The authors have declared that no competing interests exist.
Conceived and designed the experiments: AL MDT BG MDW RRD. Performed the experiments: AL MDT. Analyzed the data: AL MDT. Contributed reagents/materials/analysis tools: AL MDT BG MDW RRD. Wrote the paper: AL MDT RRD.
The evolution of ants (Hymenoptera: Formicidae) is increasingly well-understood due to recent phylogenetic analyses, along with estimates of divergence times and diversification rates. Yet, leading hypotheses regarding the ancestral habitat of ants conflict with new findings that early ant lineages are cryptic and subterranean. Where the ants evolved, in respect to habitat, and how habitat shifts took place over time have not been formally tested. Here, we reconstruct the habitat transitions of crown-group ants through time, focusing on where they nest and forage (in the canopy, litter, or soil). Based on ancestral character reconstructions, we show that in contrast to the current consensus based on verbal arguments that ants evolved in tropical leaf litter, the soil is supported as the ancestral stratum of all ants. We also find subsequent movements up into the litter and, in some cases, into the canopy. Given the global importance of ants, because of their diversity, ecological influence and status as the most successful eusocial lineage on Earth, understanding the early evolution of this lineage provides insight into the factors that made this group so successful today.
The precise habitats and conditions favoring the rise of one of the most successful groups of social organisms on earth, ants, have long been a subject of fascination and study
Today, ants occupy nearly every stratum of terrestrial ecosystems from the deep soil to the highest forest canopy, from the tropics to the subarctic and subantarctic
In the most prominent model of early ant evolution, Wilson and Hölldobler
Here, we statistically assess the predictions of the Dynastic Succession hypothesis as they relate to habitat strata for the first time. We test the hypothesis that ants arose in the leaf litter, as well as the alternative ‘Out of the Ground’ hypothesis that ants evolved in the soil and then, secondarily, colonized the leaf-litter and other strata. To do this we reconstruct ancestral states of habitat strata on the ant tree of life, and examine historical transitions between states to reveal which are likely to be ancestral to all ants and which, if any, appear only more recently. We also consider the relative frequency (and arguably, ease) with which ant lineages have made transitions among different habitat strata.
In order to reconstruct the ancestral traits of ants, we mapped characters onto terminal nodes of the phylogeny of Brady et al.
Each terminal was assigned a state for habitat strata and for biome that reflects the known biology for the genus based on literature records and consultations with experts
Strata (%) | Biome (%) | Any Soil spp. (%) | Any Arboreality (%) | ||||
Soil only | 23.6 | Trop+Subtrop | 35.4 | Any Soil | 39.6 | Any Arboreal | 30.6 |
Soil+Litter | 11.8 | Subtrop | 1.4 | No Soil | 60.4 | No Arboreal | 69.4 |
Litter only | 35.4 | Subtrop+Temp | 3.5 | ||||
Litter+Arboreal | 12.5 | Trop+Subtrop+Temp | 59.7 | ||||
Arboreal only | 14.6 | ||||||
Soil+Litter+Arb | 2.1 |
Relative proportions of the 134 included ant genera occupying each habitat stratum and biome category. Highest proportions are in bold.
All ant genera were coded as being soil dwelling, surface dwelling, arboreal or some combination of these three. We categorized ant genera according to the habitat strata in which they nest and forage, with an emphasis on light availability to distinguish soil-dwelling ants from those dwelling on the surface. The commonly used designation of ‘leaf litter dwelling’ does not clearly differentiate these differences, so we avoid using that terminology here. Instead, we define subterranean genera as those which nest and forage in a primarily low-light environment below the surface stratum. Surface dwelling genera forage on the surface of the soil or in the leaf litter, exposed to open air and direct sun- or moonlight; most of these genera nest underground or in protected spaces aboveground that are contiguous with the soil (e.g. rotting logs). Arboreal genera nest within living or dead tissue of standing trees.
We also used binary coding schemes to code taxa according to whether or not the lineage (genus) possessed any species that were soil dwelling as well as whether any arboreality was present in the lineage. Coding schemes used were 1) any soil-dwelling species vs. none and 2) any arboreal species vs. none.
We examined biome at a coarse scale. Terminals were assigned one of four character states for biome: 1) tropical+subtropical, 2) subtropical, 3) subtropical+temperate, and 4) tropical+subtropical+temperate (no terminals were found to be solely tropical, solely temperate or tropical+temperate). Occurrence data were assessed at the level of political boundaries (primarily country) from the literature and expert opinion
We evaluated the evolution of strata and biome preferences by mapping life history data onto trees that represent consensus views of ant phylogeny
Phylogenetic signal of the characters mapped onto the phylogeny was calculated using Blomberg's K statistic
Ancestral character states were reconstructed for habitat strata using several different approaches in the programs RASP and BayesTraits. All analyses were based on topologies of 500 trees filtered from the posterior distribution of Bayesian analyses done by Brady et al.
Ancestral state reconstruction of ant habitat strata based on phylogram plus 500 trees sampled from the original likelihood distribution of Brady et al.
Trait Reconstructed (Soil, Surface or Arboreal) | Outgroup coding (Soil, Surface or Arboreal) | Posterior Probability |
Soil | Soil | 0.99 |
Soil | Surface | 0.86 |
Soil | Arboreal | 0.70 |
Soil | Soil & Surface & Arboreal | 0.92 |
Soil | Soil & Surface | 0.97 |
Soil | Surface & Arboreal | 0.79 |
Soil | Soil & Arboreal | 0.91 |
Soil | Null (Neither soil, surface nor arboreal) | 0.80 |
Influence of outgroup coding on Ancestral State Reconstructions of the habitat stratum of the root node of ants, with MCMC posterior probabilities provided as support values (performed using the Bayesian Binary Method in RASP using the model F81+G). In all cases soil was reconstructed as the ancestral habitat with the highest proportional likelihood (vs. surface dwelling or arboreal).
Changes of state were summarized over the phylogeny by assessing relative MCMC transition rate coefficients on all 500 trees (as described above) in BayesTraits. ‘Global’ transition rates between characters were optimized in analyses that did not include specific ancestral node reconstructions. Timing of colonizations of different habitat strata was established by calculating a single dated chronogram representing the mean of minimum and maximum trees
In order to address concerns how about outgroup coding affected our results, we performed ancestral state reconstruction in RASP with the outgroup coded to all possible ancestral states, using every possible combination of Soil, Surface and Arboreal as well as null (neither Soil, Surface nor Arboreal).
The idea that that ants evolved as a clade adapted to the tropical leaf-litter and soil and subsequently made colonization events both out of the tropics and out of the litter into forest canopies and deep into the soil was presented by Wilson and Hölldobler
We revisited uncertainties raised about the relationships among the earliest ant lineages
Three depictions of a four-cluster likelihood map visualizing the strength of the phylogenetic signal supporting the placement of lepantillines outside the poneroid + formicoid clade. The top triangle indicates density of individual reconstructions; bottom left and right indicate percentage of points in divided cluster-space, with each tip representing a particular hypothesis (or specific arrangement of taxa). a: outgroups, b: leptanillines, c: poneroids, d: formicoids.
The reconstruction of the ancestral traits of early ants strongly suggests that the most likely habitat stratum for the ancestral ant was within the soil with subsequent, and perhaps multiple, colonizations of the surface stratum and, in some cases, forest canopies (
Trait Reconstructed | MCMC Posterior probability | ML Mean proportional likelihood | |
Soil | 0.695 | 0.631 | |
Soil/Litter | 0.125 | 0.100 | |
Litter only | 0.053 | 0.069 | |
Litter/Arboreal | 0.038 | 0.060 | |
Arboreal | 0.031 | 0.057 | |
All | 0.058 | 0.084 | |
Soil-dwelling | 0.645 | ||
Not soil-dwelling | 0.355 | ||
Arboreal | 0.310 | ||
Not Arboreal | 0.690 |
Ancestral state reconstruction of the habitat stratum of the root node of ants. Support is given as MCMC posterior probabilities and ML mean proportional likelihoods (performed in BayesTraits).
We also assessed the influence of outgroup coding on our results in order to consider the possibility that reconstruction of the ancestral stratum of ants is contingent on the strata assigned to outgroups. We found that all possible outgroup coding schemes supported a subterranean habit for the earliest ants (79–99% probability for subterranean habits) (
After ants emerged from the ground, many secondary transitions occurred, but some transitions were more likely than others.
Rate of transitions among six habitat states summarized across 500 trees sampled from the original likelihood distribution of Brady et al.
The relative rarity of transitions between strictly above- and below-ground habits on the ant tree of life suggests that such transitions are evolutionarily difficult, and likely require more physiological and behavioral changes than are required for transitions within either realm (from surface-dwelling to arboreal, for example). The leaf litter habitat that emerged with the prominence of angiosperm forests may have been influential in easing transitions between the surface strata and soil environment as it incorporated elements of both strata and offered a protected transition zone between the two.
Significant phylogenetic signal for habitat strata in both 6-state and binary coding schemes suggests that species dwelling and foraging below the soil surface are subject to stronger phylogenetic constraint than those that occupy above-ground realms, including the leaf litter and in the trees (6-state: K = 0.4336, p = 0.0001; soil-dwelling: K = 0.3631, p = 0.0001; arboreal: K = 0.1529, p = 0.0004). These results were further supported by assessment using Pagel's lambda statistic, which also showed strong phylogenetic signal (P = 8.5022E-18) for the 6-state habitat strata (⋌ = 0.9323). A likelihood ratio test indicated a significant difference in likelihood scores for the original tree with phylogenetic signal (lnl = −259.5393) as compared to the lamda transformed tree with no phylogenetic signal (lnl = −222.6229). Additionally, significant shifts in diversification rate reported in previous studies correlate with some important habitat shifts, such as the move to partial or complete arboreality within clades (e.g. the tribe Camponotini in Formicinae, the genus
The hypothesis that ants originated in tropical leaf litter
The earliest fossils representing the presumed sister-group to ants, the extinct subfamily Sphecomyrminae, are known from workers and have large eyes, suggesting they were not soil-dwelling, but rather epigaeic or arboreal
Arguments for other lineages that arose in the wet tropics suggest subsequent radiation into drier, cooler realms
Trait | K | P-value |
Biome | 0.090 | 0.1458 |
Strata (6-state) | 0.434 | 0.0001* |
Any soil spp. | 0.363 | 0.0001* |
Any arboreal spp. | 0.153 | 0.0004* |
Phylogenetic signal, presented as Blomberg's K statistic, for habitat traits in ant genera (reconstructed with outgroups). Significance of < 0.05 is designated with an asterisk.
While our data and methods have not offered new insight into the biome in which ants arose, phylogenies are not the only evidence as to the early biomes in which ants lived. The fossil record provides some indication as to the biomes in which early ants have been found, if not necessarily the biome in which the first ants arose. The paleo-reconstructions of the environments in which the earliest ant fossils have been retrieved suggest warm temperate to tropical climates
The conclusions from this study contribute to a broader understanding of the early evolution of ants. Looking toward the future, we anticipate that even more insight will be gained by revisiting these questions with a more complete sample of ant genera, as well as inclusion of targeted outgroups that represent the hymenopteran lineages most closely related to ants.
(DOCX)
(DOCX)
We thank S.G. Brady, T.R. Schultz, B.L. Fisher and P.S. Ward for making original data available for re-analysis. P.S. Ward, M. Borowiec and J.K. Lattke provided input regarding habitat strata of the ant genera. J. Hunt provided input regarding the natural history of outgroup taxa. We thank J. Breinholt for technical assistance with analyses and express our appreciation to B. O'Meara, B.M. Wiegmann, Editor C.S. Moreau and two excellent reviewers for their insightful suggestions that significantly improved the manuscript.