Honest and dishonest communication in social Hymenoptera

Imagine a world in constant darkness and silence, but full of odours and tastes. You are alarmed, soothed, and told what to do by scent, and by their perfumes you recognize the standing of those you interact with, what role they play in the society, and whether or not they belong to your family or work group. This is life in an ant society, a superorganism in which all communication is honest, free of deceit and for the good of the group –or is it?

In most communication, senders would benefit from manipulating the receivers and thus receivers should respond only to fraud-resistant signals(e.g. Maynard Smith and Harper, 1985; Zahavi and Zahavi, 1997). In contrast, communication in ants, honeybees, and other social insects has long been assumed to be utterly honest. Insect societies are characterized by a sophisticated division of labour between reproductive queens and non-reproductive workers, which forage for food, defend the nest and nurse the queen's offspring (Hölldobler and Wilson, 1990; Hölldobler and Wilson, 2009). Kin selection theory(Hamilton, 1964) states that,because insect societies are usually more or less complex families in which all individuals are related, workers maximize their fitness indirectly by helping queens to reproduce.

The magnitude of the workers' indirect fitness gains depends on the smooth and efficient functioning of the whole society. Dishonesty about the location of food sources or turning in false alarms disrupts this balance and should thus be non-adaptive. Therefore, intra-specific communication in insect societies might be less shaped by co-evolutionary arms races between senders and receivers but instead characterized by parsimonious and cost-efficient signals stabilized by selection on the level of the society. Ants in particular have been referred to as walking chemical factories and produce hundreds of chemicals in dozens of glands (e.g. Hölldobler and Wilson,1990). Experimenters have been able to elicit the complete behavioural response of an ant with secretions from glands dissected out of the insect body or with synthetic compounds previously identified in the contents of these glands (reviewed by Hölldobler, 1995),suggesting that communication may indeed rely on relatively simple signals.

The life history of insect societies also provides situations in which the interests of sender and receiver are not aligned, such as in interspecific parasitism, where the parasite must sneak into the host nest. The complex pattern of hydrocarbons present on the cuticle of ants and other social insects provides the cues that enable recognition(Lenoir et al., 1999), so slave-makers and other social parasites mimic the host cuticular profile or attempt to remain chemically invisible through the lack of cuticular hydrocarbons (Lenoir et al.,2001; Lorenzi,2006; Brandt et al.,2005). Social parasites are sufficiently rare to be ahead in the arms race with their hosts and thus are able to manipulate them by exploiting an honest signalling system (nestmate recognition) in a dishonest way(Nash and Boomsma, 2008).

In an intra-specific context, one case in which rarity has allowed deception is that of winged males of Cardiocondyla obscurior, which mimic the scent of female sexuals to avoid lethal attacks by wingless fighter males (Cremer et al., 2002). Here, chemical mimicry may be evolutionarily stable because the winged cheater males are produced only when the colony is under stress or poor food conditions (Cremer and Heinze,2003). Fighter males usually only encounter virgin females sexuals, with which mating is advantageous.

More generally, kin selection theory(Hamilton, 1964) predicts that individual nestmates might benefit from being dishonest in the context of the reproductive division of labor. Haplodiploid sex determination results in relatedness asymmetries and introduces conflict among nestmates, e.g. over the partitioning of reproduction and the origin of young males, and thereby creates conditions in which deceitful communication may be advantageous. In the evolution of sociality, workers have often lost the capability of mating and cannot produce female offspring from fertilized eggs, but in many species workers have retained their ovaries. They are thus capable of rearing sons from unfertilized, haploid eggs, to which they are more closely related than to the queen's male offspring (e.g. Bourke and Franks, 1995). In addition, workers in societies with a single, singly mated queen (monogyny, monandry) are more closely related to the sons of other workers than to their brothers. In this situation, workers could in principle maximize their inclusive fitness by raising the queen's female sexual offspring and rearing nephews (worker-produced males) instead of brothers.

When colonies contain a multiply mated queen (polyandry) or multiple queens(polygyny), the average relatedness of workers to worker-produced males may be lower than to queen-produced males. Workers may then inhibit reproduction by other workers through egg eating and aggression, leading to a complex network of mutual policing and punishment(Ratnieks, 1988; Heinze, 2004; Ratnieks et al., 2006). Similarly, worker reproduction may be inhibited if it is associated with costs that lower the total productivity of the society, e.g. due to workers competing for egg-laying rights and neglecting daily maintenance tasks(Ratnieks, 1988).

Nevertheless, in particular in monogynous, monandrous species with limited queen–worker dimorphism, workers might benefit from producing their own sons. Workers readily begin to lay eggs once the queen has been experimentally removed, a situation that in nature would occur with the death of the queen(e.g. Bourke, 1988), but genetic data reveal that workers reproduce in the presence of the queen only in a minority of species (e.g. Hammond and Keller, 2004; Heinze,2004).

Worker sterility in queenright colonies has occasionally been explained as queen inhibition, i.e. queens using pheromones to manipulate workers into sterility against their interests [`queen control'(Dejean and Passera, 1974; Passera, 1980; Hölldobler and Wilson,1983)]. However, such dishonest queen control has been argued to be evolutionarily unstable because workers would be selected to ignore it and the ensuing arms race between queens and workers would decrease overall colony performance and be selected against(Keller and Nonacs, 1993). Furthermore, the observation that honeybee workers `deliberately' distribute queen pheromones through the hive suggested an alternative explanation in agreement with honest signalling theory(Seeley, 1985; Woyciechowski and Lomnicki,1987): queen pheromones reliably signal queen presence and reproductive potential [`queen signal'(Keller and Nonacs, 1993)] and thus are honest indicators of the workers' indirect fitness benefits, to which workers react in their own interest by self-restraint. In this framework, the queen signal could be interpreted as a `fertility signal'(Seeley, 1985; Gobin et al., 1999; Ortius and Heinze, 1999; Monnin and Peeters, 1999). Zahavi and Zahavi proposed interpreting queen pheromones as handicaps(Zahavi and Zahavi, 1997). The honesty of such a signal would then be assured by its toxicity and/or maintenance costs. Individuals not showing the qualities indicated by the signal would not be capable of producing it because of constraints they cannot bear.

The function and validity of a signal can often be derived from observed signal design (e.g. Maynard Smith and Harper, 1985; Tibbetts and Dale, 2007). Dishonest and honest signals are therefore expected to differ in a number of traits. Keller and Nonacs (Keller and Nonacs, 1993) suggested that if the queen signal were dishonest and manipulative (i) its inhibitory effects would not be correlated with the queen's egg-laying ability (fertility); (ii) unmated females that produce only males would have the same inhibitory effect as queens producing males and females, and (iii) worker reproduction (e.g. ovary development) would decrease with increasing number of fecund queens in a colony, as the concentration of pheromones and thus their inhibitive effect would be higher in polygynous colonies. In the 16 years since the publication of this seminal paper,numerous studies have aimed at elucidating the nature of queen pheromones. In the following section we investigate whether the predictions by Keller and Nonacs (Keller and Nonacs,1993) were met. We then proceed to examine whether the variation of chemicals presumed to be queen pheromones shows traces of an arms race, as expected from dishonest manipulation(Keller and Nonacs, 1993), or whether they are associated with costs, as expected for honest signals (e.g. Maynard Smith and Harper, 1985; Zahavi and Zahavi, 1997).

Reproductives and non-reproductives of ants, bees and wasps usually differ conspicuously in the quantitative and qualitative composition of their cuticular hydrocarbons and/or gland contents, which seems to corroborate the hypothesis that these chemicals honestly signal fertility, i.e. egg-laying rates (reviewed by Monnin,2006). For example, in the ant Pachycondyla inversa,3,11-dimethylheptacosane is the major constituent of the cuticular blend of queens, and is preferentially detected by workers' antennae(d'Ettorre et al., 2004);mated, fertile workers (`gamergates') of the queenless ant Dinoponera quadriceps are characterized by a high percentage of 9-hentriacontene(Peeters et al., 1999); and 13,23-dimethylheptatriacontane is specific to egg layers of Harpegnathos saltator (Liebig et al.,2000). This widespread positive association between certain chemical compounds and fertility appears to disagree with the first prediction of Keller and Nonacs (Keller and Nonacs,1993) for the queen control scenario and instead to be in accordance with the honest queen signal hypothesis. At closer inspection,however, the picture becomes less clear.

First, we must caution that the association between the preponderance of particular chemical compounds and fecundity is as yet mostly correlative. The few bioassays with synthesized `queen substances' have failed, with two notable exceptions: in honeybees, β-oxodecenoic acid (9-ODA) from the mandibular gland of queens directly reduces juvenile hormone biosynthesis and inhibits ovarian development in worker bees(Kaatz et al., 1992). In the ant Aphaenogaster cockerelli, pentacosane is the most pronounced reproductive-specific hydrocarbon. When experimentally applied onto the cuticle of non-fertile, thus `cheating', workers this substance elicited aggression (Smith et al.,2009) (see also below). In other cases, however, additional components might be involved in the regulation of reproduction, in particular behavioural traits or substances of glandular origin(Le Conte and Hefetz,2008).

Second, cuticular hydrocarbons thought to be associated with fertility are occasionally produced before egg laying starts(Cuvillier-Hot et al., 2005). Furthermore, workers of Temnothorax nylanderi do not lay eggs in the presence of a queen that has stopped reproducing when preparing for hibernation, but they do so when the queen is removed (J.H., unpublished data). This suggests that queen pheromones are linked to physiological changes preceding the maturation of eggs and more reliably reflect future reproductive potential than actual fecundity.

Third, and most importantly, the chemical profiles of fertile mated and unmated workers and queens are often similar, which following Keller and Nonacs (Keller and Nonacs,1993) is suggestive of manipulative, i.e. dishonest, queen control. For example, the relative proportion of 9-ODA in the mandibular gland secretion was lower in highly fecund, mated honeybee queens than in both sterile and egg-laying virgin queens, and a multivariate analysis of all compounds did not separate fertile unmated and mated queens(Strauss et al., 2008). Likewise, the proportion of 3,11-dimethylheptacosane increased with the fecundity of both mated queens and unmated workers of P. inversa(d'Ettorre et al., 2004), and 9-hentriacontene also increased when unmated workers of D. quadricepsachieved social and reproductive dominance(Peeters et al., 1999). Similar results have been found in numerous other ants and also paper wasps(Sledge et al., 2001).

The assumption that the chemical similarity between virgin and mated egg layers indicates dishonest manipulation, though logical, is probably unrealistic, as in most natural situations unmated females do not reproduce in the presence of a fertile mated individual (e.g. Hammond and Keller, 2004). The individual emitting large quantities of queen pheromones is usually a mated female, and in most cases it is the mother of the workers. Unmated workers or virgin queens that attempt to become fertile in its presence will first show a distinctly weaker chemical signature, which, as we argue below, is associated with considerable costs. The situation is different once colonies are orphaned.

The remaining prediction of Keller and Nonacs(Keller and Nonacs, 1993)regarding the queen control hypothesis, i.e. that queen number and worker reproduction should be negatively related, is difficult to evaluate with available data. As yet, worker fertility has only rarely been compared between monogynous and polygynous colonies of the same species (but see Helanterä and Sundström,2007). However, even if future studies documented lower worker egg-laying rates in multi-queen colonies, this would not necessarily reflect pheromonal queen control. Instead, the same outcome is expected from worker policing (Ratnieks, 1988; Ratnieks et al., 2006).

In the course of evolution, manipulation is expected to lead to counter-manipulation and/or resistance to manipulation. Dishonest queen pheromones would thus evolve to more and more complex and heterogeneous mixtures, as is the case in manipulative seminal fluids in species with sexual conflict (Swanson et al.,2001), or cycle between different compounds(Foster et al., 2000). With honest signals, no such heterogeneity and complexity is expected. The large variability of substances supposed to act as queen pheromones might indicate that they change quickly. At present, no obvious trends have been identified concerning the chemical and physical properties of such compounds. In some species, the bouquets of reproductives and non-reproductives differ in long-chained hydrocarbons, while in other species, shorter-chained and branched hydrocarbons are more dominant(Endler et al., 2004; de Biseau et al., 2004; Monnin, 2006; van Zweden et al., 2009).

Several recent studies have provided data about the evolutionary stability of fertility signals, but the results are ambiguous. In paper wasps,substances that differentiated between reproductives and non-reproductives differed even between populations (Dapporto et al., 2004), suggesting relatively rapid evolution. Temnothorax ant workers had more developed ovaries in the presence of a queen from distantly related species than when kept with a more closely related or conspecific queen, suggesting a relatively slow decrease of pheromone effect with increasing phylogenetic distance (Elisabeth Brunner,unpublished). In contrast, workers of Temnothorax species rarely lay eggs in the presence of queens of certain socially parasitic ants (J.H.,unpublished), to which they are even less closely related(Beibl et al., 2005), i.e. the cross-species efficiency of queen pheromones does not vary with phylogenetic distance as expected. Social parasites are a special case, as they might have evolved manipulative, dishonest compounds and also use physical intimidation(Heinze et al., 1994). However, a similar mismatch between effectiveness and phylogeny has been observed in honeybee egg-marking substances, which allow worker policing by egg eating (Nanork et al.,2007).

The speed of evolution cannot easily be calibrated by comparison with other chemical characters, such as alarm pheromones or trail pheromones, because those might be adapted to competitors or chemical background `noise'. Hence,the meaning of differences in the nature of queen pheromones between populations and species remains unclear.

According to models of signal evolution, signals need to be unforgeable to be honest. Thus, the costs associated with the production or maintenance of queen pheromones should be too high for low quality individuals to bear them. Pheromones are often produced in minute quantities, and little is known about how costly the production is to the sender (e.g. Dicke and Sabelis, 1992). In Drosophila males, life span was found to be negatively associated with the investment in sexual pheromones, suggesting considerable production costs (Johannson et al.,2005). In contrast, pheromone attractiveness of male Tenebrio molitor beetles was actually higher following an immune challenge,implying that the signal is cheap enough that it can be produced even during a costly immune response (Sadd et al.,2006). However, such higher investment in sex might also be a reaction to threatened survival.

Almost nothing is known about the energetics of pheromone production in social insects, and as the traditional trade-off between longevity and reproductive efforts is turned upside down in perennial social insects(Heinze and Schrempf, 2008) it will be difficult to determine how costly it is to produce queen signals. Yet,considerable costs appear to be born by those individuals that, in the presence of one or several fully fertile reproductives, change their chemical bouquet as they attempt to usurp the position of the reproductive or to start laying eggs. These costs are evident from numerous experiments in which workers were allowed to become reproductive by temporarily splitting a society into a queenless and a queenright fragment. Upon re-unification of the two fragments, those workers that have become fertile in the queenless part are usually policed, i.e. attacked, expelled or killed by nestmates(Ratnieks et al., 2006).

Worker policing requires that the chemical profiles of new reproductives differ to a certain degree from those of non-reproductives and established egg layers. Indeed, workers of Camponotus floridanus that lay eggs at a very low rate are not aggressed (though their eggs are eaten), perhaps because their chemical profiles are too similar to those of sterile workers(Endler et al., 2007). In contrast, workers of the clonal ant Platythyrea punctata, which had started to lay eggs during a short separation from the fully fertile reproductive, were violently attacked by other workers when returned into the nest, while workers that had become fully fertile during a longer separation were not. Instead, such `established new reproductives' were attacked only by the old reproductive (Hartmann et al.,2003; Hartmann et al.,2005). Whether new reproductives were policed by workers was clearly reflected in their hydrocarbon profiles: they were not attacked when it matched those of infertile or fully fertile females(Hartmann et al., 2005). Recent research on A. cockerelli gives direct evidence of the costs of wrongfully bearing the queen bouquet: application of a synthetic compound typical for fertile queens to non-fertile workers elicited worker policing(Smith et al., 2009).

Worker policing is not restricted to unmated workers or virgin queens,which belatedly begin laying haploid eggs in the presence of a fertile reproductive. Mated queens in multi-queen societies, which due to experimental manipulations, queen–queen antagonism, or other causes begin to develop their ovaries later than others, may also be fiercely policed by workers (e.g. Lipski et al., 1992; Ortius and Heinze, 1999). Similarly, when queens cooperatively start a new society, workers may eliminate all but one queen, presumably the most fertile (Balas, 1995), and when queens are introduced into monogynous societies, workers also eliminate the less fertile individual (Fletcher and Blum, 1983). This suggests that individuals that develop their ovaries in the presence of a fertile nestmate and consequently change their hydrocarbon profile risk being policed regardless of caste and mating status.

Worker policing against those individuals, unmated or mated, that belatedly develop their ovaries and change their chemical signatures in the presence of established reproductives might therefore constitute the costs of unrightfully producing queen pheromones – a nice exemplification of the motto incorrectly ascribed to former Soviet leader Gorbachev: `those who are late will be punished by life'. We note here that it is obviously difficult or impossible to become fully fertile without signalling the changed reproductive status, even if individuals might benefit from hiding it. Whether this also holds for species in which freshly mated queens are adopted into mature nests and develop their ovaries in the presence of one or several fertile queens is presently unclear. It is suggestive that in many temperate and boreal species,young queens rarely lay eggs before hibernation but develop their ovaries simultaneously with the old queens in spring. In many polygynous species from tropical biomes, workers lack ovaries, and policing is therefore not expected.

The association between queen substances and fertility appears to be often much less perfect than expected from honest signalling theory, but there is also little evidence that workers are directly manipulated by dishonest signals. Instead, in agreement with recent ideas about enforced altruism(Ratnieks and Wenseleers,2008), we believe that workers refrain from egg laying even when they might individually benefit from reproducing because of the risk of being attacked and policed in the presence of a fertile reproductive. While an individual worker might potentially increase its inclusive fitness by producing sons from its haploid eggs, costs associated with worker egg laying may negatively affect the overall performance of the society and thus decrease the inclusive fitness of the average worker.

Female reproductives of several ant species may establish their reproductive monopoly through aggressive displays, such as antennal boxing and biting. Once they have become fully fertile, overt aggression disappears and its function is taken over by chemical signalling (e.g. Heinze and Smith, 1990; Ortius and Heinze, 1999; Cuvillier-Hot et al., 2005; d'Ettorre et al., 2004) and worker policing. Queen pheromones essentially lead to the same outcome as physical intimidation and the switch to pheromones might simply be a consequence of group size. A single queen can keep 10 or 20 workers in a small society in check by antennal boxing – similar to the foreman in a small family enterprise who personally directs the activities of all workers. In larger nests, chemical signals directly emitted by the queen or, in very large societies with multiple nest sites, chemical signals transferred by messenger workers or the brood may serve the same function, analogous to circulars or mass mails sent to employees in medium-sized or large international companies.

Policing might also explain why queen pheromones do not more accurately indicate the fertility and mating status of reproductives. Though fertile workers and virgin queens can in principle produce bouquets similar to that of fertile, mated queens, the chance of mistaking an unmated individual for a fertile, mated queen is minimal in most natural situations. Due to policing against unmated females that begin to develop their ovaries in the presence of already fully fertile, mated queens, the individual that produces queen pheromones is likely to be the mother of the workers.

What is to be done in the future? We certainly need more information about the chemical nature, physical properties and mode of action of queen pheromones. In addition, better bioassays are needed to identify the active chemicals in the complex pheromone blends of reproductives. We expect honest signals to be relatively simple, as complex signals would indicate the occurrence of an arms race between sender and receiver, and to be intrinsically linked to present or imminent ovarian development.