Coexistence through mutualist‐dependent reversal of competitive hierarchies

Abstract Mechanisms that allow for the coexistence of two competing species that share a trophic level can be broadly divided into those that prevent competitive exclusion of one species within a local area, and those that allow for coexistence only at a regional level. While the presence of aphid‐tending ants can change the distribution of aphids among host plants, the role of mutualistic ants has not been fully explored to understand coexistence of multiple aphid species in a community. The tansy plant (Tanacetum vulgare) hosts three common and specialized aphid species, with only one being tended by ants. Often, these aphids species will not coexist on the same plant but will coexist across multiple plant hosts in a field. In this study, we aim to understand how interactions with mutualistic ants and predators affect the coexistence of multiple species of aphid herbivores on tansy. We show that the presence of ants drives community assembly at the level of individual plant, that is, the local community, by favoring one ant‐tended species, Metopeurum fuscoviride, while preying on the untended Macrosiphoniella tanacetaria and, to a lesser extent, Uroleucon tanaceti. Competitive hierarchies without ants were very different from those with ants. At the regional level, multiple tansy plants provide a habitat across which all aphid species can coexist at the larger spatial scale, while being competitively excluded at the local scale. In this case, ant mutualist‐dependent reversal of the competitive hierarchy can drive community dynamics in a plant–aphid system.


| INTRODUCTION
The coexistence of similar species in an ecological community has long been studied (May, 1973). Research supports the theory that species must sufficiently differ in their ecology to enable coexistence (Chesson, 2000). Mechanisms that allow for the coexistence of two competing species that share a trophic level can be broadly divided into those that prevent competitive exclusion of one species within a single patch, and those that allow for coexistence only at a regional level, that is, among patches. Prevention of competitive exclusion at the regional level implies that competitive exclusion may occur within each local community patch, but at any one time different patches are occupied by different communities, leading to regional coexistence (Atkinson & Shorrocks, 1981). At the local level, coexistence of two species that share a trophic level is made possible by a range of mechanisms. This includes differential resource uptake, that is, differences in traits of the individual species (Debout, Dalecky, Ngomi, & McKey, 2009), and frequencydependent attack by natural enemies, that is, interactions with a third species (Abrams, 1999;Gliwicz & Wrzosek, 2008;Paine, 1966;Slobodkin, 1964;Vandermeer & Pascual, 2006). Predator-mediated coexistence has been described for a number of systems, from intertidal through to invertebrate terrestrial communities (Belsky, 1992;Berendse, 1985;Connell, 1961;Huntly, 1991;Olff & Ritchie, 1998;Paine, 1966).
The tansy-aphid system is ideal for studying the effect of mutualists on coexistence of aphids. Tansy plants (Tanacetum vulgare) commonly host three species of specialized aphid that vary in where they feed on the host plant (top shoots or lower leaves), and in their interaction with ants (Mehrparvar, Mansouri, & Weisser, 2014). Two of these aphid species (Metopeurum fuscoviride and Macrosiphoniella tanacetaria) both feed on the upper shoots of the plant and are rarely observed coexisting on the same host plant, yet both can be present on different host plants in the same field (Loxdale et al., 2011;Weisser & Harri, 2005). This suggests competitive exclusion at the local scale (individual plant), with coexistence only possible across multiple plants at the field or regional scale. Further, M. tanacetaria aphids are not ant-tended, whereas M. fuscoviride are highly dependent on their mutualistic interaction with ants (Senft et al., 2017;Weisser, 2000). Metopeurum fuscoviride aphid colonies visited more frequently by Lasius niger ants had higher colonization success at the beginning of the season, and those visited more frequently by Myrmica rubra had lower colonization success, lower persistence, and higher extinction rates (Senft et al., 2017). A third aphid species, Uroleucon tanaceti, feeds on the lower leaves of the plant, is not ant-tended, and is found coexisting locally with both the other aphid species in the field. It is toxic to predators, which could mediate predation pressures for the other coexisting aphid species in the community (Mehrparvar, Mahdavi Arab, & Weisser, 2013). Understanding the outcome of aphid-ant interactions within these local communities can help us to understand the processes driving local competitive exclusion, but regional coexistence of species.
In this study, we used both greenhouse (highly controlled community composition) and field experiments (natural colonization of ants and other aphid predators) to understand the relative effect of competition, predation, and mutualism on the coexistence of different species of aphid herbivores on tansy plants. In particular, we addressed the following questions: (i) What is the effect of ants on the individual occurrence and abundance of the three aphid species? (ii) How do ants affect the outcome of competitive interactions among the three aphid species? and (iii) What are the consequences of ant presence on the coexistence of the three aphid species?

| Study system
Tansy (Tanacetum vulgare) is a perennial herbaceous composite from Europe and Asia (Mitich, 1992) which preferentially grows in disturbed, well-drained, and poor soils. It often forms as isolated patches alongside river valleys, railway tracks, and on abandoned areas in cities or brown fields. Single plants comprise a "genetically identical" genet with up to 100 flowering ramets (shoots), but usually much fewer. In Jena, Germany, eight aphid species have been found on tansy (Mehrparvar, personal (Benedek et al., 2015;Flatt & Weisser, 2000;Mehrparvar, Zytynska, & Weisser, 2013;Mehrparvar et al., 2014), but also by other species such as the common red ant, Myrmica rubra (L.) (Senft et al., 2017). Metopeurum fuscoviride feeds in more compact colonies near the apex of ramets but can also occupy (at least to a certain extent) the same feeding niche as M. tanacetaria. Mixed colonies are very rarely observed in the field (Loxdale et al., 2011).
Reduced survival and reproduction of M. fuscoviride have been shown when aphids are not ant-tended (Flatt & Weisser, 2000). The third species, U. tanaceti, feeds on the underside of lower leaves of its host plant and is also not ant-tended. Loxdale et al. (2011) and Mehrparvar, Zytynska et al. (2013) discuss the life cycle of M. tanacetaria and M. fuscoviride. Both species are monoecious and holocyclic on tansy, but whereas the males of the former species are winged, those of the latter are wingless (Blackman & Eastop, 2006).
A wide range of natural enemies such as parasitoid wasps, ladybirds, lacewings, syrphid flies, predatory bugs, and spiders attacks these aphids. Even though the aphid species occupy different parts of a host plant, there may still be indirect competition for phloem nutrients (e.g., Moran & Whitham, 1990).
The experimental system, comprising the tansy plants, the three specialized aphid species, and two ant species (Lasius niger and Myrmica rubra), was studied using a greenhouse and two field experiments performed in Jena, Germany, at the Jena Experiment site on the northern outskirts of the city (50.95°N, 11.63°E) and the botanical garden of Jena (50.93°N, 11.58°E).

| Greenhouse experiment
The aim of this experiment was to determine the interactions among the three aphid species on tansy in the presence or absence of ants and absence of other potential natural enemies. L. niger and M. rubra, were collected from the same field as the tansy stolons and were thereafter maintained in the greenhouse. Each ant colony had several hundred workers, many ant larvae, and pupae. The colonies were housed in 10-L volume buckets filled with humid soil, the inside of which was coated with Fluon (Fluoropolymer Dispersion, Whitford GmbH, Germany), which acts as an ant barrier keeping them from escaping. The ants were offered boiled egg as protein in addition to carbohydrate resources that they would obtain from the aphids.

Experimental plants and insects
The soil in the buckets was sprayed frequently with water to avoid effects of desiccation.

Experimental design
This experiment was undertaken in the greenhouse, in the absence of natural enemies, and using a fully factorial randomized block design with 10 blocks. The main experimental treatments were three ant treatments (with L. niger, with M. rubra and without ants), each one with seven aphid treatments. The aphid treatments included each aphid species on its own, each aphid species with one another aphid species, and all three aphid species together, that is, M. tanacetaria Two adult individuals of each aphid species were placed onto each given experimental plant treatment using an additive design, such that when there were two or three aphid species together there were still two adults of each aphid species on the plant.
After about 4 hr, subsequent to the plant infestation by aphids, the worker ants were allowed access to the aphid infested plants of the ant-tended treatments. For each block, two buckets containing the colonies of the two ant species were placed in the experimental arena, that is, each ant colony attended all appropriate aphid treatments within a block. Access of the appropriate ant species was regulated by a series of bamboos sticks (~5 mm diam.). These connected the buckets housing the ant colonies to the plants. Plant pots were placed into water-filled plates to prevent the escape of worker ants as well as the access of vagrant workers in the greenhouse to no ant-treatments. The plants in non-ant treatments were not connected to any ant colony. All plants were irrigated gently every time they required water.
To obtain cohorts, adult aphids were allowed to reproduce for 48 hours whereupon the numbers of adults (0, 1, 2) were counted to assess early adult survival in the treatments. Then the adults and all offspring except three 1st instar nymphs were removed from the plants to obtain populations of similar reproductive age. When no initial adults survived, 1st instar nymphs from stock colony plants were added. Following this, the number of aphids was counted each day for 20 days.
The variables used for analyses were as follows: The number of adults initially put on the plant that survived until the second day (early adult survival); the cumulative number of individuals (the sum of all daily aphid counts, as measure for population growth and productivity for 20 days); and, colony persistence, calculated as the number of days until no aphid was present any more on the plant, up to day 20.

| Field experiments
The aim of this experiment was to determine the competition between aphid species and to assess the colonization success of aphids on tansy in the presence or absence of ants in the field when natural enemies are present.

Competition experiment
Here tansy rhizoms were collected, cultivated, and maintained as in the greenhouse experiment, and we used the three tansy aphid species as before. In August 2010, colonies of L. niger were located within an experiment grassland site for use in the experiment. Before placing the plant pots in the field, all small plants around the place where a plot was placed were re-moved at ground level, to exclude the access of ants to the experimental plant via the aerial parts of nearby plants. For "without ant" treatments, each potted tansy plant was placed in another empty pot without holes in its base. Insect glue (Raupenleim grün, Schacht, Germany) was daubed on the outside surface of the pots. The plants were irrigated gently, so as not to disturb the vegetation or aphid colonies, as required. To avoid any build-up of water within plots, these were checked for daily, but this only occurred once after heavy rain.
The experiment was started by placing two unwinged adult female aphids and five 3rd or 4th instar nymphs, hence a total of seven aphids, on the plant, for each of the three aphid species per experimental plant treatment. After 1 day of the experiment, aphids were checked and if the total number of aphids per species on each plant was less than five, numbers were suitably increased with new individuals. Thereafter, the numbers of aphids were counted each day in the morning for 21 days.
The following variables were used for analysis: the cumulative number of individuals after 21 days, that is, the sum of all daily aphid counts, as measure for population growth and productivity; and colony persistence, calculated as the number of days until no aphid was present any more on the plant, up to day 21. For the without ant treatment, each potted tansy plant was placed in another empty pot without holes in its base, and insect glue was daubed on the outside surface of the pots, as described above. This experiment was performed in the botanical garden of Jena, Germany, where some other tansy plants grow and are naturally infested by aphids over the season. The plants were allowed to be naturally colonized by aphids, and then they checked every week from 23 May until 3 October. The number of winged (adult) aphids and the number of unwinged adult aphids plus nymphs were counted separately.

| Statistical analysis
For the greenhouse experiment, generalized linear models using binomial distribution with logit link were used to analyze early adult survival of each aphid species. All analyses were conducted using SPSS version 22.
For both greenhouse and field experiments, the cumulative numbers of individuals at the end of the experiment were transformed as (x + 1). Generalized linear models using a Gamma distribution with log link function were used for the analysis. For each aphid species, models included main effects for the competition (aphid combination) and ant treatments, as well as interactions. Aphid competition treatments included the focal aphid species on its own, the combination of the focal

Early adult survival
The presence of ants strongly reduced early adult survival; in many cases no adults, and on average fewer than one adult survived until day two, while in the absence of ants, both individuals almost always survived (Figure 1a, Table 1). Workers of both L. niger and M. rubra were often observed killing and carrying M. tanacetaria to their nest ( Figure S1). Competition also negatively affected early adult survival, so that the average number of adults when they were accompanied by other aphid species was smaller (Figure 1a). The interaction between the competition and ant treatments was not significant (Table 1).

Colony persistence and population growth
Both the presence of L. niger and M. rubra caused a drastic reduction in colony persistence of M. tanacetaria, whereas in the absence of ants on the plant, colony persistence was on average about 2.5 times as long (Figure 2a and Table 3). Competition had no significant effect on colony persistence in the no ant treatment (Table 2). In the presence of the ant L. niger, the shortest colony persistence was observed when all the three aphid species were on the plant, while it was four times longer when M. tanacetaria was alone (Table 2). In the presence of the ant M. rubra, the longest colony persistence was also in the "no-competition" treatment, while the presence of M. fuscoviride on the plant caused a significant decrease in colony persistence (Table 2).
Considering cumulative number of individuals, the interaction between ant and competition treatments was significant (Table 1)  The ant treatment had three levels in the case of the greenhouse experiment (L. niger, M. rubra, without ant), and two levels in the case of the field experiment (with/without L. niger). For the analysis of early adult survival in the greenhouse, generalized linear models with a binomial error distribution and logit link were used. The cumulative numbers of individuals at the end of the experiments were transformed as (x + 1) and analyzed using generalized linear models with a Gamma error distribution and log link function. Significant results are indicated by *p < .05, **p < .01, ***p < .001.
one-fifth in the presence of M. fuscoviride, U. tanaceti, and in the combination of all the three species, respectively.

Early adult survival
In the presence of ants, early adult survival was generally better; in most cases both adults survived. In the absence of ants, on average only about 0.5 adults survived in the first 2 days (Figure 1b, Table 1).
Competition had not significant effect on early adult survival (Table 1).
There was also no significant interaction between the ant and competition treatments (Table 1).

Colony persistence and population growth
In the presence of L. niger, M. fuscoviride colonies were more persistent than in the absence of ants, while the differences between no ant and M. rubra treatments, and between the L. niger and M. rubra treatments, were not significant (Table 3, Figure 2b). In the absence of ants, colony persistence was longest in the "no-competition" treatment (M. fuscoviride on its own), while the presence of M. tanacetaria or U. tanaceti on the plant decreased colony persistence by about 50% and 80%, respectively (Table 2). In the three-species treatment,  (Table 2).
Considering the cumulative number of individuals, the interaction between the ant and competition treatments was significant (Table 1).
The presence of ants significantly increased the cumulative number of

| Uroleucon tanaceti (untended)
Early adult survival In the absence of ants, early adult survival of U. tanaceti was more than 1.5 times greater than in the presence of ants (Figure 1c, Table 1).

Colony persistence and population growth
Colony persistence of U. tanaceti was significantly shorter in the presence of L. niger and M. rubra than in the absence of ants (Table 3, Figure 2c). There was a tendency for lower colony persistence in competition with M. fuscoviride in the presence of L. niger ( Table 2).
The presence of ants also had a negative effect on the cumulative number of U. tanaceti individuals, and the effects of L. niger tended to be stronger than that of M. rubra (Figure 1f). Competition had no significant effect on the cumulative number of U. tanaceti individuals and also the interaction between ant and competition treatment was not significant ( Table 1) (Table 1). In the absence of ants, M. fuscoviride apparently had no effect on the population growth of M. tanacetaria, while T A B L E 2 The effect of the presence of ants and aphid-aphid competition on colony persistence, in days, of three specialized tansy aphids in the greenhouse and field experiments The ant treatment had three levels in the case of the greenhouse experiment (L. niger, M. rubra, without ant), and two levels in the case of the field experiment (with/without L. niger). For the analysis of colony persistence, a survival analysis (Kaplan-Meier) was used. Pairwise comparison between treatments was performed using the log-rank test. Mean (±SE) of colony persistence in different aphid (competition) treatments in the presence and absence of ants during the 20 days of greenhouse experiment and 21 days of field experiment is given. Means in rows with different letters and in columns with different symbols are significantly different from one another (p < .05).

| Competition experiment
the negative effect of U. tanaceti was found to be significant. In the presence of ants, competition decreased the cumulative number of individuals in all competition treatments, more especially when M. fuscoviride was present, so that it was greater in the "M. tanacetaria alone" treatment compared with the other aphid combinations (Figure 3a).

Metopeurum fuscoviride colony persistence and population growth
As with the greenhouse experiment, the ants had a positive effect on colony persistence of M. fuscoviride, and colonies persisted on average about five times as long when they were not attended by ants (Tables 2 and 3). Competition had no discernible effect on the colony persistence of M. fuscoviride (Table 2). In the "no ant treatment," colony persistence was very short independent of the competition treatment, colonies rarely survived beyond the first 3 days.
In the presence of L. niger, populations grew considerably and the cumulative number of individuals were high, about 33-fold the average of the no ant treatment, in all combinations with other aphids or when alone. In contrast, in the absence of ants, colonies became extinct in the first 3 or 4 days due to the action of predators in all aphid species combinations (Table 1, Figures 2e and 3b). The main effect of competition and the interaction between ant and competition were hence not significant (Table 1). The ant treatment had three levels in the case of the greenhouse experiment (L. niger, M. rubra, without ant), and two levels in the case of the field experiment (with/without L. niger). For the analysis of colony persistence, a survival analysis (Kaplan-Meier) was used. Pairwise comparison between ant treatments was performed using the log-rank test.

F I G U R E 3
The effect of ant presence and aphid-aphid competition on population growth (cumulative number of individuals) of specialized tansy aphids in the field experiment. Aphid competition treatments included the focal aphid species on its own, the combination of the focal with either of the two other aphids species, and the three-species combination. The ant treatment had two levels (with/without L. niger). Mean ( (Table 2).
With regard to the cumulative number of U. tanaceti individuals, the interaction between ant and competition was marginally significant, and patterns were complicated (Table 1, Figure 3c). In the absence of ants, there was no effect of competition on the cumulative number of   (Table 4).

| DISCUSSION
Our results show that the presence of ants drives community assembly of the three specialized tansy aphids at the local level ( (Clancy, Zytynska, Senft, Weisser, & Schnitzler, 2016;Senft et al., 2017). Therefore, we must begin to consider this system at the regional level, using our understanding of the local community dynamics in order to understand fully what drives species interactions and coexistence of the different aphid species.

| Local interactions among aphids and the role of ants
The ant-tended M. fuscoviride benefits from ants in different ways, for example, protection against natural enemies (Billick et al., 2007;Tilles & Wood, 1982;Way, 1963). This can indirectly benefit other species, for example, U. tanaceti aphids on the same host plant as- change from mutualist to predator due to a lower production of honeydew by M. tanacetaria and also lower honeydew quality (Woodring, Wiedemann, Fischer, Hoffmann, & Völkl, 2004), so that there is little return to the ants when tending this aphid species (see Fischer, Hoffmann, & Völkl, 2001;Sakata, 1994Sakata, , 1995Skinner, 1980). Even though U. tanaceti belongs to the same group of herbivores on the tansy plant, there is nevertheless a different and somehow complicated impact of ants on it. Ants occasionally visited the U. tanaceti colonies on the plant, but both attendance and predation were rare. It is known that U. tanaceti can be toxic to some common aphid predators (Mehrparvar, Mahdavi Arab et al., 2013) and perhaps also to the ants, thus explaining the ant avoidance of this aphid species.
F I G U R E 5 Interaction web of tansy aphids including indirect effect of ants. Blue arrows show interactions between two aphid species in the absence of ants, red arrows show interactions in the presence of ants. Gray arrows show main effect of ants on the different aphid species. + positive effect; − negative effect, 0 neutral effect. Effects in brackets are tendencies, for example, effect only significant in field experiment but not in greenhouse experiment (see Table 2). Dashed and solid lines highlight the direction of the arrow, from one aphid to the other (solid) and vice versa (dashed)  (Begon, Townsend, & Harper, 2006

| Coexistence across host plants
Overall  (Clancy et al., 2016). Thus, intraspecific plant variability may affect aphid population growth not only directly, by changing plant suitability for the aphids, as is well-known from plant resistance research and also found in natural systems, but also indirect via ant preference . In fact, variation in tansy chemotypes has been found at the level of a field (Clancy et al., 2016). This chemical variation thus creates a heterogeneous habitat for the aphids and hence community assembly processes. Further work is necessary to understand the role plant variation in driving community assembly processes of aphid communities through effects on mutualistic ants and possibly predators. The mechanisms may in fact be even more complicated as colony persistence time in the field (how long a plant is colonized for) is also ant species related, with persistence time being reduced when colonies were tended by M. rubra rather than L. niger (Senft et al., 2017). In other systems, host plant genetic variation was also able to structure aphid populations indirectly through interactions with mutualistic ants (Abdala-

| Conclusions
In the current study, we have experimentally shown that ant mutualistic/antagonistic relationships drive competitive exclusion between two aphid species within local communities (i.e., individual plants), with a third aphid less affected by these interactions. In such a system, with multiple local patches linked by dispersal, we also suggest that understanding how biotic interactions drive community assembly at the local scale can benefit future work in the field of metacommunity ecology, where the effect of these local interactions can be further investigated across multiple patches at the field or regional scale.

ACKNOWLEDGMENTS
This study is devoted to Wolfgang Völkl who started the work on tansy and its aphids. Special thanks are due to Hugh D. Loxdale