Social transmission of information about a mutualist via trophallaxis in ant colonies
Abstract
Partner discrimination is crucial in mutualistic interactions between organisms to counteract cheating by the partner. Trophobiosis between ants and aphids is a model system of such mutualistic interaction. To establish and maintain the mutualistic association, ants need to correctly discriminate mutualistic aphids. However, the mechanism by which ants recognize aphids as their partners is poorly understood, despite its ecological and evolutionary importance. Here, we show for the first time the evidence that interaction with nest-mates that have tended aphids (Aphis craccivora) allows ants (Tetramorium tsushimae) to learn to recognize the aphid species as their partner. When ants had previously tended aphids, they moderated their aggressiveness towards aphids. More importantly, ants that had interacted with aphid-experienced nest-mates also reduced their aggressiveness towards aphids, even though they had never directly experienced them, indicating that aphid information was transmitted from aphid-experienced ants to inexperienced ants. Furthermore, inhibition of mouth-to-mouth contact (trophallaxis) from aphid-experienced ants to inexperienced ants by providing the inexperienced ants with artificial honeydew solution caused the inexperienced ants to become aggressive towards aphids. These results, with further supporting data, strongly suggest that ants transfer information on their mutualists during trophallactic interactions.
1. Introduction
Many organisms engage in reciprocally beneficial relationships with other species. Interspecific mutualisms essentially involve conflict between the interacting organisms. When the investments offered to the partners involve fitness costs, natural selection should favour the use of cheating strategies by individuals, who can thus obtain mutualistic benefits while paying fewer costs by providing fewer commodities to their partners [1–3]. In these selfish interactions, it is likely necessary for mutualists to have accurate partner-recognition systems that associate cues from the partners with rewards. Thus, learning-dependent cognitive systems might help organisms in regulating mutualistic associations to minimize being cheated [4,5].
In fact, some animal species are considered to rely on associative learning to recognize their mutualists [6–9]. In such learning processes, animals can use not only their own private information from personal experience, but also social information from other individuals [10–12]. By using this social information as a strategic option in the context of mutualism, animals can potentially modulate their cooperative behaviour more precisely and can thus choose relevant partners from among potential candidates [4,13]. Therefore, to understand coordinated mutualism between organisms it is important to understand how social information is used in the recognition of mutualistic partners.
Interactions between ants and trophobiont insects such as aphids are well-known examples of mutualism: aphids produce honeydew and provide it to ants as an important food source; in return, ants offer the aphids beneficial services such as a defence [14–16]. In these associations, ants have to discriminate their partners correctly. Recent studies have shown that ants use cuticular chemicals of mutualists to recognize them, and that ants learn to associate their mutualists' chemicals with honeydew [7,8,17]. Learning-dependent recognition is considered to be an adaptive system for ants in maintaining mutualistic associations with aphids, because ants can reduce the risk of being exploited by cheaters that do not offer them honeydew but obtain the benefits of ant attendance [7]. However, the quality and quantity of honeydew produced by aphids are known to be changeable and are affected by various biotic and abiotic factors [18–21]. Because the information that ants acquire from personal experience only could be unreliable or outdated in some situations, ants might be able to more quickly discern the appropriate partners by using not only personal information but also social information supplied by nest-mate workers.
Our aim here was to determine whether ants receive information about aphids through interaction with their nest-mates, and to reveal how ants transmit information on aphids to their nest-mates. We explored these issues by using the ant Tetramorium tsushimae, the cowpea aphid, Aphis craccivora, and the pea aphid, Acyrthosiphon pisum, in laboratory experiments. Tetramorium tsushimae ants are known to tend A. craccivora aphids; prior experience tending the aphids reduces the ants' aggressive reactions to them and their cuticular hydrocarbons, indicating that the ants learn to recognize the aphid species as their partners and use their cuticular hydrocarbons to discriminate mutualists [7]. Conversely, the ants never tend A. pisum and behave aggressively towards this aphid species and their hydrocarbons [7]. First, we examined whether ants moderated their aggressiveness towards these aphid species upon interacting with nest-mates that had previously tended A. craccivora aphids. We then focused on food-exchange by regurgitation (trophallaxis) between ants as a means of transferring information, because we frequently observed this behaviour when aphid-inexperienced ants interacted with nest-mates that had tended aphids. Trophallaxis is the distribution of food resources among colony members through mouth-to-mouth feeding, and it is widely developed in social insects such as ants [22]. To investigate the role of trophallaxis in the transmission of partner information, we examined the aggressiveness of aphid-inexperienced ants towards A. craccivora aphids when trophallaxis from aphid-experienced ants to inexperienced ants was experimentally prevented.
2. Study organisms
We collected three colonies of the ant T. tsushimae, each of which contained several queens, 150–300 workers and some brood, in the city of Matsudo, Japan, in June 2010. The ants were supplied with mealworms, Tenebrio molitor, as a food source and were allowed to colonize glass test tubes (16 mm diameter, 150 mm length) in plastic cases, the inside walls of which were coated with Fluon (Asahi Glass Co., Ltd, Tokyo, Japan) to prevent the ants from escaping. Between 1000 and 3000 workers were present in each colony at the time of the experiments. Because the ant colonies had been isolated from honeydew-producing insects for more than 4 years in the laboratory, we assumed that the ant workers used in the experiments had never previously encountered aphids [7].
An apterous adult female of each of two aphid species, A. craccivora and A. pisum, was collected in Matsudo in July 2011 and April 2017, respectively. Colonies of each aphid species were established from the females and maintained on broad bean plants, Vicia faba, grown from seed in plastic pots (90 mm diameter, 75 mm depth) containing soil. We used broad bean seedlings 20–25 cm high for the experiments. The insects and plants were maintained in a climate-controlled room (24 ± 2°C; 16 L: 8 D).
3. Experiment 1: social transmission of mutualist information
To test whether interaction with nest-mates that had experience with A. craccivora aphids influenced the aggressiveness of ants towards aphids, we established ‘signaller’ and ‘receiver’ ants; signallers either had directly experienced the aphids or had not, and receivers interacted with the signallers (figure 1). To establish signallers that had experienced aphids, 20 apterous adult A. craccivora aphids that had been randomly selected from the maintained colony were placed on a leaf disc (12 mm diameter) on moisten cotton wool (13 × 13 mm) in a rectangular Petri dish (87 × 57 mm, 19 mm deep) lined with Fluon. The leaf discs were made from the top or second leaves of clean V. faba plants. We introduced 20 ant workers randomly selected from the maintained colony into the dish with the aphids. Under these conditions, the ants could freely visit the aphids on the leaf disc [7]. The ants were thus provided with honeydew from the aphids and sometimes preyed upon them. Because it was possible that differences in the hunger levels of the ants would influence their aggressiveness, as control treatments we established two other sets of conditioned signallers that had not experienced aphids: one was provided with artificial honeydew and with chopped mealworm as animal matter (satiated control), and the other was not provided with any resource (starved control). In the case of the satiated control, we presented cotton wool (10 × 10 mm) soaked in 200 µl of sugar solution, which was designed to mimic the honeydew of A. craccivora (sugar concentrations: 5 µg µl–1 fructose, 5 µg µl–1 glucose, 2.5 µg µl–1 sucrose and 35 µg µl–1 melezitose; see [23]), and 25 mg chopped mealworm to 20 ant workers in a Petri dish with a leaf disc without aphids. In the starved control, 20 ant workers in a Petri dish with a leaf disc were not provided with any food resources. While the three groups of signallers were being established, three groups of 10 receiver ants randomly selected from the same colony as the signallers were marked with white paint on the thorax under hypothermic anaesthesia. After recovery from the anaesthesia, each group of receivers was introduced to Petri dishes that did not contain any resources.
Figure 1. Experimental design of Experiment 1. Signaller ants are represented as black figures and receiver ants as white figures. Arrows indicate that subjects were transferred into Petri dishes.
Signallers and receivers were left for 8 h under each set of conditions during the photophase. The receivers were then brought together with each signaller group in new Petri dishes. After 2 h, we transferred 10 randomly selected signallers from the mixed group of 20 signallers and 10 receivers into a new Petri dish. At the same time, all of the receivers were transferred into another Petri dish. Fifteen to 20 minutes after the transfer, an apterous adult of A. craccivora or A. pisum was introduced into each dish containing signallers or receivers from the maintained aphid colony, and the ants' behaviour towards the aphid was observed. In this study, a single tester who was blind to the treatment group directly observed the behaviour for all experiments. The observer recorded the frequency of contacts with, and attacks on, the aphid by the ants for 5 min. The contacts of the ants with the aphids were distinguished as attacks or non-attacks. We recorded a contact as an attack when an ant bit the aphid and as a non-attack when an ant just touched the aphid. We performed the experiments 14 times for each treatment and each aphid species, using new ants and aphids for each replicate. In the experiment for A. pisum, we omitted ‘satiated control’ because there was no difference in ant aggressiveness towards A. craccivora aphids regardless of the hunger level (see Results). Insects used in the experiments were never replaced in the maintained colonies. All experimental procedures were conducted in a climate-controlled room (23 ± 2°C, 16 L: 8 D).
In each experiment assay data on the signallers and receivers towards two aphid species were analysed by using a generalized linear mixed model (GLMM) with a binomial error distribution and logit link function. A model was constructed by using attack probability (i.e. incidence of attacks versus incidence of non-attacks) as a response variable, treatment (i.e. aphid-experienced, satiated control and starved control) as a fixed effect, and ant group (i.e. replicate) nested within ant colony as a random effect. The effect of treatment was tested by using the likelihood-ratio chi-square (LR) test. In assays on A. craccivora, we also performed a post hoc Tukey's test for pairwise comparison among treatment. Statistical analyses were conducted with the program package lme4 v. 1.1-12 and multcomp v. 1.4-6 in R software v. 3.3.2 [24].
(a) Results
Treatment significantly affected the attack ratio of signaller ants towards A. craccivora aphids (χ2 = 47.59, d.f. = 2, p < 0.001; figure 2a). Signallers that had directly tended aphids attacked the aphids significantly less frequently than did the two control groups of signallers (attack ratio, i.e. the frequency of attacks divided by the total frequency of contacts, mean ± s.e.: aphid-experienced 0.004 ± 0.004, satiated controls 0.234 ± 0.041, starved controls 0.198 ± 0.030; aphid-experienced versus satiated controls, Z = 4.239, p < 0.001; aphid-experienced versus starved controls, Z = 4.121, p < 0.001). Hunger level did not influence the attack ratio towards A. craccivora: the difference in the ratio between the two groups of signallers that had not experienced aphids (satiated and starved controls) was not significant (Z = −0.361, p = 0.926).
Figure 2. Attack ratio (frequency of attacks divided by total frequency of contacts: mean ± s.e.) of (a) signallers and (b) receivers towards the mutualistic aphids Aphis craccivora, and (c) signallers and (d) receivers towards the non-mutualistic aphids Acyrthosiphon pisum in Experiment 1. Signallers had either previously experienced A. craccivora aphids or not, and receivers had interacted with the signallers. Different letters above bars indicate significant differences between treatments.
Similarly, treatment significantly affected the ratio of attack on A. craccivora by receiver ants (χ2 = 20.462, d.f. = 2, p < 0.001; figure 2b). Receivers that had interacted with signallers with aphid experience had significantly lower attack ratios towards A. craccivora than receivers that had interacted with aphid-naive signallers: the mean (±s.e.) attack ratios were, for receivers that had interacted with aphid-experienced signallers, 0.062 ± 0.019; for those that had interacted with satiated controls 0.226 ± 0.030; and for those that had interacted with starved controls 0.150 ± 0.020 (aphid-experienced versus satiated controls, Z = 4.619, p < 0.001; aphid-experienced versus starved controls, Z = 3.078, p = 0.006). There was no significant difference in the ratio of attack on A. craccivora between receivers that had interacted with satiated aphid-naive signallers and those that had interacted with starved ones (Z = −1.749, p = 0.185).
Conversely, treatment did not affect the attack ratio of signallers towards A. pisum (attack ratio of signallers towards A. pisum, mean ± s.e.: aphid-experienced 0.278 ± 0.037, starved controls 0.320 ± 0.034; χ2 = 1.171, d.f. = 1, p = 0.279; figure 2c). There was also no significant difference in the attack ratio towards A. pisum between receivers that had interacted with signallers with aphid experience and those that had interacted with starved aphid-naive signallers (attack ratio of receivers towards A. pisum, mean ± s.e.: aphid-experienced 0.336 ± 0.034, starved controls 0.294 ± 0.030; χ2 = 1.018, d.f. = 1, p = 0.313; figure 2d).
4. Experiment 2: role of trophallaxis in social transmission
To test whether social transmission of partner information occurred during trophallactic interactions among ants, we used an experimental design similar to that described in Experiment 1. Ten signaller ants were given experience with A. craccivora aphids as described above. As receivers, we set up two different groups of ants that had not experienced aphids: one was provided with artificial honeydew (honeydew treatment), and the other was not provided with any food resources (control treatment). The aim of providing receivers with artificial honeydew was to prevent trophallaxis from signallers to receivers; we predicted that the receivers would not accept trophallactic resources from the signallers if the receivers had fed on enough sugar resources. In the honeydew treatment, we presented 10 receivers with cotton wool (10 × 10 mm) soaked in 200 µl of the same sugar solution used in Experiment 1 on a Petri dish. In the control treatment, 10 receivers were held in a Petri dish without any food resources.
After 8 h under each set of conditions, each group of receivers was transferred into a new Petri dish with a group of 10 signallers. We left the ants for 2 h and recorded their behaviour with a digital video camera (HC-W850M, Panasonic Co., Ltd, Osaka, Japan) for the first 1 h. We measured how many times receivers were provided with solution by signallers by mouth-to-mouth contact during the observation period. Each group of receivers was then transferred to a new Petri dish. Fifteen minutes after the transfer, an apterous adult aphid of A. craccivora was introduced into the dish and a tester observed the responses of the ants, as described above. We performed the experiments 16 times for each treatment (i.e. honeydew treatment and control).
The frequency of trophallaxis was analysed by using a GLMM with a Poisson distribution and log link function. A model was constructed by using frequency of trophallaxis as a response variable, treatment as a fixed effect, and ant group nested within ant colony as a random effect. The influence of treatment was tested by using the LR test. The ratio of attack on aphids by ants was analysed by using the GLMM and LR test in the same way as that described in Experiment 1.
(a) Results
When receivers interacted with signallers that had previously tended A. craccivora aphids, the frequency of trophallaxis from signallers to receivers that had been provided with artificial honeydew was significantly lower than that to receivers that had not been provided with any resource (frequency of trophallaxis from signallers to receivers, mean ± s.e.: honeydew treatment 0.75 ± 0.281, control 10.25 ± 2.741; χ2 = 32.959, d.f. = 1, p < 0.001, figure 3a). Furthermore, the ratio of aphid attack by receivers that had previously fed on artificial honeydew was significantly higher than that by the controls (aphid attack ratio of receivers, mean ± s.e.: honeydew treatment 0.174 ± 0.030, controls 0.076 ± 0.019; χ2 = 6.932, d.f. = 1, p = 0.008, figure 3b).
Figure 3. (a) Frequency of trophallaxis from signallers that had previously experienced Aphis craccivora aphids to receivers that had either been provided with artificial honeydew or not; and (b) attack ratio (frequency of attacks divided by total frequency of contacts: mean ± s.e.) of receivers towards A. craccivora in Experiment 2.
These results support the hypothesis that aphid-associated information acquired by naive ants from their aphid-experienced nest-mates during trophallactic interactions decreased the aggression of the receivers towards the aphids (see Discussion section). The results of this experiment alone, however, could have had alternative explanations. One was that the receivers may have acquired aphid information by means other than trophallaxis during the interaction with signallers, and they may have behaved aggressively towards the aphids simply because of their feeding on the artificial honeydew. The other was that the reduction in trophallaxis from signallers to receivers may have induced aggressive behaviour in the receivers towards the aphids, irrespective of the aphid-tending experience of the signallers. We tested these possibilities as described below.
5. Control experiment 1: effect of feeding on artificial honeydew
Here, we examined whether feeding on artificial honeydew influenced the aggressiveness of aphid-experienced ants towards A. craccivora aphids. Ten ants were given experience with the aphids as described above. These ants were transferred to Petri dishes with or without artificial honeydew. After 2 h, we transferred the ants into new dishes and observed their responses towards the aphids, as described above. We performed the experiments 16 times for each treatment (i.e. with or without honeydew). The attack ratios were analysed by using the GLMM and LR test, as described the above.
(a) Results
There was no significant difference in aphid attack ratio between aphid-experienced ants that had fed on artificial honeydew and those that had not received artificial honeydew (attack ratio of ants towards aphids, mean ± s.e.: with honeydew 0.008 ± 0.006, without honeydew 0.010 ± 0.007; χ2 = 0.0032, d.f. = 1, p = 0.955, figure 4). The aphid attack ratios of aphid-experienced ants remained low, regardless of whether or not they had fed on artificial honeydew.
Figure 4. Attack ratio (frequency of attacks divided by total frequency of contacts: mean ± s.e.) of aphid-experienced ants towards Aphis craccivora aphids when the ants were provided or not provided with artificial honeydew.
6. Control experiment 2: effect of trophallaxis from aphid-naive ants
To test whether trophallaxis from aphid-naive nest-mates affected the aggressiveness of ants towards A. craccivora aphids, we followed the same procedure as in Experiment 2, except that the signallers had never experienced aphids. Signallers were provided with artificial honeydew instead of aphids, whereas receivers were exposed to two different conditions, as in Experiment 2 (i.e. provided with artificial honeydew or not). We measured the frequency of trophallaxis from signallers to receivers and the aphid attack ratio of receivers, as described the above. We performed the experiments 16 times for each treatment (i.e. honeydew treatment and control). The frequency of trophallaxis and the attack ratio of receivers were analysed by using the GLMM and LR tests in the same way as that in Experiment 2.
(a) Results
When receivers interacted with aphid-naive signallers that had fed on artificial honeydew, the frequency of trophallaxis from signallers to receivers that had been provided with artificial honeydew was significantly lower than that to receivers that had not been provided with any resource (frequency of trophallaxis from signallers to receivers, mean ± s.e.: honeydew treatment 1.000 ± 0.303, control 14.063 ± 1.368; χ2 = 58.833, d.f. = 1, p < 0.001, figure 5a), as was found in Experiment 2. However, there was no significant difference in the aphid attack ratio between receivers that had been given honeydew treatment and those that had not (aphid attack ratio of receivers, mean ± s.e.: honeydew treatment 0.187 ± 0.031, control 0.164 ± 0.024; χ2 = 0.131, d.f. = 1, p = 0.717, figure 5b).
Figure 5. (a) Frequency of trophallaxis from signallers that had been provided with artificial honeydew to receivers that had either been provided with artificial honeydew or not; and (b) attack ratio (frequency of attacks divided by total frequency of contacts: mean ± s.e.) of receivers towards Aphis craccivora aphids in Control experiment 2.
7. Discussion
Social learning—defined broadly as learning through the observation of, or interaction with, other individuals—has evolved not only in vertebrates, but also in insects, despite their small brain size [10,11]. Although there is evidence that social learning or social transmission of information occurs in the contexts of food search and predator avoidance in social insects such as honeybees, bumblebees and ants [25–30], strict evidence of learning, i.e. the use of neuronally stored memory in behavioural change, is usually difficult to obtain. Furthermore, studies of social information use in the context of mutualism are also scarce.
Here, in Experiment 1, we found that receiver ants that had interacted with signaller ants that had tended A. craccivora were less aggressive towards A. craccivora aphids but were highly aggressive towards A. pisum, as were signallers that had directly tended aphids. This indicated that ants acquired information on A. craccivora aphids through interaction with their aphid-experienced nest-mates. Furthermore, inhibition of trophallaxis between aphid-experienced signallers and receivers, by providing the receivers with artificial honeydew in advance, increased the aggressiveness of the receivers towards A. craccivora aphids (Experiment 2). It was unlikely that feeding on artificial honeydew changed the receivers' aggressiveness (Control experiment 1), and that a reduction in trophallaxis from signallers induced receiver aggression towards the aphids irrespective of the aphid-tending experience of the signallers (Control experiment 2). Therefore, our results strongly support the hypothesis that information on aphids is transferred through trophallactic interactions between nest-mate ants.
However, there might be another possibility that the reduction of aggressiveness was not due to learning by ants; putative chemicals contained in the aphid honeydew changed ant physiology and reduced aggressiveness towards aphids. In fact, secretions produced by a mutualistic lycaenid butterfly reportedly affect levels of dopamine in the brains of tending ants, resulting in the reduction of locomotory activities of the ants [31]. However, previous studies clarified that mutualist recognition by ants is based on associative learning of nutritious rewards and cuticular hydrocarbons of mutualists [8], and that hydrocarbons of aphids elicit the reduction of aggression of aphid-experienced ants [7]. In the present study, moreover, the decrease of ant aggression was observed only towards A. craccivora aphids but not towards A. pisum. We do not, however, completely exclude the chemical (hormone) hypothesis yet, because a putative chemical produced by the mutualistic aphid A. craccivora might change ant physiology leading to reduction of ant aggression specifically towards A. craccivora. At least, to the best of our knowledge, this is the first study to show evidence that mutualist recognition of ants involves social information transmission.
Previous studies have shown or suggested that, through trophallactic interactions, social Hymenoptera (bees and ants) acquire information on the floral odour associated with sugar solutions [25–27,32]. For example, Camponotus mus ants choose floral odour in Y-maze tests after receiving from nest-mates through trophallaxis sucrose solution containing the floral scent [27]. This suggests that ants acquire chemical cues associated with foods during trophallaxis among nest-mates. Ants also discriminate mutualistic aphids on the basis of the aphids' cuticular hydrocarbons [7,17]. Therefore, it is reasonable to infer that the receiver ants in this study might have acquired the aphids' hydrocarbons during trophallactic interactions among nest-mates. We intend to study this hypothesis in the future.
Provision of honeydew to ants imposes costs on aphids [18,33,34], and this can explain why mutualistic interactions with ants are often evolutionarily unstable, having evolved multiple times in aphids and having frequently been lost [35,36], because the cost and benefit balance of mutualism can change, depending particularly on ant behaviour. Some studies [37–39] have suggested that ant-tending behaviour at the colony level can change quickly: ants can attend to, ignore, or even predate on aphids, depending on the partner's reward value. The social information transmission shown here might be a proximate mechanism for these quick responses to changing environments. However, it remains to be studied whether information sharing among nest-mates through social transmission enables an ant colony to choose better partners among currently available partner options to which they could attend. In other words, sharing the wrong information could hinder the colony-level adaptive response. We believe, however, that information transmission via trophallaxis can help avoid this drawback: each individual ant can taste the honeydew and compare it with the memorized values of previously tasted honeydews. Such an information-processing mechanism at the individual ant level is another important issue for future studies. For a deeper understanding of the evolution and maintenance of mutualism between groups of organisms, the integration of approaches focusing on multiple levels—such as individual, group and community—is necessary.
Data accessibility
Raw data can be accessed from the Dryad data repository (http://dx.doi.org/10.5061/dryad.n6d50) [40].
Authors' contributions
M.H. conceived the study, participated in its design, cultured the organisms, performed the behavioural experiments and statistical analyses and participated in the drafting of the manuscript; M.K.H. participated in the design of the study and the drafting of the manuscript; M.N. participated in the design of the study and the drafting of the manuscript; K.T. participated in the design of the study and the drafting of the manuscript and coordinated the study. All authors gave final approval for publication.
Competing interests
The authors have no competing interests.
Funding
KAKENHI (16K14865, 15H04425, 15H02652, 15K18610, 17J04148).
Acknowledgements
We thank Tomonori Kikuchi, Kiyoshi Nakamuta and Yasuyuki Choh for valuable advice on design of this study, Akihito Aizawa for valuable advice on statistical analysis, Kazuhiko Tamai for supporting the plant culture and providing Acyrthosiphon pisum aphids, Ed Vargo for some advice on English expression and Tadeusz Kawecki and two anonymous reviewers for valuable comments on the manuscript.