Behavioural plasticity associated with propagule size, resources, and the invasion success of the Argentine ant Linepithema humile

Linepithema humile were collected from Hastings (39°63′ S, 176°85′ E, and 39°63′ S, 176°86′ E) and Wellington (41°28′ S, 174°76′ E) and then combined in the laboratory as they are genetically similar and display no aggression towards each other (Corin et al. 2007). Monomorium antarcticum were collected from Wellington (41°28′ S, 174°71′ E) in July 2006. Colonies were placed in separate plastic containers (11 × 7 × 6 cm) with the sides painted with fluon™ (Polytetrafluoroethylene) to prevent escape. Each container had two 10-ml nesting tubes one-third filled with water, plugged with cotton wool, and covered with aluminium foil (Hee et al. 2000). They were fed daily with a mixed diet of 25% sugar water (via absorbent cotton wool), tuna in oil and peanut butter. Incubation temperature was 25 °C and humidity 50% with a light–dark cycle of 12:12 h light:dark.

To examine L. humile propagule size establishment success in the presence of M. antarcticum, we used a factorial design with four categories of L. humile propagule size and two food treatments (low food and high food). The low food treatment had one food dish and the high food treatment had 10 food dishes. Food in these dishes consisted of approximately 0·3 g of scrambled egg, 0·6 g of peanut butter, 25% sugar water (via absorbent cotton wool), and one dead fly Lucilia sericata Meigen daily. Food items were mixed into individual shallow dishes (0·7 cm wall and 2·8 cm in diameter) and haphazardly placed inside the foraging arena. Linepithema humile propagules consisted of colonies of 10, 100, 200, or 1000 workers, each with two queens. Colonies were placed in individual nesting containers (14 × 9 × 8 cm) with the top half of all inner walls painted with fluon. Each nesting container had two nesting tubes. A plastic exit tube (5 cm long, 0·3 cm internal diameter) was inserted into one of the shorter container sides. Each L. humile propagule size (except for controls) was introduced into a foraging arena with an existing nest container of 200 workers and two queens of M. antarcticum. The foraging arena was a plastic tray (72 × 34 × 13 cm) covered with fine white sand, with the inner walls painted with fluon. The nest containers of each species were placed at opposite ends of the foraging arena.

We introduced M. antarcticum into the foraging arena 2 days prior to introducing L. humile to allow M. antarcticum to lay claim to the arena. Both species were then monitored at least twice daily. The time until one of the ant species became extinct was recorded. If however, L. humile or both species persisted, we continued to supply food for up to 2 months. After 2 months, we considered the colonies to be established and likely to persist longer, and ended the trials by counting the total number of brood, workers and queens. Prior to statistical analysis, we examined all the data below for heteroscedasticity using Levene's test, and for fit to a normal distribution using the Shapiro–Wilk normality test. The data on propagule survival times did not conform to a normal distribution (Shapiro-Wilk P < 0·001). Therefore, we used two-way univariate Permutational Multivariate Analysis of Variance (permanova) (Anderson 2001). We performed the analysis on untransformed data without standardization, with the Bray–Curtis dissimilarity as the distance measure and 9999 permutations. The explanatory variables were propagule size and food amount; the response variable was survival time.

Some of the 1000 L. humile colonies persisted for the study duration. To determine sublethal effects of M. antarcticum presence on L. humile, a control treatment was included. Controls consisted of 1000 workers and two queens of L. humile in both food treatments, but with no M. antarcticum. Five replicates were used for each colony size per food treatment, except for the colonies of 1000 workers and controls which had three replicates. A two-way anova tested for sublethal effects of M. antarcticum presence on the total number of L. humile workers alive at the end of 2 months and to compare the total number of brood (eggs, larvae and pupa) produced. Analysis was performed on untransformed data, as there was no evidence of heteroscedasticity (Levene's P ≥ 0·259) or non-normality (Shapiro–Wilk P ≥ 0·428).

Aggressive behaviour between L. humile and M. antarcticum was determined by observing both species for 120 s every 10 min for 3 h after introducing L. humile into arenas. The behavioural categories followed Rowles & O'Dowd (2007). Category ‘ignore’ (= 0) included contacts where no interest or aggression was displayed. If interest was shown via antennation, it was considered ‘touch’ (= 1). Contact where both ants retreated from each other quickly was scored as ‘avoid’ (= 2). Where contact included lunging, biting or leg-pulling it was regarded as ‘aggression’ (= 3). Prolonged (> 5 s) incidences of aggression, individuals locked together and active flexing of gasters in the use of stings or chemical defences, was scored as ‘fighting’ (= 4). Individual pairs of heterospecific ants were haphazardly chosen for observation as they moved within ∼1 cm of each other. The species that moved toward the other species and initiated behavioural interactions was recorded. We gave a score to each species in cases where both species initiated contact at the same time. The behavioural data were analyzed using multinomial logistic regression. Individual scores from all replicate arenas were pooled in each treatment. The five categorical outcomes were the aggression ratings (0–4), which were modelled as being dependent on species (L. humile or M. antarcticum), low and high food availability (1 food tray per arena or 10 trays per arena), and propagule size (two queens and 10, 100, 200 or 1000 workers). The reference category for the analysis was M. antarcticum, high food availability and a propagule size of 1000 workers. Only the main effects were included in the model. A full factorial model was undertaken, and although it fitted the data better (Cox and Snell's pseudo R² = 0·273), singularities in the Hessian matrix were encountered and P values and parameter estimates could not be calculated. Ordinal logistic regression procedures were also undertaken, and although results were similar, the fit was much poorer (Cox and Snell's pseudo R² = 0·061) than the presented analysis.

The study was conducted in forest patches at Korokoro (Lower Hutt, New Zealand) (41°220′ S, 174°870′ E) in mid-summer to early autumn 2007. The study area consisted of closed forest patches on hills. On the extreme west side of the site, there was a single L. humile colony at the base of a pedestrian overpass. This colony appeared to be spreading from a larger infestation into new territory at a rate of 2–5 m per month during the 3 months prior to this study. We placed the nearest L. humile experimental colony 15 m away and the farthest one 130 m away from this single colony, keeping the experimental L. humile colonies as near as possible to the invasion front in case of accidental release. Experimental colonies were separated by ≥ 20 m.

We introduced L. humile propagules consisting of 10, 100 or 1000 worker treatments, each with one queen. Each propagule size (colony size) was placed in a clear plastic nest container measuring 14 × 6 × 4 cm. Each container had two nest tubes. There were five 1 mm exit holes on shorter sides of the container 1–2 cm up from the base. Preliminary trials demonstrated that the exit holes were sufficient for workers to exit, but not queens. Several 0·5 mm holes under the flanges of the containers were made for ventilation. Fluon was painted over the upper holes and upper inner walls to prevent these being used as exit holes.

After placing L. humile in the containers, we sealed the exit holes and left the colonies in an incubator for 5 days (with food consisting of 25% sugar water and tuna in oil) to acclimatize. After 5 days, we took the colonies to the field and allowed the ants to exit the containers and forage. Propagule treatments of 100 and 1000 L. humile had six replicates, while due to limiting numbers of queen ants, the propagule size of 10 workers had only four replicates. We also used control colonies consisting of 10, 100 and 1000 workers and one queen per worker category with two replicates each. The control containers had no exit holes, and the ants were fed 25% sugar water, tuna in oil and a single fly (L. sericata) every 3–4 days. These controls tested the influence of abiotic conditions on L. humile. All colonies were checked every 3–4 days to monitor worker presence in the nest boxes and the survival time of the queen (thus, all queens had died at the end of the experiment).

Small temperature data loggers (Onset, Pocasset Massachusetts) were placed in all containers to record the temperature every 30 min and calculate a daily mean of minimum, maximum and mean temperatures for each container. We used these daily mean temperatures to calculate an overall mean of minimum, maximum and mean temperatures for the entire 3 months of this study. A Kruskal–Wallis analysis was used to test for differences in average daily mean temperatures between controls and experimental containers. Pitfall traps (7 cm top diameter × 9 cm deep) were used to sample relative resident ant abundance prior to introducing L. humile and at the study conclusion. Two pitfall traps were haphazardly placed 20–25 cm from each container. Pitfall traps were one-third filled with preservative made from equal parts water and monoethylene glycol with a few drops of detergent to break the surface tension. Traps were removed after 5 days and all ants were identified and counted.

A one-way ancova was used to test for the effect of propagule size on colony survival time. Only after log transformation did the data conform to a normal distribution (Shapiro–Wilk P ≥ 0·625). A Levene's test revealed homogeneity of variance after the log transformation for both variables (P ≥ 0·566). The covariates were the total abundance of two common resident ant species [M. antarcticum and Pachycondyla castanea (Mayr)] sampled from pitfall traps.

For the colony survival analysis, we compared only the 10, 100 and 200 L. humile colony treatments, as all the 1000 worker replicates persisted for ≥ 2 months. permanova test revealed propagule size to have a significant effect on L. humile colony survival (P = 0·037,Fig. 1, see Supplementary Material Table S1 for the full permanova table). Post hoc analysis revealed that a propagule size of 10 workers had a shorter survival time than that of 200 workers (P = 0·008). However, the mean survival time for a propagule size of 10 workers was not statistically different from a propagule size of 100 workers (P = 0·304), and that of a propagule size of 100 workers was not different from that of 200 workers (P = 0·170). The effect of food treatment was not statistically significant (P = 0·125, Fig. 1). There was no significant interaction between food and propagule size treatments (P = 0·682).

In the treatments with 1000 L. humile workers, entire L. humile colonies were never annihilated by M. antarcticum. Instead, L. humile colonies persisted for 61 days (~1464 h) in all replicates for both food treatments. Three of the six M. antarcticum colonies were annihilated by L. humile during this time (two colonies in the high food treatment which persisted for 21 and 31 days, and one colony in the low food treatment which persisted for 25 days). In replicates where both ant species persisted, each species appeared to take turns in completely dominating all food resources on the arena floor lasting for up to several days. The species not foraging would entirely retreat into their nest boxes, defending entrances. Two-way anova indicated that the presence of M. antarcticum in the 1000 L. humile treatments significantly increased the mortality of L. humile workers (F_1,8 = 31·416, P < 0·001). The mean number of L. humile workers alive in control colonies (655 ± 42, means ± SE) was twice that of experimental colonies (313 ± 48) where M. antarcticum were present. We speculate that the ~35% mortality in the control treatment with L. humile alone represented dietary inadequacies, as the diet we provided may not have been equally nutritious for long-term population growth of both species. Increasing food availability did not have a significant effect on the survival of L. humile workers (two-way anova F_1,8 = 2·456, P = 0·156). Nor was the interaction term between the absence of M. antarcticum and the amount of food significant for L. humile worker survival (two-way anova F_1,8 = 0·733, P = 0·417). The mean number of brood in experimental colonies (569 ± 47) was higher than control colonies (390 ± 93), although not significantly so (two-way anova F_1,8 = 2·568, P = 0·148). Nor did the amount of food (two-way anova F_1,8 = 0·688, P = 0·431) or the interaction between these two variables (two-way anova F_1,8 = 0·019, P = 0·895) have a significant effect on brood production by L. humile.

Linepithema humile propagules of 10, 100 and 200 workers were not able to defend their nests against M. antarcticum which entered nests and killed the L. humile workers and queens. During M. antarcticum raids, L. humile queens and several workers hid between spaces in the nesting tubes and aluminium foil or vacated their containers and sought refuge in the foraging arena or on top of their nesting boxes. Linepithema humile queens always attempted escape in the company of several attending L. humile workers (Fig. 2a). At a colony size of 1000 workers, L. humile resisted invasion and attacked workers of M. antarcticum at the M. antarcticum nests entrances. Monomorium antarcticum workers blocked the M. antarcticum nest exit tubes in defence by sitting in an acrobatic manner with gasters and heads pointing towards the attackers, behaviours similar to those which we observed at our field site (Fig. 2b). Despite such aggressive behaviour, L. humile continued to attack M. antarcticum at their exit tubes by biting and pulling the limbs of workers (Fig. 2c). Once exit tubes were cleared, L. humile entered the M. antarcticum nests. In contrast to L. humile, M. antarcticum queens attempted to escape alone which appeared to increase the chance of being killed by L. humile workers (Fig. 2d).

In all observed interactions, more than one L. humile worker was needed to kill individual M. antarcticum. Linepithema humile killed their opponents through joint effort by one or more workers pulling on a limb while another amputated the M. antarcticum appendage. The opponent was then left to die (Fig. 2c). In contrast, M. antarcticum workers always attacked singly, killing L. humile in one-on-one encounters. Monomorium antarcticum appeared to use stings and noxious chemicals against L. humile.

The likelihood ratio test from the logistic regression showed significant effects of species (χ² = 351·04, d.f. = 4, P < 0·001), food availability (χ² = 11·75, d.f. = 4, P = 0·019), and propagule size (χ² = 147·73, d.f. = 12, P < 0·001). Linepithema humile displayed a high frequency of interaction in all behavioural categories (Fig. 3). Conversely, M. antarcticum displayed fewer benign interactions and showed high frequencies of ‘aggression’ and ‘fighting’. Increasing propagule size substantially decreased the frequency of benign interactions and increased the frequency of aggressive behaviours (Fig. 3a). For M. antarcticum, increasing L. humile propagule size led to a less pronounced change in behaviour, with a slight increase in the frequency of benign interactions but still largely aggressive behaviour (Fig. 3b). The model produced a relatively poor fit to the full data set (Cox and Snell's pseudo R² = 0·155), which was primarily due to a low classification success in the low aggression categories. Supplementary Material Table S2 shows the nominal parameter estimates from the model, in which the response of each factor was examined for each behavioural category in comparison with a reference treatment. For example, L. humile was found to be significantly more likely to display a behavioural category of 0 than M. antarcticum (odds ratio = 2·776, P < 0·001). Similarly, L. humile was significantly more likely to display behaviours in categories 1 and 2. Linepithema humile only exhibited less behavioural aggression than M. antarcticum in aggression category 3 (odds ratio = 0·624, P < 0·001; Supplementary Material Table S2).

For L. humile, the frequency of ‘aggression’ and ‘fighting’ was high in low food treatments and only slightly decreased with high food (Fig. 3c). However, there was no difference between low and high food treatments for ‘ignore’ and ‘touch’ behaviours. This effect was observed in the multinomial logistic regression wherein the only significant factor associated with food was observed under behavioural category 2, indicating that ants were significantly less likely to display aggression in high than in low food treatments (odds ratio = 0·654, P = 0·001). In all other behavioural category comparisons, there were no significant differences in aggression displayed between the food treatments (P ≥ 0·504, Supplementary Material Table S2). Similarly to results examining propagule size, M. antarcticulm was highly aggressive irrespective of food availability (Fig. 3d).

Monomorium antarcticum entered L. humile colonies and exterminated some colonies within 3 days of introduction. We also observed L. humile workers in three colonies of 1000 workers fighting with M. antarcticum outside the nest containers. Some 100 and 1000-worker colonies established foraging trails 24 h after introduction. However, these trails gradually disappeared over three to four consecutive days. Four epigeic resident ant species were observed at the field site: Discothyrea antarctica Emery, M. antarcticum, P. castanea, and Strumigenys perplexa (Smith). Only eight D. antarctica and five S. perplexa were observed and were never observed to interact with L. humile, which were therefore not included as covariates in the ancova. The two other species M. antarcticum and P. castanea were more abundant and occurred in 41% (13 of 32) and 56% (18 of 32) of the pitfall traps, respectively.

There was no statistically significant effect of propagule size on the survival times of L. humile colonies when examining for the presence of workers only (one-way ancova F_2,11 = 0·208, P = 0·816, Fig. 4a) or for queens and workers (one-way ancova F_2,11 = 0·062, P = 0·940, Fig. 4b). The average survival times of queens was greater than the mean time until the last worker was observed in all the colony sizes, with queens living for up to 31 days in the absence of workers. No significant effects of the covariates M. antarcticum or P. castanea were observed in either ancova model (P ≥ 0·389).

Some variations in temperature were observed between propagule placement sites in the mean daily average (range = 15·5 to 20·2 °C), the mean daily minimum (13·8 to 16·2 °C), and the mean daily maximum temperatures (17·2 to 27·4 °C). However, no significant differences in any of the temperature variables were observed between treatments (Kruskal–Wallis P ≥ 0·059). The mean temperature of nest containers was marginally but significantly higher in experimental nest boxes (16·99 ± 0·31 °C; mean ± SE, N = 16) compared to controls (15·69 ± 0·45 °C; N = 6) (Mann–Whitney-U = 21, P = 0·049). In control colonies, 100% worker mortality was observed for the colony size of 10 workers. After 37 days and 65 days, queens in two colonies of the 10-worker replicates also died. Workers of the 100 and 1000 control treatments survived until we ended the experiment, but with a mean worker mortality of 66·5% and 67·9%, respectively. These mortality rates in controls were high, but similar or higher rates of mortality have been observed with L. humile even when well-fed and in controlled laboratory environments for a similar period (Grover et al. 2007).

Competition for resources and space is an important factor structuring ant communities (e.g. Hölldobler & Wilson 1990; Sanders & Gordon 2003). In the laboratory study, L. humile survived only slightly longer in high relative to low food treatments. However, this effect was not statistically significant, indicating that the availability of food had no or little role in L. humile propagule survival. Aggression in L. humile has previously been considered unrelated to resources such as food or nest sites (Holway 1999; Zee & Holway 2006). The observed nest raiding of L. humile on M. antarcticum probably reflects the species’ intolerance for co-occurring epigeic ants (DeKock & Giliomee 1989).

Interference and exploitative competition has been studied extensively in L. humile. Most studies have found that through sheer numbers of recruits, L. humile is successful at both forms of competition thereby breaking ‘dominance–discovery’ trade-off (or the trade-off in life-history patterns between the abilities to find and to control resources) (e.g. Holway 1999). However, their success in interference competition may be simply a by-product of their aggressive nature. At high densities, L. humile are more likely to display aggression toward heterospecific ants, in contrast to other ant species (Human & Gordon 1999; Zee & Holway 2006). Even in the complete absence of resources, L. humile behaves aggressively towards other ants (Zee & Holway 2006). Studies exploring behavioural interactions of L. humile with other ant species have shown that L. humile's competitive interference ability is related to numerical abundance (Human & Gordon 1996; Holway 1999; Rowles & O'Dowd 2007). But the large numbers required for this success are probably related to the methods used to kill other ants rather than direct competition for resources. In order to kill workers of other ant species, L. humile requires a group or pack attack (Fig. 2c). Thus, a large colony size enhances the ability of L. humile to kill other ant colonies. As a potentially adaptive consequence of such aggressive behaviour, limiting resources become available.

Propagule size had a much more substantial effect on L. humile behaviour in laboratory experiments than resources moderating levels of aggression. Linepithema humile appeared to be able to determine its own colony size, or relative competitor abundance, and modify the aggressive response. An ability of ants to assess competitor abundance and then modify their behaviour has been previously observed (e.g. Wilson 1976), although not previously associated with invasion success.

At small colony sizes, L. humile showed high frequencies of ‘ignore’, ‘touch’ and ‘avoid’ behaviours suggesting an avoidance or escape strategy, and they were unable to defend their nest against M. antarcticum. They often vacated their nests and attempted to climb over the walls of the foraging arenas. In such circumstances, L. humile queens were always accompanied by workers. Such behaviours may allow small incipient L. humile colonies to escape and survive in a natural environment. If small colonies do survive, they may increase their numbers quickly, as observed in other laboratory studies (Hee et al. 2000).

In their native environment, L. humile and other invasive species like the fire ant Solenopsis invicta occupy an intermediate competitive position wherein they are extirpated from resources by more dominant ants, but also dominate other species (LeBrun et al. 2007). They may have evolved behaviour associated with both assessing competitor abundance and determining when to fight or flee. When fleeing, they show considerable adaptive behaviour with groups of attending workers exiting with queens after colony invasion. Dominant ants may be less able to modify their behaviour, as we observed with M. antarcticum. Monomorium produce venom which is of high repellency towards other species (Andersen et al. 1991; Holway 1999), and they can be very aggressive. Brightwell (2002) similarly observed high levels of aggression by M. antarcticum irrespective of L. humile densities; native ants were generally highly aggressive irrespective of competitor density. Similarly, LeBreton et al. (2007) demonstrated non-adaptive responses of the dominant native New Caledonian ants to the invasive little fire ant Wasmannia auropunctata. The type of behavioural plasticity exhibited by L. humile probably facilitates invasion success and may explain why ant species are over-represented in the list of ‘world's worst invasive alien species’ (Lowe, Browne & Boudjelas 2000). It is possible that the majority of ant species on this list occupy an intermediate competitive position in their native range, thus having a considerable degree of behavioural plasticity. An absence of behavioural plasticity perhaps makes an ant community more at risk from an invasion event, or an individual species less likely to succeed as a biological invader.

Attempts to predict invasion success have had limited success for most organisms (Hulme 2006) including ants (Lester 2005). Increasing numbers of studies invoke propagule pressure as a (or the) fundamental driver of invasions (Richardson & Pyšek 2008). Our results indicate that propagule size may have limited predictive power for species displaying behavioural plasticity. In one of the few studies incorporating behavioural plasticity as a predictive factor in exotic bird establishment, enhanced predictions of invasion success were obtained (Sol, Timmermans & Lefebvre 2002). Behavioural aspects are seldom incorporated into models to predict invasion success and may be hidden from observation without detailed experiments on each potential invasive species. Echoing Holway & Suarez (1999), we recommend that aspects of animal behaviour be given much more consideration in attempts to model invasion success. Without detailed knowledge of the behavioural adaptations that may enhance invader success, an emphasis on managing invasion pathways may be more efficient (Hulme 2006). Areas of high biodiversity are thought to offer biotic resistance to invasion via the abundance of predators and competitors. For example, a diverse invertebrate community on a rocky seashore may efficiently utilize space resources and exclude an invasive species (Stachowicz, Whitlatch & Osman 1999). However, invasive pests such as L. humile appear to modify their behaviour according to local conditions and their establishment may not be related to resource availability. Perhaps the failure to find evidence for biotic resistance elsewhere may be related to behavioural plasticity of invaders in response to local communities. We cannot necessarily rely on high levels of native biodiversity to repel invasions, nor can managers assume that small propagules are unlikely to establish and thrive in a new environment. Conversely, workers in areas of translocations for conservation purposes or practitioners of biological control may not need large introduction propagules if the species of interest displays similar behavioural plasticity.