Ants (Hymenoptera: Formicidae) often engage in mutualistic interactions with aphids (Hemiptera: Sternorrhyncha, previously Homoptera; [1]), receiving carbohydrate-rich honeydew in return for a number of protective and sanitary services [2,3,4]. In this symbiotic interaction, behavioural and physiological adaptations in both species groups ensure that interacting individuals are often co-located in space and time. For example, evolutionary adaptations in ants range from herding aphids using both physical means (e.g., pulling off alate aphid wings; Kunkel 1973, as cited by [5]) and chemical means (e.g., suppressing alate aphid development; [6]), and even transporting aphids to new feeding sites [7,8,9]. Evolutionary adaptations in aphids include a reduced movement response to aphid alarm pheromones when in the presence of ants [10], an ability to modify honeydew composition [11,12,13] and, possibly, reduced apterous dispersal behavior (although it is still unconfirmed whether this is adaptive for aphids or is an extended phenotype of ants, i.e., through chemical manipulation; [14]).

The symbioses between ants and aphids, however, are not static but in a continual state of flux, depending on the changing distributions and densities of interacting species [15,16,17]. In certain circumstances, costs may exceed benefits causing the net effect of the interaction to switch sign from mutualism to antagonism, with potentially important consequences for community dynamics [18,19,20]. The key point is that each organism evolves to maximise its own benefit within the interaction. Physiological and behavioural innovations may arise which allow organisms to ‘fine tune’ interactions and derive maximal benefits. For example, aphids have evolved to modify the quality and quantity of honeydew in order to alter its attractiveness to ants [11,12,13]. Honeydew forms an important part of the diet for many ant species and its availability can structure ant communities [21,22,23,24,25]. In contrast, ants may maximise benefits by changing their behaviour towards aphids; for example, increasing predatory- rather than tending- behaviour.

Previous research shows that ants may switch from tending aphids to preying upon them if there are alternate sources of carbohydrate available to ants [26]. It has also been suggested that predation rates may increase if protein-rich prey sources are scarce [27], although a study by Offenberg [26] did not support this hypothesis. Offenberg found that alternative protein sources offered to ants did not affect aphid predation rates. Offenberg’s study did not consider the effect of protein supply together with demand of ant larvae in the colony, however. In colony developmental stages where ant larvae are present, protein may be more important for colony growth [28]. Colonies with larvae present often up-regulate foraging for protein food sources [29]. Hence, ants might increase predation of aphids unless alternative protein sources are provided.

In this study we investigated whether the presence of larvae in ant colonies, and previous diets fed to ants, affected aphid fitness. Our aim was to manipulate the developmental or physiological cues detected by ant workers. We measured the fitness effects on aphids in terms of colony reproductive success, a measure of Darwinian fitness [30].

Changes to parthenogenetic aphid colony growth rates may drive ecological and evolutionary changes in aphid-ant interactions by modifying selection pressure for mutualistic or antagonistic traits [31,32]. Our hypothesis was that the nutritional status of ant colonies can affect their behaviour towards—and the subsequent fitness of—tended aphid colonies [2]. Specifically, we test whether increasing ant colony protein demand through the presence of larvae will have a deleterious effect on aphid colony growth rates and hence switch the interaction from mutualism or commensalism towards parasitism.

We assessed the effects of ants on the fitness of mutualistic aphids using ants fed on different diets and kept with larvae present or absent. Two Lasius niger L. colonies, with brood and queens, were excavated from the grounds of Silwood Park, Ascot, UK in October 2006, and each stored in an individual five-litre plastic container (450–750 ants in each) with soil, moist cotton wool and a test tube containing sucrose solution with a cotton wool bung. Lasius niger is a common ant in temperate ecosystems, especially on disturbed ground and gardens. Foraging ants are recruited to food sources via pheromone trails left by scouts [33,34]. The species feeds on insect prey but also solicit honeydew from Sternorrhyncha [2,35], which may provide much of the ants’ nutritional requirements in terms of proteins (from amino acids) and carbohydrates [2]. The interaction with Aphis fabae Scopoli is generally considered to be mutualistic although predation by ants often occurs, and laboratory experiments often reveal a net parasitic effect on aphids, probably due to the lack of natural enemies in experimental set-ups [36,37,38,39]. Colonies were overwintered by storing in a refrigerated room at 5 °C. Before bringing back to room temperature in February or April, colonies were kept at 10 °C for one week [40]. Ants were then placed into forty eight 20 × 10 cm plastic containers with 20 workers per box. Each box contained an upturned plastic plant saucer for ants to gather beneath, a test tube with sucrose solution and two small open plastic boxes containing water and moist cotton wool.

During the experiment, the 20 ants in each box were allowed access to two Vicia faba L. bean plants, upon which five fourth instar A. fabae aphids were placed at the start of the experiment (Figure 1). The aphids were collected from Silwood Park UK and kept in culture for two years on V. faba plants in netted cages. The plants used in experiment were selected seedlings all ca. 10 cm tall (i.e., about 20 days old) and grown with the same watering regime in standard potting compost in the greenhouse. Plants were from different parents and, were therefore, randomly distributed among the three treatments because plant genotype can affect ant-aphid interactions [19]. A week before the experiment the plants were removed from soil and the roots washed. For a separate study considering host plant quality (unpublished data), one plant was placed in de-ionised water and one in rainwater. However, the current study focuses only on the effects of different ant treatments, rather than differences between plants, therefore, when referring to aphid counts we henceforth refer to the sum total of aphids on both plants. For each box, the number of foraging ants on plants and the total number of aphids surviving on both plants were recorded at midday on seven consecutive days.

Ant treatments were as follows: Colony origin (two different colonies), Larval presence (20 larvae present, or larvae absent), Diet (protein and sugar, or sugar only). To mimic different developmental stages, the experimental colonies, which were all queenless, were either kept with larvae during the winter and throughout the experiment, or larvae were absent with workers only present. At the start of the experiment, larvae were in the intermediate stage of development, approximately 2.5–3.5 mm long. For the different dietary regimes, some ants were fed sugar (20% sucrose ad libitum) and twice a week protein in the form of chopped Calliphora larvae (Diptera: Calliphoridae) (Crowthorne Fishing Supplies, Berkshire, UK), while the remaining ants subsisted on sugar alone. Each combination of factor levels was tested, which amounted to eight treatment combinations, and with six replicates of each, to 48 boxes in total. In addition, 36 replicates of a control treatment, with aphids and plants but no ants present, were conducted.

Due to space and time constraints, only 14 boxes were tested per week over six weeks. To achieve a fair test, these 14 boxes comprised the complete combination of factor levels (eight boxes) plus six control boxes, and time was included as a blocking factor in the analysis. For the first three experimental ‘blocks’, ant treatments were initially set up on diets in February and testing with aphids began in March. Thus, ants were kept on diet for four, five or six weeks before experimentation. For the final three experimental ‘blocks’, ant treatments were removed from refrigeration and set up in March. Tests began again four, five or six weeks later. Ant survival was assessed at the end of the experiment. Zero to five individuals died in each colony; however these deaths were random with respect to the treatments tested.

To summarise, our results show that larval presence in ant colonies reduces the growth rates of aphid colonies in an ant-aphid interaction. The mechanism may be direct predation of aphids by ants, or communication of a threat of increased predation leading to increased production of aphid honeydew. Whatever mechanism dominates, it appears that ants balance both the supply of carbohydrates in the environment and demand for nutrients from within the colony to determine their behaviour towards aphids. Aphids may be tended and protected as renewable (mainly sugar) resources or used as a protein-rich meal. From the results of this study we predict that aphid rewards from the ant-aphid interaction will change on a seasonal basis, as the demand from larvae within the ant colony waxes and wanes.