Influence of slug defence mechanisms on the prey preferences of the carabid predator Pterostichus melanarius ( Coleoptera : Carabidae )

Two-choice experiments on prey preferences of a generalist predator Pterostichus melanarius, and five species of slug prey, were conducted in the laboratory. Different preferences of P. melanarius for each of the slug species are described. They are interpreted as the outcome of differing slug species-specific defence mechanisms. The influence of hunger level, temperature, day/light period, condition of slugs and beetles, weight of slugs and beetles, and the sex of beetles were controlled experimentally or statistically. The order of slug species preference for predation by P. melanarius was: Deroceras reticulatum (Agriolimacidae), Malacolimax tenellus, Lehmania marginata (Limacidae), Arion distinctus and A. subfuscus (Arionidae). Efficiency of slugs’ speciesspecific defence mechanisms reflected their phylogeny. Defence mechanisms of slugs from the superfamily Arionoidea were significantly more effective at deterring an attack of non-specialised ground beetles than the defence mechanisms of slugs from Limacoidea superfamily. P. melanarius significantly preferred Agriolimacidae to Limacidae, and Limacidae to Arionidae. Slug species was the strongest factor influencing prey preferences of P. melanarius amongst slug prey. Surprisingly, this preference was much more significant than the slug weight. Weight and sex of P. melanarius had no impact on its prey preference.


Field site and sample collection
The study was carried out in South Bohemia, Czech Republic, from July to September 2002.The beetles were collected by pitfall trapping from a cornfield near eské Bud jovice, South Bohemia (49°0´N, 14°30´E, alt.400 m).Males and females were kept apart in glass containers (20 × 20 × 30 cm) filled with 5 cm of plant litter in a controlled environment (16L : 8D; 20 ± 2°C).They were fed with darkling beetle maggots (Tenebrio molitor) for several days.Prior to the experiment, to ensure that the beetles were of a similar nutritional state, they were put into individual container and fed ad libitum.They were than starved for 7 days (Symondson, 1997).Slow moving or injured beetles were not used for the experiment.
The following slugs were tested: Deroceras reticulatum Müller, 1774, Lehmania marginata Müller, 1774, Malacolimax tenellus Müller, 1774, Arion distinctus Mabille, 1868, A. subfuscus Draparnaud, 1805.They were collected in the vicinity of eské Bud jovice from an oilseed-rape field (D. reticulatum), coniferous or mixed forest (L.marginata, M. tenellus) or suburban gardens (A.distinctus, A. subfuscus).The species were kept separately in round glass containerss (8 cm in diameter, 25 cm deep), because they (especially M. tenellus) showed a surprisingly large mortality when kept in mixed-species groups.The containers were filled with plant litter and stored in a controlled environment (16L : 8D; 20 ± 2°C). A. distinctus, A. subfuscus and Deroceras reticulatum were fed on cabbage leaves, Lehmania marginata and M. tenellus on mushrooms of Russula spp.Food supply was replaced every 3rd day.Placing the slugs on saturated filter paper (>1h) prior to weighing ensured that all slugs were uniformly and fully hydrated.Usage of fully hydrated slugs maximised their mucus production and hence their defensive ability.Injured slugs were not used for the experiment.

Two choice experiment
At the start of the experiment a single beetle was placed in a 9-cm diameter Petri dish with two slugs.The dish was lined with filter paper and moistened daily.The sex and weight of the beetle and weight of the experimental slugs were recorded before the experiment.The experiment was carried out in a controlled environment (16L : 8D; 20 ± 2°C) and the status of the slugs, eaten or not eaten, was checked every 7 h for 3 days.Each trial was considered completed when the first slug was eaten.The slug was also considered eaten, if it was dead and heavily damaged by the beetle.The trials in which both slugs were eaten or one of them was dead but undamaged were disregarded, as were trials where the beetle died.In total, there were 10 pairwise combinations of slug species, corresponding to 400 slug and 200 beetle individuals in 200 successfully completed trials.The sex ratio of ground beetles was equal for each combi-nation of slugs (10 males and 10 females).The relatively small Petri dishes did not enable the slugs to fully employ their species-specific antipredator strategies.For instance, the tailwagging and fast escaping species such as D. reticulatum were disadvantaged compared to Arion spp., that rely on their skin and mucus when being attacked by P. melanarius (Pakarinen, 1993a, b).However, the Petri dishes highlighted speciesspecific differences in defence mechanisms and so ensured that the results were not disturbed by other external factors.

Statistical processing
The numbers of eaten slug individuals for all pairs of slug species were compared (against equal proportions) by Chisquare tests. 2 × 5 Contingency table was used to test total numbers of eaten slug individuals of each species against the null hypothesis that all slug species were eaten in equal proportions.
To assess if slug species, weight, or sex and weight of the beetle affected whether the slug was eaten, generalised linear modelling (GLM) multiple-regressions were used.GLMs allow for both continuous and categorical predictors, and for nonlinear responses of the dependent variables, which are transformed using specific link-functions (Venables & Ripley, 2002).The regressions were computed in the program S-plus (S-Plus, 1999(S-Plus, , 2000)), assuming a binary distribution of the dependent variable (eaten vs. not-eaten) and using the link function "logit".The model construction was as follows: (1) defining a null model, which did not contain any explanatory predictors; (2) constructing single-term regressions for all potential explanatory variables; (3) constructing multiple regressions, using stepwise addition of individual terms, and comparing them with the null model using both F-test and the Akaike information criterion (AIC, weights increasingly better fit of higher-order models by their increasing complexity).Models for effects of slug family and superfamily were constructed by the same routine.

Total numbers of eaten/uneaten slugs
Numbers of slugs eaten in each pairwise experiment are shown in Table 1.D. reticulatum was eaten in significantly greater numbers than all other slugs tested.M. tenellus was eaten in significantly greater numbers than other slug species except D. reticulatum.There were no significant differences in the number of slugs eaten amongst L. marginata, A. distinctus and A. subfuscus.
Significant differences in the total numbers of slugs of each species eaten were found for A. subfuscus, D. reticulatum and M. tenellus.Differences in total numbers of eaten individuals of L. marginata and A. distinctus eaten were only marginally significant (p = 0.060 and p = 0.075; see Table 1 1.Numbers of slug individuals eaten for each combination of slug species in the pairwise experiments (compared by Chisquare tests against equal proportions).TN -total numbers of slugs of each species eaten in the whole experiment (2 × 5 Contingency was used to test against the null hypothesis that all slug species were eaten in equal proportions).Significant differences are indicated by an asterix.

Multiple regressions
The single-term regressions of slug being/not-being eaten against all possible predictors revealed that slug species influenced the prey preferences of P. melanarius very significantly (Table 2).The slugs were again preferred in the following order: D. reticulatum, M. tenellus, L. marginata, A. distinctus, A. subfuscus (Fig. 1A).On the other hand, there was no separate effect of slug or beetle weight, or sex of the beetle.However, the best model constructed by the multiple regression contained slug weight in addition to slug species.This, together with the fact that slug weight was not significant in single-term regression, suggests that within species the beetle preferred lighter slug individuals (Fig. 1B).Inclusion of further predictors (beetle sex and weight) increased the AIC values, indicating that the predictors were redundant (see full model in Table 2).Moreover, a comparison of the fits of the final and full models did not reveal any significant differences (F = 0.53, d.f.= 2, 392, p = 0.59, delta AIC = 2.97).This indicated that the species of slug had a strong separate effect on prey preferences of the beetle even when the variability due to weight of the slug (and sex and weight of the beetle) were included in the model, acting in effect as co-variables.This indicates the importance of species-specific defence mechanisms of the slugs in carabid beetle prey selection.
The models in which family/superfamily were considered instead of slug species (Table 2) showed significant effects of these taxonomic rankings.Slugs from the superfamily Limacoidea (Limacidae + Agriolimacidae) were significantly preferred to slugs from the superfamily Arionoidea (Arionidae).Slugs from the Agriolimacidae family (D.reticulatum) were significantly preferred to slugs from the Limacidae family (L.marginata and M. tenellus).Slugs from the Limacidae family were significantly preferred to the slugs from the Arionidae family (A.distinctus and A. subfuscus).Interestingly, slug weight was a variable in the model in which slug family was considered, but was redundant in the model in which superfamily level was the focus.This suggested that the difference in readiness of the beetle to prey on Limacids vs. Arionids outweighed the preferences due to prey weights.

361
1 Deviance in the data not accounted for by the model; 2 Deviance in the data accounted for by the model;  2. Regression models showing the influences of slug species and weight, and sex and weight of the predatory beetle Pterostichus melanarious, on prey preferences of the beetle.A generalised linear modelling procedure is used with assumed binary distribution of the dependent variable (a slug either eaten or not eaten).The three multiple regression models (constructed via stepwise addition of significant predictors) consider slug species, slug family and slug superfamily as possible explanatory variables.Fig. 1.Multiple GLM regression of the effects of (A) slug species (categorical predictor) and (B) slug weight (continuous predictor) on the slug being either eaten or not eaten (with assumed binary distribution).Model with slug species as one of the independent variables, see Table 2 for model statistics.The plots show partial effects of individual predictors (full lines) with respective standard errors (dotted lines).The empty circles are individual cases, the internal tickmarks on x-axis show frequencies of respective values of independent variables.Note that the values on y-axis are fitted using the logit-link function.

DISCUSSION
The present study documents how a non-specialised ground beetle preyed upon different slug species selectively as a consequence of species-specific defence mechanisms of the slugs.Results of this study seem to be consistent with the hypothesis that the efficiency of slug defence mechanisms varies amongst species.Importantly, all other known (Wareing & Bailey, 1985;Ernsting & Vanderwerf, 1988;Wheater, 1988aWheater, , b, 1991;;Wallin & Ekbom, 1994;Hon k, 1997;Ayre, 2001;McKemey et al., 2001) or possible confounding influences (i.e., hunger level, temperature, day/light period, condition of slug or beetle, weights of slugs and beetles, sex of the beetles) were controlled for experimentally or statistically.This further supports the notion that the different preferences of P. melanarius were due to differences in efficiency of species-specific defence mechanisms of slugs.In several earlier studies (Thiele, 1977;Ernsting & Vanderwerf, 1988;Wheater, 1988a;Ayre, 2001;McKemey et al., 2001), the carabids were considered to be generalist predators, which feed on almost any suitable prey they meet and are able to subdue.The authors mentioned above considered weight of the slug to be the strongest factor influencing prey preferences of the ground beetles eating slug-prey.
P. melanarius preferred slugs from superfamily Limacoidea (Limacidae, Agriolimacidae) to slugs from superfamily Arionoidea (Arionidae); Agriolimacidae (D. reticulatum) to Limacidae (M.tenellus, L. marginata); and Limacidae to Arionidae (A. distinctus, A. subfuscus).It follows that efficiency of slug defence mechanisms reflects phylogenetic positions of the slugs as postulated by Wiktor (1984) and Reischütz (1999).This should be taken into account, especially in studies examining the potential of ground beetles as biocontrol agents against slug pests.In particular, D. reticulatum, as the most preferred slug prey in this study, might not be a suitable model for other slug pests.
Most limacids are fast-moving slugs with soft skin and a slender body (Godan, 1983;Pakarinen, 1993b) that react to an attack by the production of alarm mucus and tail-wagging, accompanied by a burst of speed.The tailwagging engages attackers attention giving the victim time to escape or hide.Limacids lift up the anterior part of their body and distend their tentacles to orientate towards shelters.On the other hand, slow moving Arionids respond by stopping and withdrawing the head under the mantle (Rollo & Wellington, 1979), relying on their tough skin for protection (Godan, 1983, Pakarinen, 1992).Heparan sulphate was identified in mucus of Arion lusitanicus, A. ater and A. rufus (Wright et al., 1997;Ditrich, 2001).In the Arionidae, autotomy has been demonstrated only in the Anadeninae subfamily (Pakarinen, 1993a).
In this study, the slugs had no chance to hide effectively and their escape strategies were constrained by the 9cm diameter of the Petri dish.The most preferred species (D. reticulatum and M. tenellus) are able to autotomize their tail when attacked (Pakarinen, 1993a).
Wagging of body parts can decrease their defence ability if their escape fails and the predator attacks them again in the dish.However, the significant differences in preference of P. melanarius between D. reticulatum and M. tenellus can not be explained in this way, because these species are similar in their autotomy/escape strategies (Pakarinen, 1993a).Moreover, attempted escapes of a slug from the place of attack in Petri dish increased the probability of P. melanarius encountering another slug species, which could be regarded as a successful use of escape strategy by the attacked slug.Despite this bias, usage of relatively small Petri dishes reduced the efficiency of hide/escape strategies of the slugs.This emphasized species-specific differences in the efficiency of slug defences, and highlighted their importance for Carabidbeetle prey selection.
The species-specific defence of slugs had a stronger effect on prey preference of P. melanarius than the prey weight, which had a significant effect only within slug species.The probability that the predator will eat a slug decreased log-linearly with slug weight.This corresponds with findings in other studies (Ernsting & Vanderwerf, 1988;Wheater, 1991;Ayre, 2001;McKemey et al., 2001).However, all these studies dealt in fact only with a partial prey-size preference.The rates of feeding on slugs of different sizes differ between the field and laboratory conditions as a consequence of slug-size related behavioural differences and other factors, which can only be simulated in the laboratory (Mckemey et al., 2003).
Sex of P. melanarius had no effect on its preypreference.This corroborates the findings of Kielty et al. (1999) and Digweed (1993), who studied feeding preferences of carabid beetles on aphids and collembolans, and terrestrial snails.The missing effect of sex might liberate further researchers from considering it in experimental design, and thus considerably reduce the requirements for material. ).