivorous genera, extreme ER values diﬀered by a factor of 128. They were lowest in brood parasitic S. parabiotica and in large, actively predatory Paraponera clavata (both ER=0.25), followed successively by three Pheidole species ( £ 1.0) and Pachycondyla apicalis (=1.0). The highest value was for Wasmannia sp. 1 (CC99-170, ER=32), followed by Wasmannia auropunctata (=9).
Body composition Nitrogen content was analyzed only for genera and tribes (including Solenopsis), because morphological traits are more likely than behavioral traits to be ﬁxed at lower taxonomic levels. With high % N and intermediate d15N, Ochetomyrmex and its tribe (Formicoxini) were omitted as statistical outliers in these analyses (see Discussion). Table 1 reports the results of two-way ANOVAs investigating eﬀects of diet (d15N), and either log ER or (correlated) log body size on log biomass % N. In the ﬁrst model, N content varied inversely with log ER (Fig. 2, dashed line ﬁt to all genera in univariate regression), but a signiﬁcant interaction term indicated dependence on diet. % N was signiﬁcantly greater for Ncarnivores than for N-OH taxa (F2,13=142.30, P<0.0001, r2=0.96) and declined signiﬁcantly with increasing N-deprivation only in the latter taxa (solid line, unﬁlled circles), largely due to the extremely low N content of Azteca and Crematogaster. For N-carnivores, values are either high or intermediate, depending on the nature of chemical weaponry (see Discussion). ANOVAs at both genus and tribe levels also revealed signiﬁcant variation in % N with body size, d15N, and their interaction (Table 1 and below). When the data set was expanded to include ten additional, ER-unavailable taxa, the distributions of both untransformed and (especially) log transformed % N values departed signiﬁcantly from normality. However, data fell naturally into three categories: (1) low; (2) intermediate, and (3) high (Fig. 3), used as response variables in ordinal logistic regressions. Separate models incorporated diet as either d15N or d15N category (NOH taxa vs N-carnivores), with the latter corresponding more closely to graphical presentation of data in Fig. 4. In all cases, ordinal logistic regressions revealed signiﬁcant eﬀects of both log size and diet. Biomass % N increased with body size across all genera (Fig. 4a, P=0.01 in Spearman Rank Test) and for genera in each trophic category (Spearman Rank Tests: P=0.009 and 0.04, for N-carnivores and N-OH taxa, respectively). (As a courtesy to other investigators, Fig. 4a,b shows lines ﬁt to parametric regressions, with the caveat that assumptions of those analyses are violated.) However, values
were lower on average for N-OHs than for N-carnivores (medians of 11.6 vs 13% N, respectively; P<0.01 in Wilcoxon test). N content also increased with d15N for all taxa (P=0.05 in a Spearman Test), and for ants larger than the median body size (ﬁlled circles, Fig. 4b, P=0.0001, in a Spearman Test), but not for smallerbodied ants. On average, % N was lower in small-bodied than in large-bodied taxa (11.7 vs 13.05%, P=0.03 in Wilcoxon test). Wasmannia (CC99-170 and Wa. auropunctata), Ochetomyrmex (Oc = CC01-97), and Acromyrmex (CC01-24) exhibited especially high % N for their body sizes, and Linepithema sp. (Li=CC01-130) displayed an exceptionally low value for its size and (especially) trophic specialization (Fig. 4a,b respectively).
Discussion Data on N-deprivation and N content are considered separately below. Overall, both small body size and less carnivorous diets contributed to greater N-limitation and lower N content in ants. These patterns reveal strong linkages among nutrient consumption ratios, Ndeprivation, and nutrient use ratios, and document the importance of ecological stoichiometry in this major component of Amazonian ecosystems. N-deprivation As indexed by high ER values, N-deprivation was signiﬁcantly, and almost ﬁvefold, greater for N-OH ant taxa than for N-carnivores on average (Fig. 1). This ﬁnding, paralleling the results of Kay (2002), was obtained despite the likelihood that the former taxa invest some fraction of their ‘‘excess’’ CHOs in functions promoting N acquisition (Introduction and below). In regression analyses, however, log ER also varied inversely with body size, despite lower % N in smaller taxa (below), and was related more strongly to log body size than to diet (d15N). Surprisingly, several taxa with especially high ERs are among the most predatory members of their genera or subfamilies, as evidenced in
Table 1 Results of two-way ANOVAs or ordinal logistic regressions investigating determinants of biomass % N at genus and tribe levels (Ochetomyrmex excluded, see Discussion); eﬀects statistics in
same order as grouping variables. In expanded data set, N=27 genera and 19 tribes, and NS interaction terms were excluded. NS not signiﬁcant or marginally so at a=0.05
Whole model statistics
ER-available taxa: ANOVAs with response variable=log biomass % N Genus log ER, d15N, interaction F3,12=4.16, P=0.03, r2=0.51 F3,11=4.02, P=0.04, r2=0.52 Tribe log ER, d15N, interaction Genus log body size, d15N, interaction F3,12=6.65, P=0.007, r2=0.62 F3,11=5.10, P=0.02, r2=0.58 Tribe log body size, d15N, interaction Expanded taxon set: ordinal logistic regressions with response variable=% N category X22=12.92, P=0.002, r2=0.25 Genus log body size, d15N Tribe log body size, d15N X22=10.38, P=0.006, r2=0.27 Genus log body size, trophic category X22=17.39, P=0.0002, r2=0.33 Tribe log body size, trophic category X22=10.00, P=0.007, r2=0.26
Eﬀect probabilities 0.04, NS, 0.03 0.03, NS, 0.03 0.002, 0.003, 0.02 0.007, 0.007, 0.07 0.005, 0.05 0.03, 0.04 0.004, 0.009 0.02, 0.035
Fig. 2 Biomass % N versus N-deprivation (ER). Lines depict all taxa (P=0.08, broken line) or N-OH taxa (unﬁlled circles, solid line, P=0.05), but with Oc omitted in both cases. Taxa identiﬁed in Fig. 4 legend
Fig. 3 For data set expanded beyond ER-available taxa, % N categories used in two-way ordinal logistic regression at genus level: 1 low<10.5, 2 10.5
isotopic data. In an earlier study (Davidson et al. 2003), C. sericeiventris and C. depressus ranked, respectively, as the fourth and eighth most predatory of 21 sampled Camponotus species, D. decollatus as the most predatory Dolichoderus, Azteca species as the ﬁrst to the tenth most predatory of 21 dolichoderines (excepting Linepithema sp., for which ER data are lacking), and Wasmannia spp., the second and sixth most predatory of 36 myrmicines surveyed. Therefore, the greater N-limitation evidenced in N-OH taxa than in N-carnivores appears to be derived from characteristics of higher level taxa (genera through subfamilies), with the opposite pattern often existing within these groups. Trends among major taxa
Fig. 4 Representing 28 taxa, including 10 for which ER data are lacking: a % N versus log body size (broken line for all taxa but Oc). Filled circles and highest line denote N-carnivores, and unﬁlled circles, N-OH genera. For carnivores, P=0.02, r2=0.30, % N=11.09+2.00 log size. For N-OHs (except Oc), P=0.02, r2=0.59, % N=7.66+5.00 log size. b % N versus d15N; unﬁlled circles denote 14 genera smaller than median size (5.7 mm), and ﬁlled circles, 14 larger-bodied taxa. For the latter, parametric regression statistics are: P=0.0002, r2=0.70; % N=10.13 0.45 d15N. Abbreviations (Figs. 2 and 4, subfamilies in parentheses): Ac Acromyrmex (M), An Anochetus (P), Ap Apterostigma (M), At Atta (M), Az Azteca (D), Ca Camponotus (F), Ce Cephalotes (M), Cr Crematogaster (M), Cy Cyphomyrmex (M), D Dolichoderus (D), Da Daceton (M), E Ectatomma (EC), Ec Eciton (E), G Gigantiops (F), Gn Gnamptogenys (EC), La Labidus (E), Li Linepithema (D), N Nomamyrmex (E), O Ochetomyrmex (M), Od Odontomachus (P), P Paratrechina (F), Pa Pachycondyla (P), Pc P. clavata (PA), Ph Pheidole (M), Pl Platythyrea (P), Ps Pseudomyrmex (PS), S Solenopsis (M), and W Wasmannia (M). Subfamily identiﬁers (Bolton 2003): D Dolichoderinae, E Ecitoninae, EC Ectatomminae, F Formicinae, M Myrmicinae, P Ponerinae, PA Paraponerinae, PS Pseudomyrmecinae
correlate with both nutrient consumption ratios (reﬂected in d15N values) and nutrient use ratios. The former are regulated in part by digestive traits associated at generic and subfamilial levels with greater reliance on Npoor liquid foods (e.g., Davidson et al. 2004). The latter appear to correlate with the elemental composition of chemical weaponry, determined largely at genus level and above (Ho¨lldobler and Wilson 1990), and perhaps with size-correlated investment in protein-rich cuticle (next section).
Within higher taxa, greater N-deprivation may be associated with especially N-poor diets. ER values were especially high among taxa tending Coccidae, as opposed to Membracidae; no regular coccid-tenders were included among species with ER<23. Future studies might proﬁtably probe whether CHO:N ratios in honeydews of coccids, which are stationary after the crawler stage, diﬀer regularly from those of mobile treehoppers, which can migrate at all developmental stages to higherN feeding sites on new growth. Highest measured ERs among eight Camponotus species were for two specialized wound-feeders, C. sericeiventris and C. depressus. The former species consumes CHO-rich sap from trunks and branches of multiple live host trees (and lianas), some of which house polygynous colonies or colony fragments (Davidson et al. 2004). Camponotus depressus nests in dead Guadua bamboo culms and consumes wound secretions produced in prodigious quantities at insect-damaged tips of live culms (D. Davidson, J. Arias, J. Mann, and S. Castro, unpublished work). Unlike wound secretions from N-rich plant parts (see Blu¨thgen and Fiedler 2004), these exudates may be particularly unbalanced in CHO:N ratios or AA composition. Despite these special cases, the most dedicated trophobiont-tenders (Azteca and Dolichoderus) did not stand out consistently from genera composed primarily of leafforagers (Camponotus and Gigantiops, Fig. 1), and this was also true within the genus Camponotus (see Results). Body composition Across sampled Formicidae, 1.8-fold variation in biomass % N (7.7–13.8%) approached the 2.3-fold variation recorded for insects generally (Fagan et al. 2002), and as in these other insects, N content was labile at genus level and above. In two-way ANOVAs including just taxa represented by ER data, both signiﬁcantly lower % N in N-OH taxa, and an inverse relationship between % N and ER in N-OH taxa, helped to produce a marginally signiﬁcant inverse relationship overall (Fig. 2, Table 1). N-carnivores alone did not show this pattern but separated into groups with intermediate versus high % N, reﬂecting elemental composition of chemical weaponry. The latter group includes ﬁve genera of active hunters with functional stings and probably proteinaceous venoms. Such venoms are well documented for Pachycondyla, Paraponera, and Ectatomma (Morgan et al. 2003), and proteins or peptides are probable in the venoms of Gnamptogenys, a relative of Ectatomma (Bolton 2003). They likely account as well for the potent sting of Wasmannia (Howard et al. 1982), in which volatile alkaloids have been ruled out (see also Odontomachus and Platytheria in expanded data set; Morgan et al. 2003); cephalic alkylpyrazines also contribute to % N in this genus (Howard et al. 1982). Among carnivorous genera with intermediate % N, just Solenopsis possesses a functional sting (Ho¨lldobler and Wilson 1990), and its alkaloidal venom (Leclercq et al. 2000b) should be less
N-dense than are proteins and peptides. Carnivorous Paratrechina produces N-free, volatile formic acid as an alarm/defense product, and Apterostigma, not really predators, cultivate fungi on feces and dead wood (Ho¨lldobler and Wilson 1990); like other attines (below), they may store fungal enzymes in the gut. N-free alarm/ defense compounds are also known from most chemically well-characterized N-OH genera, including Camponotus and Gigantiops (volatile formic acid), Dolichoderus and Azteca (volatile iridoids), and Crematogaster [somewhat less volatile (more C-dense) compounds smeared directly onto enemy ants] (reviewed in Attygalle and Morgan 1984; Ho¨lldobler and Wilson 1990; Leclercq et al. 2000a). Pseudomyrmex may be exceptional, but the single studied species is a pugnacious resident of myrmecophytes (e.g., Pan and Hink 2000) and may not be representative of the genus. Other pseudomyrmecines produce less N-dense alkaloidal venoms (Leclercq et al. 2000b), and feeding on pollen and fungal spores (see Results), many Pseudomyrmex may deploy venoms mainly defensively. Consistent with the relationship of N-deprivation to both body size and diet, biomass % N was signiﬁcantly associated with both log size and d15N in ER-available taxa and the expanded data set (Table 1). % N increased signiﬁcantly with carnivory over all taxa combined in a non-parametric test, and in large-bodied taxa, which are less N-deprived (Fig. 2) but produce proteinaceous venoms if actively predatory (above). Most small-bodied genera fell below the regression line ﬁt to values for larger taxa. Disparities in the elemental composition of chemical weaponry clearly contributed to this ‘‘size effect’’. Except for Wasmannia, and perhaps Ochetomyrmex, most small-bodied genera possess N-free or alkaloidal chemical weaponry (above, and iridoids in Linepithema), whereas larger-bodied genera with known chemistry include N-OH taxa with N-free compounds and N-carnivores with proteinaceous venoms (above and Fig. 4b). Although N content increased with body size in both N-OH and N-carnivorous taxa, the median value for N-carnivores was 12.1% greater than that of N-OH taxa (Fig. 4a). A parallel statistic was slightly higher for insects generally (15%, Fagan et al. 2002), but many N-OH ant genera are predatory to a degree. In the earlier study, N content increased signiﬁcantly with body size for predatory but not herbivorous insects (Fagan et al. 2002). A possible explanation is that, as in ants, C:N dietary ratios of herbivores with diﬀerent host-plant or host-part specializations determine the sorts of chemical weaponry those insects sequester or synthesize, but chemistry was not reported. Additionally, volatiles (often N-free, or N-sparse) may generally be more eﬀective than are proteinaceous venoms in the chemical arsenals of small-bodied insects. If a threshold amount is required for eﬃcacy, independently of insect size, smaller-bodied taxa may also invest disproportionately in chemical weapons, leading to higher C:N ratios in those groups. In attempting to explain allometric scaling of % N in insects generally, Fagan et al. (2002) mention possible
variation in protein-to-chitin ratios in insect cuticle, as well as disparities in relative allocation among muscle, cuticle, fat body, and other tissues. Larger insects require disproportionately high investment in N-rich cuticle to resist buckling of exoskeleton under muscle force (Prange 1977; Anderson et al. 1979). Although thickness of exoskeleton scales allometrically at mass1.12, muscle crosssectional areas, and their force outputs, scale isometrically with mass2/3 (e.g., Nauen and Shadwick 1999). If cuticle regularly contains more N than does insect muscle tissue, the disparity could account for higher % N in larger ants. However, given existing tissue-speciﬁc data, this scenario seems unlikely (see Sterner and Elser 2002; Fagan et al. 2002), and N investment in both muscle mass and cuticular protein may vary independently for other reasons. In ants and insects generally, much of an organism’s cuticle may not be subject to high mechanical stress during locomotion, and cuticular investment could scale diﬀerently on those body parts. For example, among exudate-foragers (N-OHs), a requirement for an expansible gaster could necessitate reductions of cuticle on that structure. Moreover, because surface:volume ratios vary inversely with size, greater demand for cuticular protein could both account for higher ERs of smaller ants, and select for reduced % N in non-supporting cuticle (Figs. 2 and 4; see Hood and Tschinkel 1990, for habitat-correlated diﬀerences in ant cuticle). In favor of this argument, protein-bound lipids in epicuticle are a principal barrier to desiccation in insects (Gulan and Cranston 2000), and carnivorous Linepithema is characterized by both especially low % N and extraordinary susceptibility to desiccation (Walters and MacKay 2003). Also with very low % N, Azteca were the taxa most likely to feed on very low concentrations of solutes in ER trials, and unique in occasionally taking distilled water. Future comparisons of % N allometry in ants versus other insects may eventually show whether there is also a social component to such relationships. If colonies can make either many, low-N workers or few, high-N workers (Oster and Wilson 1978), then positive selection for rapid colony growth (or many independent searchers) may partly oﬀset the negative eﬀects of reduced cuticle. Rapid growth early in colony establishment may be particularly important for small-bodied species that hunt and defend their colonies cooperatively (McGlynn 1999), and numerical superiority often determines the outcome of combat and confrontations among ants (Franks and Partridge 1993; Palmer 2003). Extremely low worker N in carnivorous Linepithema (Fig. 4) is consistent with the observation that even very small ‘‘propagules’’ of congeneric L. humile are capable of rapid colony growth (Hee et al. 2000). At least three exceptional genera appear to have broken the constraints leading generally to low biomass % N in small-bodied ant taxa (Fig. 4b). Isotopically, Acromyrmex looks carnivorous, but it actually cultivates fungi on fresh vegetation (Ho¨lldobler and Wilson 1990). Its relatively high d15N value and N content may be aberrant, due in part to storage of diverse fungal enzymes
in the gut for use in fungal gardening (Ronhede et al. 2004). For Wasmannia, and perhaps Ochetomyrmex, high worker biomass % N may be possible by virtue of highly carnivorous diets. Tiny Wasmannia forage cooperatively in extremely populous and pugnacious colonies (McGlynn 1999), and W. auropunctata workers even attack Eciton burchelli army ants in the bivouac (M. Merwin, personal communication). High ER values in these tiny ants are harder to explain, but could either reﬂect exceptionally high N demand for cuticle due to relative large surface area, or be artifacts of how ants treated feeding assays. If Wasmannia depend on AAs for both energy and N, they may have viewed assays as choices between CHO (sucrose) alone and CHO plus N (AAs), and strongly preferred the latter. Based on a single anecdotal observation, the lifestyle of poorly studied Ochetomyrmex may somewhat resemble that of Wasmannia. d15N values typical of omnivores came from various annual collections at what were thought to be small, independent colonies in a restricted forest area. When that area was revisited in 2002, the ‘‘colonies’’ had coalesced into an extremely populous column resembling that of cooperatively hunting Wasmannia ‘‘supercolonies’’ (author’s unpublished observation). Diets of some omnivorous ant taxa become more carnivorous with colony age (author’s unpublished data), and this pattern was deemed particularly likely here. This statistical outlier was therefore deleted from analyses of % N. Although the proximate cause of high % N is unclear, Ochetomyrmex have functional stings (Bolton 2003), with as yet uncharacterized venoms. Finally, worker bodies were analyzed whole here, under the presumption that gut contents contribute little to % N of foraging workers, and despite Blu¨thgen et al.’s (2003) demonstration that % N values can diﬀer between workers with and without gasters. Gaster removal can aﬀect N content in multiple ways. If cuticle is reduced disproportionately on expansible gasters of exudatefeeders, removal could bias toward higher whole body % N. Gasters also contain the principal reservoirs of chemicals used for alarm, combat, recruitment and dispatching of prey (Ho¨lldobler and Wilson 1990). The assumption of negligible contribution of gut contents in unladen workers is likely justiﬁed overall, if not in every case (e.g., attines, above), because ER and % N varied less conspicuously with diet than with size and elemental composition of chemical weaponry. For example, carnivorous taxa in high and intermediate % N categories overlap substantially in their isotopic ratios (Fig. 4b). In summary, exocrine chemistry appears to account for much of the variation in biomass % N. Although this relationship could conceivably be determined independently of nutrient supply ratios, and solely on the basis of eﬃcacy under particular conditions, the association between % N and ER, coupled with the relationship of ER to worker size and diet, together suggest that dietary nutrient ratios have strongly inﬂuenced the evolution of chemical weaponry in ants.
Possible preadaption in myrmecophyte-associated genera Ideally, coevolved associates of tropical myrmecophytes (‘‘ant-plants’’) should be taxa proﬀering the best protection against herbivores, the most damaging of which are insects (Leigh and Windsor 1982). Based on isotopic evidence that Azteca are highly carnivorous members of their exudate-dependent subfamily, Davidson et al. (2003) suggested that members of this genus may have been preadapted to become prominent plant-ants in New World rain forests (rivaled only by Pseudomyrmex; Davidson and McKey 1993). Data here bolster that case for free-living Azteca species, which have the highest measured ER values (Fig. 1). Perhaps due to thin cuticle and low desiccation resistance, these taxa also use and even recruit to aqueous solutions with the lowest solute concentrations for both sucrose and AAs (D. Davidson and S. Cook, unpublished work, and above), as well as to exudates of damaged leaves (author’s observation). EFNs with high CHO and AA concentrations may not be required to attract Azteca, because water itself is a reward. (Interestingly, in view of carnivory in Azteca, myrmecophytic Cecropia produce rewards of glycogen, the CHO stored in muscle tissue.) Correlated with extraordinary N-deprivation, which should stimulate ardent pursuit of prey, Azteca are also marked by extremely low % N, possibly related to rapid colony growth, granting host plants large numbers of defenders (quickly) for the amount of N supplied ‘‘intentionally’’ as ant rewards, or perhaps ‘‘unwillingly’’ in honeydews of tended trophobionts. Because worker populations are supported mainly by CHOs, taxa with excess CHOs can have populous colonies, as Azteca often do. Finally, some of the strongest evidence for energy (CHO)dependent spatial territoriality in ants comes from studies of Azteca (Adams 1994), and constant patrolling of space by these territorial ants should be advantageous to host plants, as documented for both myrmecophyte inhabitants (Schupp 1986; Vasconcelos and Casimiro 1997) and free-living species (Overal and Posey 1990; Majer and Delabie 1993; De Medeiros et al. 1995, 1999; Vandermeer et al. 2002, though eﬀects of coccid-tending are seldom evaluated). Some of these attributes also apply to other taxa. Crematogaster, the preeminent genus inhabiting Old World myrmecophytes (Davidson and McKey 1993), shares most of these traits (Figs. 2 and 4) and protects at least some hosts from herbivory (e.g., Stanton et al. 2002; Heil et al. 2001). However, at least in free-living species, low % N may be related to C-dense chemical weaponry rather than to reductions in cuticle, because sensitivity to desiccation was not so apparent as in Azteca. Defense of true spatial territories also characterizes the two most N-deprived, wound-feeding Camponotus species (author’s observation). In contrast, for their d15N values (Fig. 4b), pollen/fungal feeding Pseudomyrmex have intermediate % N, and that predicted by their sizes (Fig. 4a).
Promising new directions for research on social insects This and other recent research may just have begun to uncover the consequences of resource imbalances for the ecology and evolution of social and solitary insects. For example, although ergonomic optimization models have greatly advanced our understanding of strategic colony design in relation to foraging, combat, and other tasks (e.g., Oster and Wilson 1978), the obvious next step is to integrate resource imbalances with energetic considerations. Second, N and C are just two of many essential resources of ants and other insects, and it may be proﬁtable to look at others (see also McKey, in press), including, e.g., salts, which are extremely limiting in most exudates, and phosphorus, needed for growth and reproduction (Elser et al. 2000b). With proper attention to the ants’ perspective (see above caution for Wasmannia), analyses of exchange ratios, using precisely composed nutrient solutions, are a promising way to measure relative limitation of various key nutrients without confounding them with other resources as in some previous studies (e.g., Kaspari and Yanoviak 2001, for tuna vs sucrose). Third, as systematists provide increasingly complete and rigorous phylogenetic templates for ants and other insects, it will be important to determine whether and how responses of these organisms to resource imbalances vary with divergence times and depth in the phylogeny. For example, do transitions from carnivory to omnivory precede or succeed reduced N investment in chemical weaponry during lineage evolution? Finally, better understanding of elemental stoichiometry in particular insect taxa should spawn novel and testable hypotheses about species interactions in ecological and evolutionary time, including a possible role of ecological stoichiometry in coevolution. Acknowledgements For granting or facilitating permissions to study and collect inside the Manu National Park, D.W.D. gratefully acknowledges Peru’s oﬃce of A´reas Naturales Protegidas, Instituto Nacional de Recursos Naturales (ANP-INRENA), and oﬃcials of the Parque Nacional Manu, Museo de Entomologı´ a de la Universidad Nacional Agraria La Molina, and Museo Nacional de Historia Natural. Voucher specimens have been deposited in entomological collections of these museums, as well as in the Natural History Museum of Los Angeles County. S.C. Cook commented helpfully on the manuscript and in discussion. This study was supported by the U.S. National Science Foundation (award IBN-9707932).
References Adams ES (1994) Territory defense by the ant Azteca trigona: maintenance of an arboreal ant mosaic. Oecologia 97:202–208 Anderson JF, Rahn H, Prange HD (1979) Scaling of supportive tissue mass. Q Rev Biol 54:139–148 Attygalle AB, Morgan ED (1984) Chemicals from the glands of ants. Chem Soc Rev 13:245–278 Baker HG, Opler P, Baker I (1978) A comparison of the amino acid complements of ﬂoral and extraﬂoral nectars. Bot Gaz 139:322– 332
230 Baroni Urbani C, De Andrade ML (1997) Pollen eating, storing and spitting by ants. Naturwissenschaften 84:256–258 Blu¨thgen N, Fiedler K (2002) Interactions between weaver ants Oecophylla smaragdina, homopterans, trees and lianas in an Australian rain forest canopy. J Anim Ecol 71:793–801 Blu¨thgen N, Fiedler K (2004) Competition for composition: lessons from nectar-feeding ant communities. Ecology (in press) Blu¨thgen N, Gebauer G, Fiedler K (2003) Disentangling a rainforest food web using stable isotopes: dietary diversity in a species-rich ant community. Oecologia 137:426–435 Bolton B (1995) A new general catalog of the ants of the world. Harvard University Press, Cambridge Bolton B (2003) Synopsis and classiﬁcation of Formicidae. Mem Am Entomol Inst 71 Buller AHR (1950) Researches on fungi, vol VII. Royal Society of Canada. University of Toronto Press, Toronto Caetano FH, Cruz-Landim C da (1985) Presence of microorganisms in the alimentary canal of ants of the tribe Cephalotini (Myrmicinae): location and relationship with intestinal structures. Naturalia 10:37–47 Carroll CR, Janzen DH (1973) Ecology of foraging by ants. Annu Rev Ecol Syst 4:231–257 Chiotis M, Jermiin LS, Crozier RH (2000) A molecular framework for the phylogeny of the ant subfamily Dolichoderinae. Mol Phylogenet Evol 17:108–116 Davidson DW (1978) Experimental tests of optimal diet predictions in a social insect. Sociobiology 4:35–41 Davidson DW (1997) The role of resource imbalances in the evolutionary ecology of tropical arboreal ants. Biol J Linn Soc 61:153–181 Davidson DW (1998) Resource discovery versus resource domination in ants: breaking the trade-oﬀ. Ecol Entomol 23:484–490 Davidson DW, McKey D (1993) The evolutionary ecology of symbiotic ant-plant relationships. J Hym Res 2:13–83 Davidson DW, Cook SC, Snelling RR, Chua TH (2003) Explaining the abundance of ants in lowland tropical rainforest canopies. Science 300:969–972 Davidson DW, Cook SC, Snelling RR (2004) Liquid feeding performances of ants: ecological and evolutionary implications. Oecologia 138:255–266 De Medeiros MA, Fowler HG, Bueno OC (1995) Ant (Hym., Formicidae) mosaic stability in Bahian cocoa plantations: implications for management. J Appl Entomol 119:411–414 De Medeiros MA, Delabie JHC, Fowler HG (1999) Predatory potential of the ant Azteca chartifex spiriti (Hymenoptera: Formicidae) in cocoa plantations of Bahia, Brazil. Cie Jaboticabal 27:41–46 Delabie JHC (2001) Trophobiosis between Formicidae and Hemiptera (Sternorrhyncha and Auchenorrhyncha): an overview. Neotrop Entomol 30:501–516 Eisner T (1957) A comparative morphological study of the proventriculus of ants (Hymenoptera: Formicidae). Bull Mus Comp Zool 116:441–490 Elser JJ, Fagan WF, Denno RF, Dobberfuhl DR, Folarin A, Huberty A, Interlandi S, Kilham SS, McCauley E, Schulz KL, Siemann EH, Sterner RW (2000a) Nutritional constraints in terrestrial and freshwater food webs. Nature 408:578–580 Elser JJ, Sterner RW, Gorokhova E, Fagan WF, Markow TA, Cotner JB, Harrison JF, Hobbie SE, Odell GM, Weider LJ (2000b) Biological stoichiometry from genes to ecosystems. Ecol Lett 3:540–550 Fagan WF, Siemann E, Mitter C, Denno RF, Huberty AF, Woods HA, Elser JJ (2002) Nitrogen in insects: implications for trophic complexity and species diversiﬁcation. Am Nat 160:784–802 Franks NR, Partridge (1993) Lanchester battles and the evolution of combat in ants. Anim Behav 45:197–199 Gil R, Silva FJ, Zientz E, Delmotte F, Gonza´lez-Candelas F, Latorre A, Rausell C, Kamerbeek J, Gadau J, Ho¨lldobler B, Van Ham RCHJ, Gross R, Moya A (2003) The genome sequence of Blochmannia ﬂoridanus: comparative analysis of reduced genomes. Proc Natl Acad Sci USA 100:9388–9393
Gullan PJ, Cranston PS (2000) The insects: an outline of entomology. Blackwell Science, Oxford Hee JJ, Holway DA, Suarez AV, Case TJ (2000) Role of propagule size in the success of the invasive Argentine ant. Conserv Biol 14:559–563 Heil M, Fiala B, Maschwitz U, Linsenmair KE (2001) On the beneﬁts of indirect defense: short- and long-term anti-herbivore protection via mutualistic ants. Oecologia 126:395–403 Ho CR, Khoo KC (1997) Partners in biological control of cocoa pests: mutualism between Dolichoderus thoracicus (Hymenoptera: Formicidae) and Cataenococcus hispidus (Hemiptera: Pseudococcidae). Bull Entomol Res 87:461–470 Ho¨lldobler B, Wilson EO (1990) The ants. Harvard University Press, Cambridge Hood WG, Tschinkel WR (1990) Desiccation resistance in arboreal and terrestrial ants. Physiol Ecol 15:23–36 Howard DF, Blum MS, Jones TH, Tomalski MD (1982) Behavioral responses to an alkypyrazine from the mandibular gland of the ant Wasmannia auropunctata. Ins Soc 29:369–374 Kaspari M, Yanoviak SP (2001) Bait use in tropical litter and canopy ants: evidence of diﬀerences in nutrient limitation. Biotropica 33:207–211 Kay A (2002) Applying optimal foraging theory to assess nutrient availability ratios for ants. Ecology 83:1935–1944 Kay A (2004) The relative availabilities of complementary resources aﬀect the feeding preferences of ant colonies. Behav Ecol 15:63–70 Leclercq S, Biseau J-C de, Daloze D, Braekman J-C, Pasteels JM (2000a) Five new furanocembranoids from the venom of the ant Crematogaster brevispinosa ampla from Brazil. Tetrahedron Lett 41:633–637 Leclercq S, Braekman JC, Daloze D, Pasteels JM (2000b) The defensive chemistry of ants. Prog Chem Org Nat Prod 79:115– 229 Leigh EG Jr, Windsor DM (1982) Forest production and regulation of primary consumers on Barro Colorado Island. In: Leigh EG Jr, Rand AS, Windsor DM (eds) The ecology of a tropical forest: seasonal rhythms and long-term changes. Smithsonian Press, Washington DC, pp 111–122 Majer JD, Delabie JHC (1993) An evaluation of Brazilian cocoa farm ants as potential biological control agents. J Plant Prot Trop 10:43–49 Markow TA, Dobberfuhl D, Breitmeyer CM, Elser JJ, Pfeiler E (1999) Elemental stoichiometry of Drosophila and their hosts. Funct Ecol 13:78–84 McGlynn TP (1999) Non-native ants are smaller than related native ants. Am Nat 154:690–699 McKey D, Gaume L, Brouat C, Di Gusto B, Pascal L, Debout G, Dalecky A, Heil M. Multitrophic interactions among tropical plants, ants and plant herbivores. In: Burslem DRRP, Pinard MA, Hartley SE (eds) Biotic interactions in the tropics. Cambridge University Press, Cambridge (in press) Morgan ED, Jungnickel H, Keegans SJ, Do Nascimiento R, Billen J, Gobin B, Ito F (2003) Comparative survey of abdominal gland secretions of the ant subfamily Ponerinae. J Chem Ecol 29:95–114 Nauen JC, Shadwick RE (1999) The scaling of acceleratory aquatic locomotion: body size and tail-ﬂip performance of the California spiny lobster Panulirus interruptus. J Exp Biol 202:3181–3193 Nonacs P, Dill LM (1991) Mortality risk versus food quality tradeoﬀs in ants’ patch use over time. Ecol Entomol 16:73–80 Oster GF, Wilson EO (1978) Caste and ecology in the social insects. Princeton University Press, Princeton Overal WL, Posey DA (1990) Use of Azteca sp. ants for biological control of agricultural pests among the Kayapo indians of Brazil. In: Posey DA, Overal WL (eds) Ethnobiology: implications and applications. Museu Paraense Emilio Goeldi, Belem, pp 219–226 Palmer TM (2003) Spatial habitat heterogeneity inﬂuences competition and coexistence in an African acacia ant guild. Ecology 84:2843–2855
231 Pan J, Hink WF (2000) Isolation and characterization of myrmexins, six isoforms of venom proteins with antiinﬂammatory activity from the tropical ant, Pseudomyrmex triplarinus. Toxicon 38:1403–1413 Prange HD (1977) The scaling and mechanics of arthropod exoskeletons. In: Pedley TJ (ed) Scale eﬀects in animal locomotion. Academic, New York, pp 169–181 Reichardt PB, Chapin FS III, Bryant JP, Mattes BR, Clausen TP (1991) Carbon/nutrient balance as a predictor of plant defense in Alaskan balsam popular: potential importance of metabolic turnover. Oecologia 88:401–406 Ronhede S, Boomsma JJ, Rosendahl S (2004) Fungal enzymes transferred by leaf-cutting ants in their fungal gardens. Mycol Res 108:101–106 Sandstrom JP, Moran NA (2001) Amino acid budgets in three aphid species using the same host plant. Phys Entomol 26:202– 211 SAS Institute (2001) JMP, version 4. SAS Institute, Cary Schupp EW (1986) Azteca protection of Cecropia: ant occupation beneﬁts juvenile trees. Oecologia 70:379–385 Stanton ML, Palmer TM, Young TP (2002) Competition-colonization trade-oﬀs in a guild of African acacia-ants. Ecol Monogr 72:347–363
Sterner RW, Elser JJ (2002) Ecological stoichiometry. Princeton University Press, Princeton Tobin JE (1994) Ants as primary consumers: diet and abundance in the Formicidae. In: Hunt JH, Nepala CA (eds) Nourishment and evolution in insect societies. Westview, Boulder, pp 279– 308 Van Borm S, Buschinger A, Boomsma JJ, Billen J (2002) Tetraponera ants have gut symbionts related to nitrogen-ﬁxing rootnodule bacteria. Proc R Soc Lond B 269:2023–2027 Vandermeer J, Perfecto I, Ibarra-Nunez G, Phillpott S, GarciaBallinas A (2002) Ants (Azteca sp.) As potential biological control agents in shade coﬀee production in Chiapas, Mexico. Agrofor Syst 56:271–276 Vasconcelos HL, Casimiro AB (1997) Inﬂuence of Azteca alfari ants on the exploitation of Cecropia trees by a leaf-cutting ant. Biotropica 29:84–92 Walters AC, MacKay DA (2003) An experimental study of the relative humidity preference and survival of the Argentine ant, Linepithema humile (Hymenoptera: Formicidae): comparisons with a native Iridomyrmex species in South Australia. Ins Soc 50:355–360 Wheeler WM, Bailey IW (1920) The feeding habits of pseudomyrmine and other ants. Trans Am Philos Soc 22:235–279