Introduction

The enormous ecological success of ants, bees, and wasps is based on their efficient division of labor among the two female castes [1, 2]. Reproductive queens and non-reproductive workers are well adapted to their respective tasks and they often differ strikingly in size and morphology, with ant queens often being much larger and more long-lived than their workers. Though recent studies have revealed a more or less pronounced influence of genotype on caste differentiation, whether a female larva grows into a worker or a queen is typically controlled by the environment [3], making caste diphenism one of the most conspicuous and abundant cases of phenotypic plasticity.

In particular in ants, the ancestral pattern of queen-worker diphenism has evolved additional, more or less complex modifications. For example, in leaf-cutting ants, Atta, worker-destined larvae may develop into different worker morphs, which behaviorally specialize in leaf-cutting, leaf transportation, defense, or fungus processing etc. [4]. Similarly, queens with reduced wings and simplified thoracic structures may complement the “normal” queens, which at least early in their adult lives are endowed with wings and capable of active flight [5–7].

Other ants show exactly the opposite trend, i.e., a reduction of the number of reaction norms of female larvae. On the on hand, morphological queens have been completely replaced by reproductive workers in a number of Ponerinae [8]. On the other hand, several species of socially parasitic ants have lost the worker caste, i.e., parasite queens rely on host workers from another species to rear the male sexuals and young queens of the parasite species [9–11]. Monophyla, in which queens produce both workers and sexuals in some species and only sexuals in others, might be ideal model systems to elucidate the genomic basis of queen-worker diphenism and to better understand the evolutionary dynamics of caste ratios. Myrmoxenus, a genus famous for the bizarre founding behavior of its queens, provides a particularly promising example of such variation. Myrmoxenus queens sneak into colonies of Temnothorax ants, where they eliminate the resident queen by slowly throttling it with their mandibles for days or weeks. Of the 11 presently recognized species of this genus, at least six are active slave-makers [12]: the queens of these species produce workers, which instead of engaging in the daily maintenance activities of the colony pillage pupae from neighboring Temnothorax nests to replenish or increase the own stock of host workers [13]. M. kraussei is a “degenerate slave-maker,” i.e., worker number is greatly reduced and though slave raids have been elicited in the lab it was thought to be unlikely that they occur in nature [14] (but see [15]). Finally, at least three taxa have completely lost the worker caste and queens produce only female and male sexuals. Variation in caste ratio is associated with additional variation in mating behavior (mating flights vs. intranidal mating, i.e., mating in the nest) and the pattern of brood production [12]. Based on the detailed description and analysis of the various life histories evolutionary pathways linking the various taxa were proposed [12].

Aim of our present study is to determine how the different life histories of Myrmoxenus are interrelated, i.e., whether the worker caste and mating flights were lost once or multiply in evolution. We therefore constructed a phylogeny of this genus based on sequences of mitochondrial CO I / CO II and the nuclear gene wingless and use it to reconstruct the evolution of the various life history traits of this fascinating genus.

Methods

Results

Within Myrmoxenus, GC content of CO I / CO II ranged from 28.8% (M. bernardi) to 30.3% (M. birgitae), with a median of 29.6%, that of wingless from 63.1 (M. tamarae) to 63.9% (M. gordiagini), with a median of 63.7%. The alignment included 328 variable sites and 239 parsimony informative sites in CO I / CO II and 16 variable sites in wingless, of which 8 were parsimony informative. The largest differences were found between T. recedens / T. unifasciatus and M. ravouxi (247 and 251 of 1430 bp of CO I / CO II, 17.3% and 17.5%, 8 differences in 355 bp of wingless, 2.3%). GenBank accession numbers are given in S1 Table.

Both Bayesian (Fig 1) and maximum likelihood analyses (Fig 2) of the concatenated sequences of wingless and COI/COII resulted in similar, robust tree topologies with Bayesian posterior probabilities of 1 for several of the more basal major nodes. M. gordiagini consistently formed a well-supported clade that split early from the other Myrmoxenus lineages. Of the remaining taxa, M. stumperi and M. bernardi are well-separated from the other lineages. The other taxa form two clusters, A) with M. kraussei and M. algerianus, and B) with M. ravouxi, M. tamarae and the three workerless species M. adlerzi, M. birgitae, and M. corsicus, with a few inconsistent placements. For example, specimens from Algeria and Italy, by morphology assigned to M. kraussei and M. corsicus, form the outgroup to the two branches A and B, and one sample referred to as M. kraussei from Venaco clusters with clade A. Because of the morphological similarities among the different taxa we cannot exclude that this presents incorrect determination and that some of the “misplaced” specimens in fact present undescribed species. The placement of material assigned to M. algerianus from Mts de Belezma, Algeria, as outgroup to other Epimyrma is presumably erroneous as the sequence contained a number of ambiguous positions.

Download:

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

Fig 1. Bayesian tree of the socially parasitic ant genus Myrmoxenus inferred from sequences of the nuclear gene wingless and the mitochondrial gene CO I / CO II.

Bayesian posterior probabilities (as percentages) are given at the nodes. Workerless species are given in red.

https://doi.org/10.1371/journal.pone.0131023.g001

Download:

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

Fig 2. Maximum likelihood tree of the socially parasitic ant genus Myrmoxenus inferred from sequences of the nuclear gene wingless and the mitochondrial gene CO I / CO II.

Figures at the nodes represent the percentage of replicate trees with a particular branching pattern (only > 50%). Workerless species are given in red.

https://doi.org/10.1371/journal.pone.0131023.g002

A phylogeny based only on the nuclear gene wingless did not help to solve these apparent “misplacements”. While it supported M. gordiagini, M. tamarae, M. bernardi, and M. stumperi as separate lineages with Bayesian probabilities of > 0.9, it did not distinguish between M. ravouxi and M. kraussei. We could not amplify wingless in the workerless species.

Discussion

Our phylogeny of the socially parasitic ant genus Myrmoxenus reveals the existence of two well-supported lineages, one with M. gordiagini and the other with the remaining taxa. This matches the original taxonomic treatment of all species other than M. gordiagini as members of a separate genus, Epimyrma. The two genera were merged first as Epimyrma [32] and later, because of priority, as Myrmoxenus [33]. Within the “Epimyrma”-clade, one or two species (M. stumperi, M. bernardi) are rather well supported. The remaining lineage apparently has undergone a relatively recent range expansion and radiation, which makes it difficult to clearly resolve its phylogeny (M. adlerzi, M. algerianus, M. birgitae, M. corsicus, M. kraussei, M. ravouxi, M. tamarae). Recent divergence is also supported by the observation that several taxa of “Epimyrma,” though strictly isolated in nature, readily hybridize and produce fertile offspring in the lab [34, 35]. In contrast, cross-breeding of M. kraussei and M. gordiagini did not yield any progeny [35] (no mating experiments were done with M. stumperi). The specialization of most taxa on particular host species, together with the frequent replacement of mating flights by sib-mating in the nest might promote speciation and prevents natural hybridization (see also [36]]. The observed “incorrect” placements of some taxa (e.g., M. “corsicus” from Abruzzi or M. “kraussei” from Azrou) might reflect incomplete lineage sorting in this rapidly diverging genus or the existence of previously not recognized novel species.

Despite of the shallow branching pattern and the predominance of mtDNA data in the analyses, the trees allow the tentative reconstruction of the evolution of the different life histories and colony demographies of Myrmoxenus. The workerless taxa obviously do not form a monophyletic cluster, but workers were lost convergently in M. adlerzi within M. ravouxi, in M. birgitae and M. corsicus as sister taxon of M. ravouxi, and in M. “corsicus” from Abruzzi as a third separate lineage. A further independent origin of workerlessness is suggested by the observation that worker number varies tremendously within M. kraussei, with one population from Crete having lost workers [37] (no molecular data available). The stable production of queens from all female brood regardless of social and environmental conditions suggests an extensive loss of phenotypic plasticity. The sporadic occurrence of workers in naturally workerless species in the lab (e.g., one worker in a total of 745 female offspring of M. adlerzi, [38]) indicates the retention of cryptic plasticity rather than genetic accommodation. Though numerous socially parasitic ants have lost the worker caste, the close relatedness among Myrmoxenus species with and without workers and the large variation of queen/worker ratios even within single species make this genus a suitable model to investigate the functional genomics of caste differentiation in ants.

One might assume that workerless parasites with mating in the nest evolved from slave-maker species, which already had switched from nuptial flights to mating in the nest, such as M. kraussei or M. bernardi [35]. Our phylogeny does not support this obvious view. Instead, workerless M. adlerzi, M. birgitae, and M. corsicus, all with intranidal mating, appear to have evolved within the clade of M. ravouxi, a widespread species in which mating is thought to occur during nuptial flights. An excess of homozygous genotypes in several populations of M. ravouxi nevertheless suggests a mixed mating system with occasional sib-mating [22]. Mating behavior therefore appears to be a rather labile trait and nuptial flights that guarantee outbreeding have been convergently replaced by exclusive intranidal mating among close relatives in both species with and without workers. The driving force of this change of the mating location is the difficulty of finding mates when population sizes are low (e.g., [12, 39]). Queens, which mated in their natal nests, disperse on foot and usurp host colonies nearby. Because of local resource competition this selects against slave-raids and promotes the loss of workers [12].

Inbreeding is particularly detrimental for species with haplodiploidy and single-locus complementary sex determination, the typical mode of sex determination in the social Hymenoptera, because it results in the production of inviable or sterile diploid males. Breeding experiments in Myrmoxenus with intranidal mating did not yield diploid males, suggesting that the convergent transition to mating in the nest is associated with a change in the mechanism of sex determination [12]. This again makes Myrmoxenus an interesting system to study the evolution of sex determination in social Hymenoptera, which appears to be more variable than occasionally assumed (see also [40, 41].

Finally, our phylogeny indicates that the production of sexuals from non-overwintering brood (“rapid brood”) evolved convergently in M. kraussei, M. corsicus, and M. adlerzi, while sexual larvae of the other taxa hibernate in the natal nest before pupation [12]. Rapid brood may be an adaptation to worker loss, because the absence of slave-raids limits the life expectancy of the parasite colony to that of the host workers present during usurpation. Furthermore, it allows a very quick exploitation of a local patch of host nests.

At present it remains unclear which ecological factors triggered the evolution of the loss of workers, but climate might be important. Most workerless taxa have been collected in very dry and hot places with sparse vegetation cover (e.g., [37, 42], while the active slave-makers tend to occur either in more temperate woodland, alpine pastures, or, in the Mediterranean, in shady places in deciduous forests (e.g., [12, 43]. Slave raids are probably difficult in summer when they are most profitable because of the large numbers of worker pupae present in Temnothorax colonies. Host nests might be more difficult to locate because many host species protect themselves from desiccation and heat by moving deeper down in rock crevices or other nest materials. The observation that neighboring populations of M. kraussei greatly differ in average worker number indicates, however, that additional factors, e.g., microclimate or evolutionary history, might also play a role [44].

Supporting Information

Acknowledgments

We thank Jeanette Hulak, Nana Gratiashvili, Holger Martz, Diethe Ortius and Andreas Schulz for providing samples of Myrmoxenus.