The impact of an ant–aphid mutualism on the functional composition of the secondary parasitoid community



    Corresponding author
    1. NERC Centre for Population Biology, Imperial College London, Ascot, Berkshire, U.K.
    2. Centre for Ecology & Conservation, School of Biosciences, University of Exeter, Penryn, Cornwall, U.K.
    Search for more papers by this author

    1. NERC Centre for Population Biology, Imperial College London, Ascot, Berkshire, U.K.
    2. Centre for Ecology & Conservation, School of Biosciences, University of Exeter, Penryn, Cornwall, U.K.
    Search for more papers by this author

Dirk Sanders, NERC Centre for Population Biology, Imperial College London, Silwood Park Campus, Ascot, Berkshire, SL5 7PY, U.K. E-mail:


1. Mutualistic and antagonistic interactions, although often studied independently, may affect each other, and food web dynamics are likely to be determined by the two processes working in concert.

2. The structure, and hence dynamics, of food webs depends on the relative abundances of generalist and specialist feeding guilds. Secondary parasitoids of aphids can be divided into two feeding guilds: (i) the more specialised endoparasitoids, which attack the primary parasitoid larvae in the still living aphid, and (ii) the generalist ectoparasitoids, which attack the pre-pupa of the primary or secondary parasitoid in the mummified aphid.

3. We studied the effect of an ant–aphid mutualism on the relative abundance of these two functional groups of secondary parasitoids. We hypothesised that generalists will be negatively affected by the presence of ants, thus leading to a greater dominance of specialists.

4. We manipulated the access of ants (Lasius niger) to aphid colonies in which we placed parasitised aphids. Aphid mummies were collected and reared to determine the levels of endo- and ecto-secondary parasitism.

5. When aphids were attended by L. niger the proportion of secondary parasitism by ectoparasitoids dropped from 26 to 8% of the total number of parasitised aphids, with Pachyneuron aphidis most strongly affected, while endoparasitoids as a group did not respond. However, among these Syrphophagus mamitus profited from ant attendance becoming the dominant secondary parasitoid, while parasitisation rates of Alloxysta and Phaenoglyphis declined.

6. The shift to S. mamitus as dominant secondary parasitoid in ant-attended aphid colonies is likely due to the behavioural plasticity of this species in response to ant aggression, and a release from tertiary parasitism by generalist ectoparasitoids.

7. The reduction of secondary parasitism by generalist ectoparasitoids reduces the potential for apparent competition among primary parasitoids with consequences for the dynamics of the wider food web.


A challenge in ecology is to uncover the link between food web structure and ecosystem function and stability (Montoya et al., 2006; Ives & Carpenter, 2007). In order to understand complex ecosystems, and therefore to be able to predict the effects of environmental change on them, we need to understand the biological and ecological mechanisms that determine connectance of food webs (Dunne et al., 2002). Connectance, the proportion of all possible links that is realised in a food web, is determined by the diet breadth of the different species in the web. Diet breadth itself is largely dependent on the feeding mode, with guilds that have more intimate interactions with their host/prey, like parasitoids for example, having a narrower diet than those with less intimate interactions, like predators (van Veen et al., 2008). Webs that are dominated by intimate interactions therefore have a lower connectance than those dominated by generalist predatory interactions, and indeed a similar pattern has been observed in plant–animal mutualistic networks (Blüthgen et al., 2007; Guimarães et al., 2007). So we can understand why some species are generalist feeders while others are specialists, but what determines the relative frequencies of these different guilds (and thus connectivity) in ecological communities?

Food webs represent networks of antagonistic interactions but organisms also interact in other ways. In the last decade there has been much research on the properties of mutualistic interaction webs (e.g. plant–pollinator) and how these compare with those of antagonistic webs (Montoya et al., 2006; Fontaine et al., 2009). There is however an increasing realisation that these different types of interaction do not act independently from each other (Ings et al., 2009). We used an ant–aphid–parasitoid–secondary parasitoid system for studying the effect of a mutualism on the structure of a secondary parasitoid food web.

Ants are a group of eusocial insects that almost any other insect group is likely to encounter at some stage because of their abundance and territoriality, the relative stability of their populations, as well as their feeding habits and aggressiveness. The outcome of these encounters can be positive or negative (Stadler & Dixon, 2005). In temperate regions, aphids are the most important honeydew producers and many have developed a mutualistic relationship with ants (Buckley, 1987; Hölldobler & Wilson, 1990). Aphids benefit from this mutualism with ants by reduced predation and parasitism and by a reduced risk of fungal infection (e.g. Banks, 1962; Way, 1963; Samways, 1983; Völkl, 1992; Fischer et al., 1997; Müller & Godfray, 1999).

Aphids are commonly attacked by hymenopteran parasitoids (Brodeur & Rosenheim, 2000) and these parasitoids have to deal with the aggressive behaviour of aphid-attending ants. The larvae of these parasitoids first feed on the host's haemolymph and later kill the aphid by feeding on other tissues creating a mummy. Primary aphid parasitoids are found in two families of Hymenoptera: the Aphidiinae (Braconidae) and the Aphelinidae with over 600 described species, all of them being solitary endoparasitoids (Mackauer & Starý, 1967). Primary aphid parasitoids are attacked by two functional groups of secondary parasitoids (Fig. 1, Müller et al., 1999): the first group of species are referred to as endophagous koinobiont hyperparasitoids (‘endoparasitoids’ from hereon), which nearly all belong to the Alloxystini (Figitidae, Charipinae). Endoparasitoids lay their eggs in the parasitoid larva in the still-living aphid (parasitised aphid host), where they remain to hatch after mummification of the aphids. The second group are generalist ectophagous idiobiont secondary parasitoids (‘ectoparasitoids’ from hereon) that attack the mummy and deposit their egg on the parasitoid larva or pre-pupa on which the larva feeds externally and can also cause mortality among secondary parasitoids (Fig. 1). This group contains several unrelated genera of Pteromalidae and a single genus (Dendrocerus) of Megaspilidae. Because of the more direct interactions with host defences, endoparasitoids are typically more specialised and have a more restricted host range than the generalist ectoparasitoids (Müller et al., 1999; Sullivan & Völkl, 1999; Brodeur, 2000; van Veen et al., 2003). Because of these differences in host range, shifts in the relative abundance of these guilds can have a major impact on the structure of the food web (Bukovinszky et al., 2008).

Figure 1.

Parasitoid food web with primary parasitoids and two functional groups of secondary parasitoids: endoparasitoids, which parasitise the primary parasitoid larva inside the aphid before mummification; the second group, the ectoparasitoids, attack the mummy and deposit their egg on the parasitoid larvae or pre-pupa, on which the ectoparasitoid larva feeds externally.

A number of studies have documented that attending ants reduce the parasitism of honeydew-producers through attacks and/or disturbances against ovipositing parasitoid females (e.g. Bartlett, 1961; Völkl & Mackauer, 1993; Itioka & Inoue, 1996; Stechmann et al., 1996), whereas several studies have shown that ant-attendance enhances primary parasitism by some species of aphid parasitoids (Völkl, 1992; Mackauer & Völkl, 1993; Völkl & Stechmann, 1998; Kaneko, 2002). Although the majority of secondary parasitoids have evolved escape and avoidance strategies that reduce ant-inflicted mortality of adult parasitoids, these lead to severe disturbance of foraging activities and generally lead females to give up an ant-attended aphid colony (Hübner & Völkl, 1996). The presence of ants can therefore lead to a significant reduction in secondary parasitoid foraging success, thereby potentially providing a kind of ‘enemy-free’ space for primary parasitoids (Völkl, 1992; Novak, 1994; Kaneko, 2003). Some Alloxystini wasps, however, are specialised on ant-attended hosts (Völkl et al., 1994). For example, the species Alloxysta brevis releases a mandibular gland secretion in response to an attack of the ant Lasius niger (Völkl et al., 1994), which functions both as self-defence and as a repellent that prevents ant attacks during subsequent encounters. It enables A. brevis females to hyperparasitise ant-attended aphids that constitute a major proportion of their host range. The presence of ants can therefore be expected to have a greater negative effect on generalist ectoparasitoids than on specialist endoparasitoids by shaping behaviours linked to resource finding and exploitation with consequences for the structure and dynamics of the aphid–parasitoid food web (Höller et al., 1993; Sunderland et al., 1997; Rosenheim, 1998; Boivin & Brodeur, 2006; Bukovinszky et al., 2008).

We manipulated access of ants to experimental aphid colonies in the field to test the hypothesis that in the presence of ants, secondary parasitism is dominated by specialists, leading to a relatively isolated food chain, while we expect more generalists in the absence of ants, resulting in more potential for interactions with the wider food web.

Materials and methods

Study organisms

A laboratory population of the primary parasitoid Lysiphlebus fabarum (Marshall) was initiated from mummified Aphid rumicis aphids collected in the field on Rumex obtusifolius and R. crispus in spring 2009 in Bracknell, Berkshire, U.K. This parasitoid species was used because it is known to exploit aphid colonies attended by Lasius niger L. Ant-attended Aphis fabae Scop. colonies in the field suffer from even higher parasitisation by Lysiphlebus than unattended colonies (Völkl, 1992). Prior to the experiments, parasitoids and their host A. fabae were maintained on broad bean (Vicia faba, variety the Sutton) and kept in a controlled environment room at 20 °C, relative humidity 75%, with a LD 16:8 h cycle at Silwood Park, Berkshire, U.K. Aphis fabae cultures were initiated by aphids gained from an existing laboratory strain at Silwood Park. This strain consists of a single clone and because aphid genotype can itself affect interactions at higher trophic levels, by using a single clone we avoid additional potentially confounding source of variation. Aphis fabae is often attended by ants and both attended and unattended aphids colonies are found in the field (Völkl, 1992).

Field experiment

To assess the impact of aphid-attending ants on the parasitisation rate of the two functional groups of secondary parasitoids, we manipulated the access of ants to aphid colonies in the field. Potted bean plants containing aphids infested by the primary parasitoid Lysiphlebus fabarum were placed close to L. niger ant colonies, with two treatments: (i) ant exclusion or (ii) ants present at aphids colonies. Prior to the experiment bean plants were grown for 2 weeks in the greenhouse. Then, each plant was infested with 20 wingless adults of A. fabae in the laboratory, to allow them to produce offspring for 24 h to gain aphid cohorts of similar age. Three days later, four females of L. fabarum were placed on each plant (covered with gauze). After another 3 days the aphid colonies were placed in the field with almost all aphids being parasitised, except for some adults, which were kept on the plants to attract the ants, because they produce larger amounts of honeydew.

The aphid colonies in the field were protected from predators using a wire cage method (Müller & Godfray, 1999) to avoid confounding effects of higher predator density and hence lower aphid density on ant free colonies. The wire cages keep out hoverflies (Syrphidae) and other insect predators but have no effect on parasitoids (mesh size is 5 mm). The cages were used as blocks each containing both treatments, an aphid colony attended by L. niger and an aphid colony with ant exclusion. Each treatment was replicated 15 times resulting in 30 aphid colonies in the field. The experiment run in August and was repeated end of August/September 2009.

Ant exclusion was achieved by a sticky barrier (tree grease) on the outside of the pots and by placing the pots on small platforms surrounded by water. Ants were encouraged to forage on the aphid-infested plants by linking the nest and the plant with a wooden stick smeared with honeydew (Müller & Godfray, 1999). Lasius niger was the only ant species observed on the bean plants in this experiment.

All aphid mummies were collected 3–5 days after the first mummies appeared and the first primary parasitoids hatched to ensure that the primary parasitoids had been exposed to the natural community of secondary parasitoids for the full duration of their development. Mummies were stored in individual gelatine capsules until adult parasitoids emerged. All parasitoids were identified to species level and categorised into the two functional groups of secondary parasitoids: ectoparasitoids and endoparasitoids.


The impact of ant presence on the parasitisation rate of secondary parasitoids was tested using linear mixed-effects models with a binomial error structure. We used lme4 package by D. Bates and M. Maechler in the open source software R 2.7.0. Linear mixed-effects models were used to partition the variation deriving from random effects of experimental blocks and repetition of the experiment from variability in the fixed (treatment) effects of the ant exclusion experiment.


The naturally assembled secondary parasitoid guild consisted of seven species with the three endoparasitoids Syrphophagus mamitus (Walker) (Encyrtidae), Phaenoglyphis villosa (Figitidae), and Alloxysta pusilla (Kieffer) (Figitidae). Phaenoglyphis villosa and A. pusilla are strict hyperparasitoids attacking the primary parasitoid larvae in the still living aphid, while S. mamitus is endophagous koinobiont, but has a dual oviposition behaviour, attacking the parasitoid larvae in live aphids and parasitoid prepupae or pupae in mummified aphids. All other secondary parasitoids Pachyneuron aphidis (Bouché), Asaphes vulgaris, Asaphes suspensus (Nees) (all Pteromalidae), and Dendrocerus carpenteri (Megaspilidae) were ectoparasitoids attacking the parasitoid prepupae or pupae inside mummified aphids.

The parasitisation rate (based on all collected mummies) by secondary parasitoids dropped in the presence of L. niger ants from 49 to 26% (Fig. 2, Table 1). This reduction was mostly related to a decline in the ectoparasitoids, from 26 to 8%. In contrast, parasitisation rate of endoparasitoids did not respond to ant presence in the aphid colonies (Fig. 2, Table 1). The rate of emerging primary parasitoids L. fabarum, which was the only primary parasitoid species in our experiment, increased from 51% in ant-exclusion treatments to 74% when attended by L. niger.

Figure 2.

Impact of aphid-attending ants on the parasitisation rate (mean + 1 SE) of all secondary parasitoids, endo- and ectoparasitoids. Differences in parasitisation rate were tested by using linear mixed-effects models (ns, not significant; ***P < 0.001).

Table 1.  Impact of ant presence on secondary parasitoid functional groups and species.
  1. Differences in parasitisation rate were tested by linear mixed-effects models, showing degrees of freedom (d.f.), estimates, z-values and probability (P).

Secondary parasitoids1, 48−0.7809−3.7500.0002
Endoparasitoids1, 480.31381.6040.1090
Ectoparasitoids1, 48−1.3691−6.110<0.0001
Alloxysta pusilla1, 48−0.6400−1.9860.0470
Phaenoglyphis villosa1, 48−0.3374−0.9640.3350
Syrphophagus mamitus1, 481.28884.567<0.0001
Dendrocerus carpenteri1, 48−0.9439−2.0700.0385
Pachyneuron aphidis1, 48−1.8374−5.556<0.0001
Asaphes suspensis1, 48−0.8280−1.9340.0531
Asaphes vulgaris1, 480.44180.5820.5610

The species A. pusilla (endoparasitoid), D. carpenteri and P. aphidis (both ectoparasitoids) had lower parasitisation rates in ant-attended aphid colonies (Table 1) while we found no statistical evidence for differences for P. villosa (endoparasitoid), and the ectoparasitoids A. vulgaris and A. suspensus (Fig. 3, Table 1). The only secondary parasitoid species that profited from ant presence was the endoparasitoid S. mamitus, increasing in rate from 5 to 12% (Fig. 3, Table 1). In the ant-exclusion treatment, the hyperparasitoid community was dominated by the ectoparasitoid P. aphidis, while in presence of ants the endoparasitoid S. mamitus became the dominant species (Fig. 4). In general, there was a shift from equal proportion of both functional groups in the absence of ants to the dominance of endo-hyperparasitoids on plants with ants (Fig. 4).

Figure 3.

Impact of aphid-attending ants on the parasitisation rate (mean + 1 SE) on different species of secondary parasitoids. Differences in parasitisation rate were tested by using linear mixed-effects models (ns, not significant; *P < 0.05; ***P < 0.001).

Figure 4.

Proportion of emerging secondary parasitoids for parasitoid communities in aphid colonies with and without attending ants (Lasius niger).


The presence of aphid-attending ants (L. niger) strongly reduced the success rate of secondary parasitoids, which is in accordance to the results of other studies (Völkl, 1992; Kaneko, 2002). However, only one of the two functional groups of secondary parasitoids responded to ant presence. Success rate of ectoparasitoids, dominant in ant exclusions, dropped dramatically while endoparasitoids, when treated as a group, were not affected. This result reflects our expectations that because of the more direct interactions of endoparasitoids with host defences they are typically more specialised than the generalist ectoparasitoids (Müller et al., 1999; Sullivan & Völkl, 1999; Brodeur, 2000) and have evolved strategies in dealing with ants, which allow them to exploit ant-attended aphids to a certain degree. We expected that in the presence of ants, secondary parasitism is dominated by specialists, resulting in less potential for interactions with the wider food web. While the main functional groups responded in the predicted way, both species with a strong response to ants, one with a decline and the other with an increase in parasitisation rate, were in fact relatively generalist. Differences for the response in the two functional groups to ant presence were mainly due to the strong decline of the common ectoparasitoid species P. aphidis, in contrast to an increase in numbers of the endoparasitoid S. mamitus, which was the dominant secondary parasitoid species in ant-attended aphid colonies (Fig. 4). Syrphophagus mamitus has been reported to attack many different parasitoids of aphids (Hoffer & Stary, 1970). So, while the presence of ants had a significant effect on the relative abundance of ecto- and endo-secondary parasitism, this did not necessarily lead to a reduced connectivity to the wider food web.

There are three main factors influencing the success of secondary parasitoids in ant-attended aphid colonies: (i) the degree of specialisation in dealing with ants, (ii) the time spent to find the host and to oviposit, and (iii) the plasticity of behavioural responses to encounters with ants. Endoparasitoids generally attack the still living aphid, which leads to more frequent encounters with ants and might result in adaptations (Hübner et al., 2002). Secondary parasitoids are, as far as we know, not as well adapted to ant attacks as for example the primary parasitoid Lysiphlebus cardui (Völkl, 1992) using chemical mimicry (Liepert, 1996; Liepert & Dettner, 1996). The functional group of endoparasitoids contains the Alloxistini, some of which have evolved adaptations in dealing with ants (Völkl et al., 1994). Like Alloxysta brevis (Völkl et al., 1994), A. pusilla and P. villosa might be able to release a mandibular gland secretion in response to an attack by ants (Hübner et al., 2002). In our study, however, they suffered from ant presence with a significant decline in numbers, although not as strong as the ectoparasitoids, which suggests this strategy is either not particularly effective or not used in these species. Phaenoglyphis villosa in particular has an unusually broad host range for an Alloxystini hyperparasitoid (Evenhuis & Barbotin, 1977), so it is therefore not surprising that it has not evolved specific strategies for ant-attended hosts. Alloxysta pusilla on the other hand, has only been recorded from Lysiphlebus primary parasitoid hosts, which tend to be associated with ant-attended aphid species. There was however no evidence that among endoparasitoids this specialist species was better adapted to the presence of ants than the generalist P. villosa.

Another important factor influencing the success of secondary parasitoids in ant-attended aphid colonies is the duration of finding the host and successful oviposition. In ant-attended aphid colonies, time needed to successfully parasitise an aphid is an important limiting factor because of the frequent danger of ant disturbance. The time required to parasitise a mummy is at least four times longer than for parasitising a larvae within live aphids (Buitenhuis et al., 2005), therefore ectoparasitoids need much more time for oviposition, and are more exposed to ant aggression. While most parasitoid species have evolved specific adaptations to exploit a single host stage, Syrphophagus displays a dual oviposition behaviour as they have the ability to attack either the primary parasitoid larva when the aphid is alive or the primary parasitoid prepupa or pupa after the parasitoid has killed and mummified its host (Kanuck & Sullivan, 1992; Buitenhuis et al., 2004). In both cases, the female lays a single egg inside the primary parasitoid, where the larva first develops as an endophagous parasite, but feeds ectophagously in later larval stages. If continuously disturbed by ants, Syrphophagus could switch to attacking predominantly live aphids, which would considerably reduce the time needed for oviposition. Völkl (1992) demonstrated that the presence of ants can lead to a significant reduction in secondary parasitoid foraging success, providing a kind of ‘enemy-free’ space for a primary parasitoid. It is possible that the increased parasitisation rate of the endoparasitoid S. mamitus in ant-attended aphid colonies is similarly due to reduced mortality caused by larval competition between secondary parasitoid larvae inside one aphid or tertiary parasitism from ectoparasitoids (Fig. 1), which may attack later than S. mamitus. A reduction in parasitism by the ectoparasitoid P. aphidis in the presence of ants may release S. mamitus from larval competition and tertiary parasitism, which could explain the positive effect of ants on this species.

The rate of successful oviposition in ant-attended aphid colonies is related to the strategy the secondary parasitoid species have evolved for dealing with ant encounters. Hübner and Völkl (1996) studied the behaviour of several species of secondary parasitoids when encountering ants. They classified two main response types: (i) parasitoids of the ‘avoidance type’ try to avoid contact with ants but stay on the plant while (ii) the ‘flight type’ quickly leaves the plant through jumping or flying. In contrast to the ‘flight type’, the ‘avoidance type’ stays on the plant and may still be successful. For example, Syrphophagus and similar, but less effective Asaphes species have both avoidance and flight behaviour, while P. aphidis, classified as ‘flight type’, stayed only for a short time on plants with ants (Hübner & Völkl, 1996) which explains why this species suffered most from ant presence.

The positive influence of ant presence on Syrphophagus is remarkable when compared with the pattern observed for all other species and may be due to a combination of behavioural responses to ant encounters, and a release from competition and tertiary parasitism by ectoparasitoids. Parasitisation of primary parasitoids by endoparasitoids was not as severely affected as that by ectoparasitoids. If aphid colonies attended by ants are favouring specialists then this will restrict the interactions with the wider food web. However, our results suggest that while there are both strong species-specific positive and negative responses to ant presence, these were not associated with generalism and specialism. By changing the composition of the secondary parasitoid community of their associated aphid species, ants may alter the indirect interactions with other primary parasitoid species and their aphid hosts in the surrounding habitat.


Thanks to Ki Jung Nam (Silwood Park Campus, Imperial College) for providing A. fabae. This study was financially supported by the German Research Council (DFG).