Ellen van Wilgenburg, Department of Zoology, University of Melbourne, Melbourne, Vic. 3010, Australia. Tel.: +61 3 8344 6264; fax: +61 3 8344 7909; e-mail: email@example.com
The cuticular hydrocarbons (CHCs) of ants provide important cues for nest-mate and caste recognition. There is enormous diversity in the composition of these CHCs, but the manner in which this diversity has evolved is poorly understood. We gathered data on CHC profiles for 56 ant species, relating this information to their phylogeny. We deduced the mode of evolution of CHC profiles by reconstructing character evolution and then relating the number of changes in CHC components along each branch of the phylogeny to the length of the branch. There was a strong correlation between branch length and number of component changes, with fewer changes occurring on short branches. Our analysis thereby indicated a gradual mode of evolution. Different ant species tend to use specific CHC structural types that are exclusive of other structural types, indicating that species differences may be generated in part by switching particular biosynthetic pathways on or off in different lineages. We found limited, and contradictory, evidence for abiotic factors (temperature and rainfall) driving change in CHC profiles.
Pheromones, chemical signals secreted externally by one individual and eliciting a specific response in a conspecific individual (Karlson & Luscher, 1959), play an important communication role in a variety of contexts, including mate attraction, kin recognition, raising alarm, defining a territory and foraging. Pheromones may be a single compound or, more typically, are composed of blends of several compounds. The types of compounds and the relative proportions in which they occur are often highly species specific. Wyatt (2009, 2010) draws a distinction between pheromones and ‘signature mixtures’; the former being a compound or subset of compounds that elicit an innate response in conspecifics, the latter being variable subsets of molecules of an animal’s chemical profile that are learnt by other animals. This study takes the broader definition of pheromones (Karlson & Luscher, 1959; Wyatt, 2003), because the main focus is signal evolution rather than the mechanisms of perception.
Phylogenetic-based studies of chemicals used in communication (such as pheromones) are still rather uncommon, but can provide insight into the evolutionary forces driving their diversity (Symonds & Elgar, 2008). Understanding the mode of pheromone evolution lies at the heart of this approach. Under a gradual mode of evolution, species accumulate relatively small changes in type and number of the chemicals over time in relation to the overall variation across species, resulting in closely related species having more similar pheromone blends than more distantly related species (Roelofs et al., 1982). The alternative mode of evolution consists of more major ‘saltational’ shifts in pheromone blends that occur at speciation events, generating phenotypes that are substantially different from their close relatives (Baker, 2002). The saltational mode of evolution is likely to occur when species benefit from having unique pheromone blends, for example, when closely related, reproductively isolated, species live in sympatry. By relating phenotypic change to branch length (genetic distance or evolutionary time), it is possible to ascertain the mode of evolution (Pagel, 1999). A gradual mode of evolution predicts a close correlation between branch length and the number of changes (loss or addition of components) in pheromone blends, whereas saltational evolution implies large changes in profiles that may be associated with short branches (Baker, 2002; Symonds & Elgar, 2004, 2008).
Investigations into the mode of pheromone evolution indicate that the evolution of these signals depends on the functional context of the chemical signal. For example, blends of aggregation pheromones of closely related species of bark beetles are just as, or even more different than those of more distantly related species (Symonds & Elgar, 2004), suggesting a saltational, rather than gradual mode of evolution. Similar patterns exist for sex pheromones of Yponomeuta moths and Drosophila flies (Lofstedt et al., 1991; Lofstedt, 1993; Ferveur, 2005). These signals are important in species recognition and maintaining species isolation, and thus are predicted to evolve via sudden shifts in composition (Butlin & Trickett, 1997; Symonds & Elgar, 2008). In contrast, the aggregation pheromones of Drosophila, which do not require a species-specific response, evolve via a gradual mode of evolution, with small changes in composition over evolutionary time (Symonds & Wertheim, 2005). Given these differences in the patterns driving the evolution of different types of pheromones, we require insights into the mode of evolution of pheromones that are involved in other functions.
Analysis of the chemical composition of ant CHCs profiles over the past 40 years has yielded the identity of almost 1000 different CHCs (Martin & Drijfhout, 2009a). However, the manner in which this diversity has evolved and the ecological processes that drive this evolution remain poorly understood. The available evidence is contradictory. Closely related ant species often differ in their use of alkenes or di-methylalkanes (Akino et al., 2002; Lucas et al., 2005; Martin et al., 2008b), suggesting that substantial changes in CHC composition have occurred at speciation events. Further, a dendrogram based on CHCs from 78 species of ants bore only partial resemblance to a phylogeny based on morphological characters (Martin & Drijfhout, 2009a), suggesting a saltational mode of evolution. However, similar analyses within the genera Cataglyphis and Formica (Dahbi et al., 1996; Martin et al., 2008a,b respectively) revealed differences in CHC profiles that partially reflect taxonomic groupings between species, which might indicate that the evolution of hydrocarbon profiles may be a more gradual process. Using CHC profile information to group together species is interesting, but potentially problematic if the underlying mode of evolution is strongly saltational. However, to elucidate accurately the mode of CHC evolution, we need to map information about CHCs directly onto a phylogeny produced by other means. We predict that as CHC profiles are probably of only minor importance in species recognition, there should not be strong selective pressures driving apart CHC blends. Consequently, we expect the mode of evolution to be gradual. Under such circumstances, closely related species should have similar CHC profiles, and more specifically, there should be a direct relationship between the length of branches in the phylogeny and the amount of change in CHC blends along them.
In addition to identifying the mode of evolution, we sought to identify if climate drives CHC evolution in ants. Climate may play a role in CHC evolution because in addition to playing a role in recognition, CHCs prevent desiccation, and desiccation resistance may be a significant constraint for species living in warmer and drier climates (e.g. Rouault et al., 2004). Hydrocarbon melting temperatures and permeability to water decrease with decreasing backbone length, the presence of double bonds and methyl branching (Hadley, 1978; Toolson, 1984; Gibbs, 1995; Gibbs, 1998; Rourke & Gibbs, 1999; Gibbs & Rajpurohit, 2010). We therefore predict that species living in warmer and drier climates would have CHC blends with on average longer chain lengths and be less likely to utilize alkenes and branched alkanes.
Our comparative analysis maps CHC profiles from 56 species of ants onto a composite evolutionary tree based primarily on Moreau et al.’s (2006) extensive molecular phylogeny of the family Formicidae. First, we investigated the mode of evolution that has generated CHC diversity. Second, we investigated the role that climate may have had on CHC evolution in ants.
Choice of study species and data collation
We collated data for 56 ant species whose phylogenetic relationships and CHCs have been identified in sufficient detail (see Table S1). Data on CHC composition were taken initially from the recent review by Martin & Drijfhout (2009a), but verified and modified by consulting the original sources. Slavemaking species were excluded from our analysis as their CHC profiles are often derived from their host species (Brandt et al., 2005). For each species, we noted the presence or absence of compound groups (alkenes, alkadienes, mono-methyl branched alkanes, di-methyl branched alkanes, tri-methyl alkanes and methyl alkenes) and the presence or absence of homologous hydrocarbons series. Homologous hydrocarbon series are series of structurally related hydrocarbons that possess the same methyl group or double bond, but have a backbone that varies in length by increments of two carbons (e.g. 15-MeC35, 15-MeC37, 15-MeC39). We included these series rather than individual hydrocarbons in our analysis because the relative amounts of compounds within a homologous series are often constant, both within colonies and species (Martin & Drijfhout, 2009a; van Wilgenburg et al., 2010). We noted the presence of hydrocarbons rather than their relative proportions in the CHC profile because CHC profiles of conspecific colonies differ in HC ratios but not in the presence or absence of compounds, whereas species differ most prominently in the presence or absence of compounds. We chose to not include the n-alkanes in our analysis as they are present in all species and probably do not function as colony-mate recognition cues (Bonavita-Cougourdan et al., 1987; Dani et al., 2005; Lucas et al., 2005). For species in which the location of the double bond in alkenes and alkadienes has not been identified, we only noted the presence or absence of alkenes and alkadienes in general. In addition to treating each of the homologous CHC series as separate characters, we also performed the analysis using the five structural groups outlined above as characters. We also noted the backbone length of the shortest CHC, the backbone length of the longest CHC and the total number of different CHCs.
We reconstructed a phylogenetic tree (Fig. 1) based on recently published molecular phylogenies of ants. The primary phylogeny depicting intergeneric relationships was that of Moreau et al. (2006). Three genera in our analysis (Acromyrmex, Manica, Nothomyrmecia) were not included in that phylogeny, and so we derived information on their relationships following Brady et al. (2006). We resolved the remaining intrageneric relationships using the following phylogenies: Formica– (Goropashnaya, 2003); Lasius– (Maruyama et al., 2008); Temnothorax– (Beibl et al., 2005); and Cataglyphis– (Agosti, 1990). Branch length information (millions of years) was derived from the primary phylogeny (Moreau et al., 2006), with branch lengths for the additional genera and intrageneric branches being re-scaled from their source publications to fit this scheme. ‘Re-scaling’ was achieved by setting the age of the basal node in the generic phylogeny as the age of the branching point for that genus in the primary phylogeny, and using this basal reference age to recalculate the branch lengths based on their relative lengths in the intrageneric phylogenies. In one case (Cataglyphis), there were no intrageneric branch length data, so we assigned equal branch lengths within the genus (each branch length = 7.1 million years). While this approach undoubtedly adds error to the analysis, it is an acceptable default in comparative analyses where no branch length information is available (see e.g. Purvis et al., 1994; Diaz-Uriarte & Garland, 1998).
Mode of CHC evolution
We assessed whether CHC profile differences were related to phylogeny by estimating the number of changes, both in groups and in individual homologous hydrocarbon series, along each branch of the phylogeny. To do this, we used a Bayesian approach, as implemented through the program simmap (Bollback, 2006) to reconstruct character histories for each component along the phylogeny. One thousand reconstructions were generated for each component, from which we could derive a probability of change in that component (i.e. loss or gain) for each branch in the phylogeny. These probabilities were then summed across all components, providing us with a measure of total estimated number of component changes for each branch. This total was then correlated with the branch length, to test the mode of evolution (Pagel, 1999). The analysis was also repeated using only the presence or absence of the five structural groups whose presence varied across species (see Results).
As with earlier analyses (Symonds & Elgar, 2004; Symonds & Wertheim, 2005; Symonds et al., 2009), we also assessed the relationship between phylogenetic distance and pheromonal differences. We constructed a pairwise matrix of phylogenetic distances by summing branch lengths between each species pair, and a similar matrix of pheromonal differences by measuring the binary squared Euclidean distance (the total number of CHCs that are absent in one species but present in the other, and vice versa). The correlation between phylogenetic distance and pheromonal differences was calculated using Mantel tests, with the rows and columns of the distance matrix being randomly perturbed and the correlation coefficient recalculated 999 times to generate a null reference distribution. These tests were performed using the program GenAlEx (Peakall & Smouse, 2006).
These two tests of evolutionary change are not perfect as indicators of the mode of evolution. Specifically, the presence of positive relationships between number of differences/changes and phylogenetic distance/branch lengths need not automatically indicate a gradual mode of evolution. If saltational changes in chemical composition at speciation events are followed by subsequent gradual chemical divergence, or there is very great diversity in the number of chemical compounds available for use as part of the CHC profile of a species, phylogenetic distance and pheromonal differences (or the estimated number of component changes for each branch and branch length) can also be correlated under a saltational mode of evolution (Symonds et al., 2009). Therefore, in addition to looking at these correlations, we assessed the mode of evolution by looking at the location at which the best-fit lines intercept the y-axis. If state changes are concentrated at speciation events, the y-axis is expected to intercept away from the origin, whereas under gradual evolution, the best-fit line is expected to intercept at the origin. Additionally, we calculated the phylogenetic signal in the presence or absence of each of the five structural groups of CHCs. We used the Continuous program of the BayesTraits package (Pagel, 1999; Pagel & Meade, 2004) to calculate the maximum likelihood (ML) values for λ, the extent to which the trait is predicted by the phylogeny (if λ = 1, it indicates that trait evolution mirrors the phylogeny, whereas λ = 0 indicates no relationship between phylogeny and the trait). We also calculated the ML value for the branch length scaling parameter, κ, which indicates the mode of evolution for individual traits with κ = 1 indicating that character change is directly related to branch length (i.e. a gradual mode of evolution) and κ = 0 indicating the character change is unrelated to branch length (i.e. a punctuated mode of evolution for that trait).
Climatic factors driving CHC evolution
Information on geographical location of the ant species was collated from the CHC source literature, by taking the latitude and longitude of the collecting location of the samples used in the CHC analyses. Where exact source location was not specified, such as when a country only was stated as the location, we used the latitudinal and longitudinal midpoint of this area as a geographical location for the species. The geographical location data were used to extract temperature and precipitation data for the location from NOAA’s National Geophysical Data Center (http://map.ngdc.noaa.gov/website/timeline/viewer.htm). Annual mean temperature (°C) and annual mean rainfall (mm) data were taken. Rainfall data were unavailable for two species (Dinoponera quadriceps and Pachycondyla villosa), so the sample was reduced to 54 species for the rainfall analysis.
We assessed whether species in colder and wetter climates (i.e. low temperature, high rainfall) had more volatile hydrocarbons in their profiles by testing the correlations between species’ environmental temperature and rainfall, and the carbon chain length of the longest and shortest CHC (for all compound groups), and between the climatic factors and the presence of methyl branched alkanes (di and trimethyls), alkenes and alkadienes, respectively. Species do not provide independent data points for analysis, on account of their shared evolutionary histories (Harvey & Pagel, 1991). Consequently, our analysis of the influence of climate on CHC variation controlled for phylogenetic effects by using phylogenetic generalized least squares (PGLS) regression (Martins & Hansen, 1997), as implemented through the program COMPARE (Martins, 2004). In these models, CHC characteristics were the dependent variables, and the rainfall and temperature were independent predictors. In the case of each dependent variable, we compared the results from the models with these predictors individually with the model that included both predictors, using Akaike’s information criterion (Burnham & Anderson, 2002; Symonds & Moussalli, 2011). The two predictors are correlated (r = 0.441, n = 54, P < 0.001). In no case did the two predictor model provide the best approximating model; hence, we restricted our interpretation to the single factor models.
The ant species have on average 18.75 ± 9.84 different homologous alkene and methyl branched alkane series within their profile. However, this number is an underestimate because in 48% of the species, the positions of the double bonds were not characterized for all or some of the alkenes and alkadienes. Of the 56 species included in our analysis, all produced methylalkanes, 88% produced dimethylalkanes, 23% produced trimethylalkanes, 66% produced alkenes, 18% producted alkadienes and 7% produced methylalkenes. Ancestral state reconstruction using simmap identified that the presence of methylalkanes and dimethylalkanes is ancestral to ants, whereas the expression of alkadienes, methylalkenes and trimethylalkanes is derived in the lineages that express them. The ancestral state of alkenes is equivocal with simmap assigning a 0.53 probability, and the expression of alkenes is ancestral.
We found that shorter branches in the phylogenetic tree were associated with smaller changes in CHC profiles (Fig. 2). The result was the same whether we counted all homologous hydrocarbon series as individual characters (r = 0.446, n = 110, P < 0.001, Fig. 2a) or using structural groups as characters (r = 0.560, n = 110, P < 0.001, Fig. 2b). Both results thereby support a gradual rather than saltational mode of evolution. In addition, the best-fit lines for Fig. 2a,b cross the y-axes very close to the intersection, also indicating that small branches are associated with small amounts of change in CHC profiles.
There was no overall correlation between phylogenetic distance and CHC differences for individual homologous hydrocarbon series (Mantel test: r = 0.048, P = 0.282; Fig. 3a). However, a significant positive correlation was evident when structural groups were considered as characters (Mantel test: r = 0.181, P = 0.001; Fig. 3b), indicating that closely related species tend to use the same groups of compounds, if not the exact compounds. The results for homologous CHC series as individual characters exhibit two other effects. First, the values in Fig. 3a suggest a curvilinear relationship with a positive relationship being more evident among the less phylogenetically distinct species. Second, the majority of the values disruptive to a significant relationship in the left-hand part of the graph stem from comparisons with a single species, Cataglyphis rosenhaueri. This species contributes most of the outlying points on the part of the graph up to 50 million years (see Fig. 3a), as well as the outlier with the shortest branch length in Fig. 2a. Removal of this species from the analysis results in a highly significant correlation between phylogenetic distance and CHC differences for species separated by up 50 million years (r = 0.341, P < 0.001).
Analysis of the degree of phylogenetic signal in the presence or absence of the five variable structural groups (Table 1) showed that the ML values for Pagel’s λ were either 0 (no phylogenetic signal) or not significantly different from 0 (as judged by a likelihood ratio test), in all cases except for dimethylalkanes, where the phylogenetic signal is strong (λ = 1). The branch-length scaling parameter κ was 1, or very close to 1, in all cases, indicating the evolutionary change in the use of CHC structural groups by ant species is strongly related to branch length (i.e. a gradual mode of evolution).
Table 1. Analysis of phylogenetic signal in the presence/absence of the five structural groups of cuticular hydrocarbons (CHCs) that vary across the ant species in the analysis, estimated as the maximum likelihood values for λ and κ. For explanations, see Methods.
Variation in CHC composition was not related to temperature. For the 56 species in the analysis, there was no significant relationship between annual average temperature and the length of the shortest or longest carbon chain used by the species (shortest: r = 0.130, n = 56, P = 0.335; longest: r = 0.138, n = 56, P = 0.306). Nor was there any association between latitude and the presence of any specific structural group (alkenes: r = −0.087, P = 0.520; alkadienes: r = −0.042, P = 0.756; methylalkenes: r = 0.015, P = 0.912; dimethylalkanes: r = −0.174, P = 0.196; trimethylalkanes: r = −0.019, P = 0.888; n = 56 in all cases).
There was a significant positive correlation between rainfall and length of the shortest CHC chain (r = 0.372, n = 54, P = 0.005) but not with length of the longest CHC chain (r = 0.193, n = 54, P = 0.158). There was also a tendency for species living in wetter environments to be more likely to utilize alkadienes (r = 0.313, P = 0.02), but there was no association between rainfall and the presence of any other specific structural group [alkenes: r = 0.201, P = 0.141; methylalkenes: r = 0.061, P = 0.658; dimethylalkanes: r = 0.044, P = 0.750; trimethylalkanes (r = −0.078, P = 0.571); n = 54 in all cases].
Our comparative analyses suggest that the interspecific diversity in ant CHC profiles has arisen via a gradual mode of evolution. These results contrast with analyses of pheromone evolution in Yponomeuta moths, scolytid bark beetles and Bactrocera fruit flies, which all involved pheromones used in mate attraction and showed evidence of a saltational mode of evolution (Lofstedt et al., 1991; Symonds & Elgar, 2004; Symonds et al., 2009). Rather, the mode of evolution of CHC diversity in ants is more consistent with the gradual mode evident in Drosophila aggregation pheromones (Symonds & Wertheim, 2005). The most likely explanation for why we found gradual rather than saltational evolution of CHCs lies in their function. A saltational mode of evolution is characteristic of pheromones that play a role in mate and species recognition, presumably because these signals reinforce mating isolation (Paterson, 1985). Small changes in pheromone composition at speciation events would increase the likelihood of cross-attraction and hybridization between closely related species. Perhaps the CHCs of ants may evolve gradually because there may not be a need for closely related species to have divergent CHCs.
The interspecific relationship between phylogenetic distance and CHC differences revealed a significant positive association when structural groups of compounds were considered, indicating that closely related species tend to use the same structural groups (and hence, presumably, similar biosynthetic pathways). Surprisingly, this pattern was not evident for individual homologous hydrocarbon series, suggesting that the closely related species in our analysis are as distinct in their CHC profiles as more distantly related species. However, this interpretation needs to be treated with caution.
First, there are likely to be effects caused by the sampling of taxa in the analysis. The conclusions of any comparative analysis are clearly dependent on the sample of species being analysed. In this analysis, the sample of species was broad taxonomically, but sparse (i.e. many instances in which speciose genera are represented by a single species). We related phenotypic change to branch length to elucidate the mode of evolutionary change that has generated interspecific differences. However, because of the sparse taxon sampling, the lengths of branches leading to the tips of the evolutionary tree (i.e. the branches leading to the individual species) are generally very long compared with the internal branches in the tree (see Fig. 1). The shortest terminal branches are found within the genus Temnothorax (6.9 million years), and the average length of terminal branches is 39.5 million years compared with 16.4 million years for internal branches. In other words, it is not surprising that closely related species show considerable differences in CHC profiles, because they have still been separated by considerable stretches of evolutionary time. This sparse sampling probably accounts for the weak, or absent, phylogenetic signal (λ) in four of the five CHC structural group characters. It may also explain why studies that have sampled more intensively but within smaller branches of the ant evolutionary tree (for example examining variation within genera –Martin et al., 2008a,b; Dahbi et al., 1996), generally find fewer CHC differences between species (and more concordance with phylogeny) than analyses among higher taxonomic groups (Martin & Drijfhout, 2009a).
Second, Mantel tests are unable to test for nonlinear relationships, and the data here (see Fig. 3a) suggest a curvilinear relationship between these two variables, with the more-closely related species tending to exhibit a stronger relationship. There was a significant positive relationship between phylogenetic distance and number of differences in species separated by up to 50 million years, but not in species separated by more than 50 million years. This kind of asymptotic relationship where closely related species exhibit a positive trend between the two variables, but more distantly related species do not, is theoretically predicted for simulations of phenotypic change via a gradual mode of evolution (Symonds & Elgar, 2004). Given a limited number of chemical components that species can use, one would expect that, once a certain phylogenetic distance between species has been reached, any further changes in CHC profiles will not actually result in any greater differences in those profiles.
Third, one species in our sample, Cataglyphis rosenhaueri, has a substantially different CHC profile than its congeners. This species provided large outliers to the general trend between phylogenetic distance and CHC differences between closely related species. The unusually distinct CHC profile of C. rosenhaueri has been noted previously by Dahbi et al. (1996). Their explanation that this sizeable difference stems from the importance of CHCs in colony-mate recognition in this species seems unconvincing, as CHCs are used for precisely this purpose in many ant species. Rather, it might be timely to reinvestigate the nature and function of the CHCs in C. rosenhaueri to gain a clearer understanding of its extreme distinctiveness. In addition, investigations into the intrageneric branch length data within this genus may be warranted, as it is has the most uncertain branching information (cf. Methods).
Finally, as with any phylogenetic comparative study, all our results are of course dependent on the reconstruction of phylogeny used as the basis for analysis. Inaccuracies in the topology or branch lengths will influence the results to an unknown degree. Although the phylogeny employed is most probably a reasonable approximation of the true pattern of evolutionary relationships, all our results should be considered with the appropriate caveats in regard to phylogenetic accuracy.
Ants use two major biochemical pathways to produce CHCs: the production of double bonds and methyl branches. Plotting compound groups on our phylogeny revealed that both alkenes and di-methyl alkanes are ancestral traits and that alkadienes and tri-methylalkanes have been gained later in the evolutionary history of ants. Moreover, species will only produce the latter two compound groups if they already produce the more simple structure. Interestingly, we found that as individual groups, alkenes and dimethylalkanes have been lost frequently, but no species has lost both groups. Both alkenes and dimethylalkanes have been identified as important in colony-mate recognition (Martin & Drijfhout, 2009b; Guerrieri et al., 2009). Perhaps ants never lost both compound groups because it would severely impede colony-mate recognition or because both groups function in keeping the CHC mixture flexible and soft, whereas a cuticle covered with alkanes only would be brittle and inflexible.
Environment is likely to play an important role in CHC evolution. Rates of water loss are temperature dependent and positively associated with shorter chain lengths and the presence of double bonds and methyl branching (Gibbs, 1998). We expected species living in warmer climates to have CHC blends with on average longer chain lengths and fewer alkenes and branched alkanes. However, we found no such relationships. One reason is that some of the longer chain compounds were missed in those studies that used conventional GC columns, which do not always detect some of the longer chain compounds in the 36–48 carbon range (Akino, 2006; Martin & Drijfhout, 2009b). Another possible explanation is that species of different latitudes do not differ in the groups of CHCs they use, but only in the relative concentrations of the different HC groups. Wagner et al. (2001) showed that under dryer and warmer conditions, Pogonomyrmex barbatus workers retain the same CHC composition, but they have higher proportions of saturated and unbranched alkanes. A similar pattern may exist across species.
We provide evidence that the shortest CHC in species from environments with higher rainfall was longer than those in dry environments. Additionally, these species were more likely to use alkadienes in their CHC profile. The latter result is consistent with our expectation that species in wetter environments are more likely to use compounds with double bonds (as they are associated with water loss, which should be less of a problem in wet environments). However, the former result contradicts the expectation that species in drier climates would have longer carbon chains, because they were at greater risk of desiccation. We are unable to provide an explanation for this counterintuitive result, but simply note that the relationship between aspects of climate and the evolution of CHC diversity in ants is not straightforward.
Our results add to the growing body of research that elucidates patterns of the evolution of chemical signals and the behavioural and ecological processes driving this evolution (Johansson & Jones, 2007; Symonds & Elgar, 2008). The large diversity of compounds utilized by ants in their CHC profiles seems most likely to have evolved through a gradual process of accumulated changes in compounds and structural groups. While we could not identify general and consistent environmental factors influencing CHC profiles, it remains possible that other aspects of their ecology could drive the evolution of these vital aspects of chemical communication in ants.
We thank David Morgan for discussion of ant CHC chemistry. Patrizia d’Ettorre and an anonymous reviewer provided constructive criticism on the earlier draft of the manuscript. This research was funded by the Australian Research Council (DP0879610 and DP0987360).