Naturwissenschaften (2012) 99:833–842 DOI 10.1007/s00114-012-0965-6

ORIGINAL PAPER

Social prophylaxis through distant corpse removal in

Lise Diez & Jean-Louis Deneubourg & Claire Detrain

Received: 7 May 2012 /Revised: 21 August 2012 /Accepted: 23 August 2012 /Published online: 7 September 2012 # Springer-Verlag 2012

Abstract Living in groups raises important issues concerning Introduction waste management and related sanitary risks. Social insects such as ants live at high densities with genetically related One of the major drawbacks of social life is the increased individuals within confined and humid nests, all these factors risk of exposure to pathogens due to high population densi- being highly favorable for the spread of pathogens. Therefore, ties and frequent contact rates with conspecifics (Anderson in addition to individual immunity, a social prophylaxis takes and May 1979; Schmid-Hempel 1998). In social insects, the place, namely, by the removal of risky items such as corpses breaking of a pathogen into the colony can become life- and their rejection at a distance from the nest. In this study, threatening since most pathogens will proliferate under the we investigate how Myrmica rubra workers manage to reduce warm and humid conditions found within the nest and will encounters between potentially hazardous corpses and nest- easily spread out within those groups of genetically related mates. Using both field and laboratory experiments, we de- individuals. scribe how the spatial distribution and the removal distance of At the individual level, there are two lines of defense waste items vary as a function of their associated sanitary risks against pathogens that are common to both solitary and (inert item vs. corpse). In the field, corpse-carrying ants social insects: the cuticular barrier and the innate immunity. walked in a rather linear way away from the nest entrance First, the tough cuticle covering the entire body of insects as and had an equal probability of choosing any direction. There- well as the thin cuticular layer on mouthparts, pharynx, and fore, they did not aggregate corpses in dedicated areas but intestinal tract act as efficient physical barriers against scattered them in the environment. In both field and laboratory invaders such as bacteria, viruses, and fungi. But, if patho- experiments, ants carrying corpses dropped their load in more gens break this first line of defense, innate immunity kicks remote—and less frequented—areas than workers carrying in by triggering a physiological defensive reaction in the inert items. However, for equidistant areas, ants did not avoid hemolymph of infected individuals. Innate immunity dropping corpses at a location where they perceived area involves a cellular response (i.e., phagocytosis and encap- marking as a cue of high occupancy level by nestmates. Our sulation of pathogens) and a humoral response (production results suggest that ants use distance to the nest rather than of antimicrobial compounds or melanization) (Schmid- other occupancy cues to limit sanitary risks associated with Hempel 2005; Siva-Jothy et al. 2005; Feldhaar and Gross dead nestmates. 2008). In insects, adaptive immune responses have also been reported: a prior exposure to a pathogen will launch a Keywords Necrophoresis . Myrmica rubra . Ants . Social more efficient defense the next time the same pathogen insects . Area marking . Threshold distance attacks and will improve protection of the individual against further challenges (e.g., in social insects such as , Rosengaus et al. 1998;bumblebees,SaddandSchmid- Communicated by: Sven Thatje Hempel 2006; and ants, Konrad et al. 2012). Despite the L. Diez (*) : J.-L. Deneubourg : C. Detrain potentially higher exposure to pathogens due to group liv- Unit of Social Ecology, Université Libre de Bruxelles, ing, social insects such as honeybees surprisingly lack sev- CP 231, Bd du Triomphe, 1050 Brussels, Belgium eral genes involved in immune responses that are present in e-mail: [email protected] nonsocial insects such as fruit flies and mosquitoes (Evans 834 Naturwissenschaften (2012) 99:833–842 et al. 2006). This suggests that immunity in social insects harvester ants (Hölldobler and Wilson 1990), or occasion- does not rely only on physiological and physical defenses at ally of inflorescences near Myrmica schencki and Myrmica the individual level but rather on lines of defenses against rubra nests entrances (Czechowski et al. 2008). Likewise, pathogens specifically achieved at the colony level (Wilson- corpses can be occasionally gathered at a location—so- Rich et al. 2009). called a cemetery. In the field, such cemeteries are mainly Social insects proved to be very successful in colonizing observed in some species living in populous colonies with almost all terrestrial ecosystems, and a reason of this success corpses being put with other wastes on the refuse pile like in is that they developed efficient ways to prevent or to limit army ants (Hölldobler and Wilson 1990)orleaf-cutting mortality due to pathogens through collective defenses species (Moser 1963). In other species, corpses can be (Cremer et al. 2007). First, social insects can upregulate scattered in the nest surroundings and individually aban- their immune response while the adult population is rapidly doned at remote locations (Howard and Tschinkel 1976). increasing (Ruiz-González et al. 2009). Social insects also From a prophylaxis perspective, the rejection of wastes and display a wide range of prophylactic behaviors (Cotter and corpses in areas that are remote from the nest or weakly Kilner 2010). Autogrooming and allogrooming (Rosengaus foraged could be a simple and efficient way to reduce encoun- et al. 1998) combined with the secretion of antibiotic com- ters between workers and those potentially hazardous items. pounds produced by the metapleural glands (Hart et al. To achieve this spatial segregation of corpses, transporting 2002) are common ways to remove most pathogens from ants should decide to drop their load at locations where the the cuticle. Another line of prophylaxis consists in reducing local density or activity level of nestmates is expected to be the density of pathogens within the confined space of the low. Therefore, corpses carrying ants should be able to assess nest. This is achieved by avoiding to dig the nest chambers nestmate density either through the number of interactions into infected soils (Drees et al. 1992), by using antimicrobial with congeners (Gordon et al. 1993; Detrain and Deneubourg material into the nest such as propolis in honeybees (Bankova 2009) or through the density of footprint marks (Devigne and et al. 2000) or tree resin in wood ants (Chapuisat et al. 2007), Detrain 2006; Lenoir et al. 2009) that are passively laid by and by removing food wastes (Zeh et al. 1999; Hart et al. walking ants. Such an area marking is used by workers as a 2002). A last way to prevent the spread of pathogens is density cue to tune their foraging behavior (e.g., in Lasius through the spatial isolation of dead or diseased individuals niger workers, Devigne et al. 2004) and could influence the from the healthy nestmates. In honeybees, dead or diseased behavior of corpse-carrying ants. adults and larvae are dropped outside the hive (Visscher In this paper, we investigate how ants manage sanitary 1983). In termites, corpses are buried into the nest (Crosland risks associated through the removal of corpses. We deter- et al. 1997; Ulyshen and Shelton 2012), while in most ant mine whether hazardous items—i.e., corpses—or inert items species, they are rejected and transported away from the nest are spatially segregated from areas usually explored by ants (Wilson et al. 1958; Choe et al. 2009; Diez et al. 2011). outside the nest. Therefore, we perform both field and In ants, necrophoresis—i.e., corpse removal—can be an laboratory experiments. First, we observe at which locations important issue for the survival of the colony. Indeed, ants the common red ant, M. rubra (L.), chooses to remove and that died from infectious diseases are potentially contagious: drop dead nestmates in the field. We investigate which they may contaminate nestmates staying in their vicinity parameters, such as the distance to the nest, the level of inside the nest or foragers fortuitously encountering corpses occupancy by nestmates, or the local abiotic conditions, are on commonly explored areas. Some studies have reported the most influential on the decision of ants to deposit corp- cases in which diseased or moribund ants leave the nest on ses. Then, we study during laboratory experiments whether their own and die away from their congeners (Heinze and area marking is used by carrying ants as a cue for ant Walter 2010; Bos et al. 2012). When workers die inside the occupancy and accordingly shape the spatial distribution nest, there are several ways for the colony to set them apart of corpses or inert items outside the nest. from healthy congeners and larvae. Dead individuals may be buried into the nest (Renucci et al. 2010) or may be put with other wastes in refuse chambers as in the case of leaf cutting Material and methods ants (Bot et al. 2001). Besides, several ant species exhibit necrophoresis by actively rejecting dead workers outside the Field experiments nest as they usually do for any other waste items. The rejection of waste items is known to lead to the formation Ant species and field site of piles around the ant nest, providing those items are available in high quantities such as heaps of scavenged The red ant M. rubra is a common species inhabiting forests insects (Czechowski et al. 2009), fungal wastes in leaf- or open areas and is widespread in Northern Europe. This cutting ants (Hart and Ratnieks 2001), seed husks in polygynous and polydomous species became invasive in Naturwissenschaften (2012) 99:833–842 835

Northern USA and Eastern Canada (Groden et al. 2005). M. trajectories were characterized by a straightness index given rubra is mainly carnivorous but also forages for other food by the ratio between the beeline distance from the first to the sources such as aphid honeydew or floral sugar exudates. last visited point over the total length walked during the trip Experiments were performed in August 2009 and April (Batschelet 1981). We calculated a straightness index only 2010 on a polydomous colony with four nest entrances for ants that walked for at least 4 min (for which we located in a semiopened area of the campus of Gembloux recorded at least four location points). The total length Agro-Bio Tech, Belgium (50°33′49″ N; 4°41′40″ E). The walked was estimated by summing the walked distance experimental site we used had a 1 % slope oriented to the during each 2-min interval. Speeds of corpse-carrying and north. We delimited a study area to a 6×6-m square that was exploring ants were calculated using the total length walked centered on the main entrance and including the three other divided by the total time of the trip. Locations of corpse nest entrances. Within this area, we marked 144 observation deposits were recorded by measuring their distance from the squares resulting in a 0.5×0.5-m grid by pinning down nest entrance with an accuracy of 1 cm and their direction yellow nails at each corner. with a compass (accuracy, 2°). Temperature at the ground level was measured with an infrared thermometer (Oakton Field protocol TempTestr IR; accuracy, 1 °C) and light intensity with a photometer Testo 545 (accuracy, 1 lx). In addition, we We compared the trajectories of corpse-carrying workers assessed the spatial distribution of ants over the studied area and exploring ants in the field. During 7 days in August to see whether corpses were deposited in areas that are 2009 and 4 days in April 2010, we observed, on average, six weakly or regularly visited by workers. This was done every corpse-carrying and six exploring ants each day, between 2 days during the week of field experiments in August 2009 10 am and 5 pm. To trigger corpse-carrying activity, we by looking carefully at each of the 144 observation squares placed a nestmate corpse in front of one of the four tested during 20 s and counting the number of foraging ants in nest entrances. Corpses were nestmates that were killed by each 0.5×0.5-m grid square. This gave us an overview on freezing at −24 °C during 30 min. They were left at ambient how the occupancy level changed with distance and orien- temperature for 7 days before being tested. Preliminary tation around the nest entrance. experiments actually showed that, after a 1-week time lapse, all dead ants were recognized as corpses and accordingly Laboratory experiments were quickly rejected by workers. As soon as an ant took the corpse, we followed it until it dropped the corpse and Collection and maintenance of ants abandoned it for at least 30 s before returning back to the nest. We also put every day an inert item—i.e., a piece of Three M. rubra colonies were excavated from earth banks of vermiculite of similar size as a corpse—near the nest en- a semiopened area in the campus of Gembloux Agro-Bio trance. But, since ants never transported such items for a Tech, Belgium. In the laboratory, colonies were kept in greater distance than 1 cm, we had no data about the spatial plaster nests (Janet type; 100×100×2 mm) connected to a distribution or the behavior of ants transporting inert waste. foraging arena (43.5×35 cm) of which borders were coated As regards ants carrying corpses, we compared their behav- with Fluon (polytetrafluoroethylene) to prevent ants from ior to that of exploring ants that were not involved in the escaping. Each of the 3 nests contained 3 gynes, 200–250 removal of dead corpses. Therefore, we randomly chose workers, brood covering 10 to 15 % of the nest area, which ants walking unloaded near the nest entrance, at an average is a colony composition commonly found in the field (Elmes distance of 30±16 cm. In order to make reliable compar- 1973). Laboratory conditions were kept at 23±2 °C and 40± isons under similar conditions, an exploring ant was ob- 5 % HR, with a constant photoperiod of 12 h/day. Each nest served immediately after the following of a corpse- was provided with a sucrose solution (ad libitum, 0.3 M) carrying ant and was chosen near the same nest entrance. and one mealworm (Tenebrio molitor) per day as a protein Each noncarrying ant was followed until it entered the nest and lipid supply. or disappeared within the litter for more than 2 min. We observed 63 corpse-carrying and 64 exploring ants of which Experimental protocol we recorded: (1) its trajectory by marking its location every 2 min, (2) the time elapsed until the corpse was dropped or A filter paper (43.5×35 cm) of which half was covered with the time of observation for exploring ants, (3) the location a plastic sheet (43.5×17.5 cm) was placed for 24 h over the where the corpse was dropped or the point along the explor- whole foraging area of the tested colony. Exploring ants ing ant’s trajectory that was the more distant from the tested were allowed to walk over the paper and to passively lay nest entrance, and (4) the light intensity and temperature at footprint marks. During this exploration phase, water, sugar this location. The corpse-carrying and exploring ants’ solution, and dead mealworms were placed in the middle of 836 Naturwissenschaften (2012) 99:833–842 the foraging area—on the line between marked and un- Comparisons between two samples were thus performed marked area—in order to avoid asymmetries in the density with a Wilcoxon rank sum test. Comparisons between of ants and hence in the level of area marking by footprints. more than two samples were performed using Kruskal– Likewise, we prevented orientation biases due to visual cues Wallis test, followed by Behrens–Fisher test for multiple by placing light sources (four neon lamps OSRAM L 18 W) comparisons when it was needed. Temperature and light ex- in a symmetrical way and by surrounding the setup with posure followed a normal distribution (after a log transforma- white walls. Before the beginning of the experiment, we tion for light exposure). Then, comparisons of means were removed the plastic sheet as well as food and water supplies. performed using ANOVAs. At this time, only one half of the paper was chemically In the laboratory experiment, we carried out 10 trials on marked while the other half that was previously covered each of the 3 colonies. For each trial, we put 10 inert items and by the plastic sheet remained free of any area marking. We 10 corpses inside the nest. Distribution of corpse deposits counted the number of ants present on the right and the left between the marked vs. unmarked areas were analyzed by sides of the area every 5 min during 15 min. These four taking into account only the trials with a minimum of five observations allowed us to estimate the average distribution transports of corpses or five transports of inert items. For of ants on both sides of the area before the experiment. At corpse transports, 22 out of 30 experiments and for inert items, the beginning of the experiment, 10 inert items (vermiculite) 18 out of 30 experiments fulfilled these conditions and were were put inside the nest by releasing them into a hole made used for further analysis. The variables “distance to the nest,” in the center of its glass roof. We chose vermiculite pieces “distribution on right vs. left side,” and “distribution on that were sized around 3 mm3 and weighted 9.13×10−4± marked vs. unmarked side” were not normally distributed. 4.24×10−4g. As vermiculite pieces were not significantly For the distance to the nest, comparisons between two samples heavier than corpses that weighed 9.58×10−4±2.21×10−4g were performed with the Wilcoxon rank sum test and compar- (Wilcoxon rank sum test, W044, N020, P00.677), ants isons between more than two samples were performed using could easily remove them. After 30 min, those inert items the Kruskal–Wallis test. The distributions of items either on were rejected by ants or moved far away from the insertion the right vs. left sides or on the marked vs. unmarked sides hole. Then, we added 10 corpses inside the nest by intro- followed a binomial distribution. In order to compare these ducing them through the insertion hole. Corpses were nest- distributions in the different colonies, we used a generalized mates killed in the same conditions as described in the field linear model (binomial distribution, logit link). The distribu- experiment. During 5 h, we recorded every 5 min the spatial tions of deposits or the distribution of noncarrying ants over location (on marked vs. unmarked side of the area) of both the right vs. left sides and marked vs. unmarked sides of the types of items (i.e., inert items and corpses) that were area were compared with expected binomial distributions. dropped before removing it from the area to avoid second- Those theoretical distributions were obtained by computing ary transports or pile formation. 1,000 simulations, each one corresponding to a set of random binomial distributions that had the same number of trials than Data analysis the total amount of transported items in our experiments. Theoretical curves were compared to experimental data using In field experiments, we tested one colony with four the Kolmogorov–Smirnov tests. All tests were two-tailed with entrances during two observation periods, one in August alpha set at 0.05. Average values are all provided with stan- 2009 (N040) and the second in April 2010 (N023). dard deviations. We used software R 2.12.2 (http://www. Location of deposits and ants’ trajectories were normal- r-project.org)forallstatisticalanalyses,modeling,and ized by being represented as originating from a same artwork. single nest entrance. The angles between north and the direction of the most remote point reached by exploring or transporting ants were also measured. These angles Results were circular data of which the distribution was com- pared by using an analysis of variance (ANOVA) test Field experiments for circular data as they followed a von Mises distribu- tion (a circular analog to normal distribution). For non- In our field experiments, we compared the trajectories of circular data, we tested the variables “straightness,” exploring ants with those of ants carrying corpses until they “distance to the nest,”“speed,”“ant densities,”“temper- dropped their load. Exploring ants foraged around the nest ature,” and “light exposure” for normality using Pear- and their trajectories showed frequent changes of direction. son’schi-squaretest.Straightness,distancetothenest, By contrast, corpse-carrying ants headed away from the nest speed, and ant densities were not normally distributed. and walked in a rather linear way. Indeed, the trajectory of Naturwissenschaften (2012) 99:833–842 837 undertakers was characterized by a straightness index (0.807± significantly longer than the mean distance of locations at 0.197, N035) that was significantly higher than for exploring which each of the 64 exploring ants were seen (42.6± ants (0.431±0.261, N039) (Wilcoxon rank sum test, W0549, 26.2 cm, N064) (Wilcoxon rank sum test, W01,387, P< P<0.0001). The average speed of transporting ants (8.4± 0.003). Thus, transporting ants dropped corpses further than 7.8 cm/min, N063) and the average speed of exploring ants areas mainly foraged by the colony, but not further than the (8.8±6.7 cm/min, N064) were not significantly different most remote areas explored by nestmates. The entrance of (Wilcoxon rank sum test, W01,845, P00.411). the nest had no effect on the distance of corpse deposits During our experiments, we never observed “cemeteries” (Kruskal–Wallis test, χ205.73, df03, P00.125) or on the or aggregation of corpses: dead nestmates were all scattered maximal distance of exploring ants (Kruskal–Wallis test, in the nest surroundings. In order to characterize the spatial χ204.41, df03, P00.221). Likewise, the observation period distribution of corpse deposits (Fig. 1a)andcompareit (August or April) had no effect on the distance of corpse to the distribution of exploring ants (Fig. 1b), we ana- deposits (Wilcoxon rank sum test, W0437, N1040, N2023, lyzed their angular and radial distribution around the nest P00.230) or on the maximal distance of exploring ants entrance. (Wilcoxon rank sum test, W02,102, N1040, N2023, P0 The distribution of the angular direction of corpse depos- 0.100). its was uniform (Watson’s test for circular uniformity, test The percentage of corpses dropped by carrying ants can statistic, 0.160, N063, P00.187). Thus, ants had an equal be described as a function of the distance from the nest. To probability of choosing any direction when removing corp- describe this relation, we used a simple model that fitted our ses away from the nest. The angular distribution of the most experimental data. We observed that the proportion of corp- remote locations visited by exploring ants was also uniform ses still transported by ants at a distance d from the nest N(d) (Watson’s test for circular uniformity, test statistic, 0.177, (Fig. 2a) decreased in a nonlinear way and could be de- N063, P>0.05), suggesting that there was no preferential scribed by the following sigmoidal curve: direction for the exploration of the nest surroundings. More- 1 over, the angular distributions of locations visited by ex- N d 1 ð Þ¼1 eη d T ð Þ ploring ants and by corpse-carrying ants were not þ ðÞÀ significantly different (circular ANOVA, F1,12600.046, P0 In Eq. 1, T is a threshold corresponding to the distance at 0.83). This indicates that corpse-carrying ants did not which 50 % of transporting ants had dropped their corpse. The choose directions that were less visited by exploring ants. parameter η indicates the steepness of the sigmoidal curve Corpse-carrying ants dumped their load, on average, at around values close to the distance threshold. High η values 72.2±60.2 cm (N063) from the nest entrance. The distances mean that ants carrying corpses are very sensitive to small at which corpses were deposited were highly variable, as it differences between their actual distance d and the threshold was comprised between 7 and 289 cm from the nest en- distance T. A steep curve indicates that even a small increase trance. On average, the distance of corpse deposits was not of distance around the threshold value T makes corpse- significantly longer than the maximum distance at which carrying ants very likely to drop their load. The best fit ants explored the nest surroundings (62.0±40.1 cm, N064) between the experiment and the model corresponded to values (Wilcoxon rank sum test, W01,877, P00.228) but it was of T052.62±0.38 cm and η00.0599±0.0015.

Fig. 1 a Spatial distribution of corpse deposits around the nest a b entrance. Filled black circle nest entrance, black circles corpse deposits. b Spatial distribution of the most remote locations of exploring ants around nest entrance. Filled black circle nest entrance, black 0 100 200 South (cm) triangles the most remote point − visited by an exploring ant North 100 − 200 −

−300 −200 −100 0 100 200 −300 −200 −100 0 100 200 West−East (cm) West−East (cm) 838 Naturwissenschaften (2012) 99:833–842

Fig. 2 a N(d): proportion of d corpses still transported at a (a) distance d from the nest entrance. Black circles experimental data values, gray line curve fitting using function 1 with parameters set as η0 0.0599 and T052.62. b P(d): probability for an ant of dropping its corpse when reaching the distance d. c Ant density as a function of the distance from the nest. Black lines represent the median value, boxes represent the first and third quartiles, and ) Proportion of ants still carrying a corpse at distance d 0.0 0.2 0.4 0.6 0.8 1.0 whiskers represent 1.5 times the ( interquartile range of the box. N 0 50 100 150 200 250 300 Letters above each bar refer to significant differences of ant density (P<0.05) between each (b) class of distance from a nest entrance using a post hoc Behrens–Fisher multiple comparisons ) Probability of dropping a corpse d ( P 0.00 0.02 0.04 0.06 0.08 0 50 100 150 200 250 300

N=8 (c) a 2

N=24 b N=26 N=27 N=27 N=32 b c ccc Number of ants/m 051015

0 50 100 150 200 250 300 Distance to the nest (d in cm)

The derived function of N(d) gives the probability for a Out of Eq. 2, one can describe how the individual prob- corpse to be dropped at a distance d: it is directly related to ability P(d) of each corpse-carrying ant to drop its load in the number of ants still transporting a corpse N(d) and to the function of its distance from the nest: probability for each transporting ant P(d) to drop its load at N 0 d η this distance d: P d À ð Þ 3 ð Þ¼ N d ¼ 1 e η d T ð Þ ð Þ þ À ðÞÀ N d According to Eq. 3, a corpse-carrying ant had a higher ð Þ N d P d 2 dd ¼ ð ÞÂ ð ÞðÞ probability of dropping its load with increasing distance from Naturwissenschaften (2012) 99:833–842 839 the nest, and this probability reached a maximal plateau value the right and left sides was not colony-dependent (GLM,

(η00.0599) at approximately 1 m away from the nest entrance t2,4001.63, P00.112). The distribution of corpses and inert (Fig. 2b). This shows that each transporting ant tends to keep items between the right or left side of the area differed on carrying the corpse close to the nest entrance and then significantly from a binomial distribution with a probability becomes more likely to drop it as they move away from the of choosing the left side set at 0.5 (Kolmogorov–Smirnov nest. From a prophylaxis perspective, a slight increase in test, corpses: D00.727, N022, P<0.0001; inert items: D0 corpse’s distance may deeply influence its probability to be 0.556, N018, P00.0077). Indeed, carrying ants showed a contacted by foraging ants since ant density is expected to slight preference for the left area with, on average, 61.5 % of decrease with the square of their distance to the nest. Therefore, cadavers and 57.4 % of inert items being dropped on the left we investigated how distance from the nest actually influences side (see Fig. 3a). We could not relate this preference to the density of exploring ants. First, we observed that the some external cues that would specifically orient transport- distance to a nest entrance significantly influenced ant’sdensity ing ants. Most probably, this slight deviation of deposits (Fig. 2c) (Kruskal–Wallis test, χ2028.9, df05, P<0.001). In from an equal distribution on both sides originates from areas close to the nest entrance (0–50 cm)—i.e., for dis- memory effects (Diez et al. 2011): some ants may have tances below the threshold distance of corpse deposit contributed to several items’ transports and thus may have (52.62 cm)—the ant’s density was significantly higher than biased the choice of a direction by returning several times at in all areas further than 50 cm (Behrens–Fisher test, P<0.05). a previous location of items deposit. Moreover, at distances longer than 1 m from the nest— The distribution of inert items and corpses on the marked where the probability for an ant to drop its corpse reaches its and unmarked sides was not colony-dependent (GLM, maximum—the ant density was the lowest, with a median t2,4000.314, P00.755). Since transporting ants showed a value below 3 ants/m2. slight preference for the left side (60 % of observed ants), The deposit of corpses was not triggered by specific the marked area was equally located either on the left side abiotic conditions. Indeed, corpse-carrying ants did not dump their load in areas that were warmer or more en- a lightened. On average, the temperature at sites of corpse deposits (20.1±3.4 °C) did not differ from that measured at the most distant sites explored by foragers (19.9±4.3 °C)

(ANOVA, F1,11500.023, P00.880). Likewise, light exposure at places where dead nestmates were removed (10,490± 6,194 lx) did not differ significantly from those foraged by exploring ants (11,080±6,840 lx) (ANOVA, F1,11500.016, P00.897). Number of experiments Laboratory experiments 0 2 4 6 8 10 12 0−20 21−40 41−60 61−80 81−100 There was no colony effect on the distance of deposit of % of depots on the left area 2 corpses (Kruskal–Wallis test, χ 02.99, df02, P00.224). b For inert items, there was a colony effect with one nest dropping these items at longer distances (Kruskal–Wallis test, χ2040.3, df02, P<0.001). Nevertheless, on average, corpses were dumped at a distance of 31.6±14.9 cm (N0 234) that was significantly more remote from the nest en- trance than for rejected inert items (19.1±15.2 cm, N0155) (Wilcoxon rank sum test, W09,914, P<0.0001). It is well known that passive area marking left by explor- Number of experiments ing ants provides an averaged estimate of the occupancy level of an area by nestmates. We tested whether ants that 0 2 4 6 8 10 12 0−20 21−40 41−60 61−80 81−100 perceive such a chemical area marking used it as a cue for % of depots on the marked side of the area dropping corpses at locations that were weakly explored by nestmates. Before the introduction of items, the distribution Fig. 3 a Distribution of trials depending on the percentage of trans- of exploring ants on each of the two areas did not differ from porting ants that deposited their load on the left area. b Distribution of trials depending of the percentage of transporting ants that deposited a binomial distribution (Kolmogorov–Smirnov test, D00.2, their load on the marked area. Black bar corpse deposits (N022), gray N0120, P01). The distribution of inert items and corpses on bar inert items deposits (N018) 840 Naturwissenschaften (2012) 99:833–842

(50 % of the experiments) or on the right side (50 % of the around nest entrance (Wehner 1970). The lack of polariza- experiments) in order to avoid a systematic bias. For the tion of waste items around the nest could result from the experiments in which the marked area was located on the activity of many individual ants heading to different direc- left side, we compared the choice of carrying ants for the tions, memorizing the orientation of previous trips, in the marked or unmarked area to a theoretical binomial distribu- case of soil material (Robinson et al. 2008) or corpses (Diez tion where the probability for an ant of choosing the marked et al. 2011). In addition, we hypothesized that transporting area was set at 0.6. Likewise, for the other half of the ants might use density-related cues such as area marking for experiments in which the marked area was located on the identifying less foraged areas. Unlike some ant species that right side, we compared the choice of carrying ants to a tune their laying of a recruitment trail (Devigne and Detrain binomial distribution with a probability of choosing this 2006) or their walking behavior depending on their percep- right side set at 0.4. The distribution of corpses and inert tion of density cues (Detrain and Deneubourg 2009), we items between the marked or unmarked area (see Fig. 3b) showed that M. rubra corpse-carrying workers do not use did not differ significantly from a binomial distribution area marking as a criterion for depositing dead nestmates. (Kolmogorov–Smirnov test, corpses: D00.409, N022, P0 M. rubra transporting ants dropped corpses at highly vari- 0.0504; inert items: D00.389, N018, P00.131). On aver- able distances, between 7 and 289 cm from the nest en- age, 45.6 % of corpses and 47.4 % of inert items were trance. The same order of magnitude was observed for S. dropped on the marked area. Hence, whatever the nature invicta workers that drop corpses at distances ranging from of their load, carrying ants were not influenced by the 7 to 315 cm (Howard and Tschinkel 1976). The distance of presence of area marking: they did not tune their dropping corpse deposits directly depends on how the probability of behavior according to density-related cues such as the level dropping a corpse varies along the distance from the nest. In of area marking. the case of M. rubra, the probability of dropping the corpse follows a threshold function and increases nonlinearly with the distance to the nest. As a result, a corpse-carrying ant Discussion will be more likely to let down its load as it is going further away from the nest entrance. For inert waste items such as Waste management and corpse removal are important issues digging pellets, the probability curve of deposits by Messor for social insects since group living increases the risks of barbarus workers is quite different from the one of corpse pathogens’ transmission (Schmid-Hempel 1998). Here, we removal since it increases till a maximum at 5 cm from the describe how corpses are spatially organized when rejected nest entrance before decreasing for longer distances. As a in the surroundings of M. rubra nests. Corpse-carrying ants result, this leads to the formation of soil craters around the did not drop their load in “cemetery-dedicated” areas but nest entrance whose tops are situated where there is a scattered them around the nest. Although they dropped their maximal probability of dropping pellets (Theraulaz et al. load in remote areas, corpse-carrying ants did not specifi- 2002). Because of the long distances of removals and the cally choose zones that were less frequented by nestmates. low number of dead ants that are concurrently removed, Several behavioral traits indicate that ants tend to quickly necrophoresis is very unlikely to generate aggregation of drop corpses far away from the nest entrance. First, corpse- corpses in the field. Therefore, we never observed any carrying ants follow a fairly straight centrifugal path while “cemeteries” in the field as the ones that can be formed in carrying a corpse in remote areas. In Solenopsis invicta, the laboratory (Theraulaz et al. 2003). In open areas, the corpse-carrying workers also perform long-distance travels conditions required for the amplification process—and in a straight line, using a sun compass (Howard and hence the formation of piles—to occur can only be met if Tschinkel 1976). Such straight trajectories are the best ways there is a sufficient quantity of wastes to be transported and for corpse-carrying ants to maximize the distance of corpse deposited. Indeed, waste piles have been observed in removal from the nest while minimizing the contact time many ant species, but always when items were very with a dead nestmate. In field experiments, corpses were abundant like fungal wastes (Hart and Ratnieks 2001), scattered around the nest, and we observed a high variability seed husks (Hölldobler and Wilson 1990), or pine in both directions and distances of corpse deposits. Indeed, inflorescences (Czechowski et al. 2008). M. rubra workers carrying corpses had an equal probability The nature of waste items is a key feature that determines of choosing any direction after exiting the nest entrance. how transporting ants will behave and will deposit their load Such a lack of polarization in the pathways followed by in the nest surroundings. Differences at the individual level transporting ants was previously reported in S. invicta mov- lead to quite different spatial patterns of waste items that are ing over flat areas in the nest surroundings (Howard and related to specific functional value: an elevated structure of Tschinkel 1976). Likewise, for nest building, transports of soil pellets put in the vicinity of nest entrances provides excavated material lead to a uniform distribution of deposits colony protection against predators and thermoregulation Naturwissenschaften (2012) 99:833–842 841

(Sakaluk et al. 1989), while an increasing trend to deposit the colony. Indeed, corpse-carrying ants walked straight corpses at remote distances favors their isolation from the trajectories to the deposit locations, reducing the time core of ants’ activity. By dropping corpses twice further than allowed for the task. By not venturing into areas that are inert items in laboratory conditions, M. rubra workers de- less frequented or even unknown to their colony, they also crease the probability of hazardous encounters between diminish the risk of getting lost or predated. dead ants and nestmates since the density of foraging ants decreases drastically with distance to the nest. Besides the Acknowledgments We thank Xavier Massart, Ariane Catala, and sanitary advantage of choosing remote areas for corpse Hélène le Borgne for their contributions to the experiments. We also deposits, rejected corpses might also act as territorial thank Dr. Vincent Fourcassié for his helpful advice for data analysis. markers. In Pogonomyrmex badius, the presence of a mid- We also thank the anonymous referees for their comments. This study den decreases the number of other competing species in the was funded by a Ph.D. grant from FRIA (Fonds pour la Recherche dans l’Industrie et dans l’Agriculture). C.D. and J.-L.D. are senior surroundings of the nest (Gordon 1984). Wood ants use research associates from the Belgian National Fund for Scientific dead corpses in conflict situations: Formica cinerea workers Research (FNRS). put dead nestmates near the nest entrance as a defensive “guard,” while Formica rufa ants carry wastes and corpses Ethical standards We declare that the experiments comply with the current laws of Belgium. to places of confrontation with other species (Czechowski et al. 2009). Dropping potentially hazardous corpses in remote areas has obvious advantages, but this behavior could also have References drawbacks for corpse-carrying individuals as well as for the whole colony. By analogy with foraging costs (Hughes and Anderson RM, May R (1979) Population biology of infectious dis- Ward 1993), several costs related to corpse transport in- eases: part 1. Nature 280:361–367 crease with distance from the nest such as higher energetic Bankova VS, de Castro SL, Marcucci MC (2000) Propolis: recent advances in chemistry and plant origin. Apidologie 31:14 metabolic costs (Franks et al. 2004) for walking and trans- Batschelet E (1981) Circular statistics in biology. Academic, London porting items, a longer time allowed to the task, a higher Bos N, Lefèvre T, Jensen AB, D’ettorre P (2012) Sick ants become exposure to predation, or a higher risk of becoming lost due unsociable. J Evol Biol 25:342–351 to a poor knowledge of remote areas. The contact between Bot ANM, Curries CR, Hart AG, Boomsma JJ (2001) Waste manage- ment in leaf-cutting ants. Ethol Ecol Evol 13:225–237 the transporting ant and the potentially hazardous item is Chapuisat M, Oppliger A, Magliano P, Christe P (2007) Wood ants use also longer-lasting. Finally, it is important to consider the resin to protect themselves against pathogens. Proc Biol Sci “missed opportunity” cost, i.e., the cost of not depositing the 274:2013–2017 corpse elsewhere or of not engaging in alternative activities Choe D-H, Millar JG, Rust MK (2009) Chemical signals associated with life inhibit necrophoresis in Argentine ants. P Natl Acad Sci like foraging. Here, ants dropped corpses in the same areas USA 106:8251–8255 and at about the same maximal distances, but further than Cotter SC, Kilner RM (2010) Personal immunity versus social immu- the mean distance explored by nestmates. This suggests that nity. Behav Ecol 21:663–668 the benefits of dropping corpses in locations that are never Cremer S, Armitage SAO, Schmid-Hempel P (2007) . Curr Biol 17:693–702 explored by congeners are not worth the associated costs or Crosland M, Lok C, Wong T, Shakarad M, Traniello J (1997) Division risks. Moreover, since corpse-carrying ants do not use a of labour in a lower : the majority of tasks are performed chemical trail while transporting corpses, they may face by older workers. Anim Behav 54:999–1012 disorientation problems when reaching unknown areas Czechowski W, Marko B, Radchenko A (2008) Rubbish dumps reveal the diet of the ant colonies: Myrmica schencki Em. and Myrmica (Diez et al. 2011), in particular at remote distances from rubra (L.) (Hymenoptera: Formicidae) as a facultative pollen- the nest (Merkle et al. 2006). eaters. Pol J Ecol 56:737–741 In conclusion, corpse-carrying ants are dealing with an Czechowski W, Marko B, Godzinska E (2009) Corpse carrying in ants important issue, that is, the management of “sanitary risks.” (Hymenoptera: Formicidae): behavioural side effect of agressive arousal of competitve signalling? Pol J Ecol 57:341–352 This is achieved, namely, by dropping their load at locations Detrain C, Deneubourg J-L (2009) Social cues and adaptive foraging that can be safely reached and that minimize encounters strategies in ants. In: Jarau S, Hrncir M (eds.) Food Exploitation between nestmates and corpses. The “strategy” followed By Social Insects. CRC Press, Boca Raton, pp. 29–51 by workers to remove waste items differs greatly depending Devigne C, Detrain C (2006) How does food distance influence for- aging in the ant Lasius niger:theimportanceofhome-range on the expected sanitary risk associated to this item. Inert marking. Insect Soc 53:46–55 items are dropped near the nest entrance, whereas corpses Devigne C, Renon AJ, Detrain C (2004) Out of sight but not out of are transported at greater distances where the encounter, and mind: modulation of recruitment according to home range mark- hence, the transmission of pathogens between corpses and ing in ants. Anim Behav 67:1023–1029 Diez L, Deneubourg J-L, Hoebeke L, Detrain C (2011) Orientation in exploring nestmates becomes very unlikely. In addition, ants corpse-carrying ants: memory or chemical cues? Anim Behav are able to deal with sanitary risks at a relatively low cost for 81:1171–1176 842 Naturwissenschaften (2012) 99:833–842

Drees BM, Miller RW, Vinson BS, Georgis R (1992) Susceptibility and Moser JC (1963) Contents and structure of Atta texana nest in summer. behavioral response of (Hymenoptera: For- Ann Entomol Soc Am 56:286–291 micidae) to selected entomogenous nematodes (Rhabditida: Stei- Renucci M, Tirard A, Provost E (2010) Complex undertaking behavior nernematidae Heterorhabditidae). J Econ Entomol 85:365–370 in Temnothorax lichtensteini ant colonies: from corpse-burying Elmes GW (1973) Observations on the density of queens in natural behavior to necrophoric behavior. Insect Soc 58:9–16 colonies of Myrmica rubra L. (Hymenoptera: Formicidae). J Robinson E, Green K, Jenner E, Holcombe M, Ratnieks F (2008) Anim Ecol 42:761–771 Decay rates of attractive and repellent pheromones in an ant Evans JD, Aronstein K, Chen YP, Hetru C, Imler J-L, Jiang H, Kanost foraging trail network. Insect Soc 55:246–251 M, Thompson GJ, Zou Z, Hultmark D (2006) Immune pathways Rosengaus RB, Maxmen AB, Coates LE, Traniello JFA (1998) Disease and defence mechanisms in honey Apis mellifera. Insect Mol resistance: a benefit of sociality in the dampwood termite Zooter- Biol 15:645–656 mopsis angusticollis (Isoptera: Termopsidae). Behav Ecol Socio- Feldhaar H, Gross R (2008) Immune reactions of insects on bacterial biol 44:125–134 pathogens and mutualists. Microbes Infect 10:1082–1088 Ruiz-González MX, Moret Y, Brown MJF (2009) Rapid induction of Franks NR, Britton NF, Sendova-Franks AB, Denny AJ, Soans EL, immune density-dependent prophylaxis in adult social insects. Brown AP, Cole RE, Havardi RJ, Griffiths CJ, Ellis SR (2004) Biol Lett 5:781–783 Centrifugal waste disposal and the optimization of ant nest cra- Sadd BM, Schmid-Hempel P (2006) Insect immunity shows specificity ters. Anim Behav 67:965–973 in protection upon secondary pathogen exposure. Curr Biol Gordon DM (1984) The harvester ant (Pogonomyrmex badius) mid- 16:1206–1210 den: refuse or boundary? Ecol Entomol 9:403–412 Sakaluk SK, Mason A, Sakaluk MC (1989) Northeasterly orientation Gordon DM, Paul RE, Thorpe K (1993) What is the function of and defense of nest entrances in the harvester ant, Pogonomyrmex encounter patterns in ant colonies? Anim Behav 45:1083–1100 owyheei. J Insect Behav 2:261–266 Groden E, Drummond FA, Garnas JR, Franceour A (2005) Distribution Schmid-Hempel P (1998) Parasites in social insects. Princeton Univer- of an invasive ant, Myrmica rubra (Hymenoptera: Formicidae), in sity Press, Princeton maine. J Econ Entomol 98:1774–1784 Schmid-Hempel P (2005) Evolutionary ecology of insect immune Hart AG, Ratnieks FLW (2001) Task partitioning, division of labour defenses. Annu Rev Entomol 50:529–551 and nest compartmentalisation collectively isolate hazardous Siva-Jothy MT, Moret Y, Rolff J (2005) Insect immunity: an evolu- waste in the leafcutting ant Atta cephalotes. Behav Ecol Sociobiol tionary ecology perspective. In: Simpson SJ (ed.) Advances in 49:387–392 Insect Physiology. Academic Press, New York, pp. 1–48 Hart AG, Bot ANM, Brown M (2002) A colony-level response to disease Theraulaz G, Bonabeau E, Nicolis SC, Solé RV, Fourcassié V, Blanco S, control in a leaf-cutting ant. Naturwissenschaften 89:275–277 Fournier R, Joly J-L, Fernández P, Grimal A, Dalle P, Deneubourg Heinze J, Walter B (2010) Moribund ants leave their nests to die in J-L (2002) Spatial patterns in ant colonies. Proc Natl Acad Sci social isolation. Curr Biol 20:249–252 USA 99:9645–9649 Hölldobler B, Wilson EO (1990) The ants. Harvard University Press, Theraulaz G, Gautrais J, Camazine S, Deneubourg J-L (2003) The Cambridge formation of spatial patterns in social insects: from simple behav- Howard DF, Tschinkel WR (1976) Aspects of necrophoric behavior in the iours to complex structures. Philos T Roy Soc A 361:1263–1282 red imported fire ant, Solenopsis invicta. Behaviour 56:157–180 Ulyshen M, Shelton T (2012) Evidence of cue synergism in termite Hughes JJ, Ward D (1993) Predation risk and distance to cover affect corpse response behavior. Naturwissenschaften 99:89–93 foraging behaviour in Namib desert gerbils. Anim Behav 46:1243– Visscher PK (1983) The honey way of death: necrophoric behav- 1245 iour in Apis mellifera colonies. Anim Behav 31:1070–1076 Konrad M, Vyleta ML, Theis FJ, Stock M, Tragust S, Klatt M, Wehner R (1970) Études sur la construction des cratères au-dessus des Drescher V, Marr C, Ugelvig LV, Cremer S (2012) Social transfer nids de la fourmi Cataglyphis bicolor (Hymenoptera, Formici- of pathogenic fungus promotes active immunisation in ant colo- dae). Insect Soc 17:83–91 nies. PLoS Biol 10:e1001300 Wilson EO, Durlach N, Roth L (1958) Chemical releasers of necro- Lenoir A, Depickère S, Devers S, Christidès J-P, Detrain C (2009) phoric behaviour in ants. Psyche 65:108–114 Hydrocarbons in the ant Lasius niger: from the cuticle to the nest Wilson-Rich N, Spivak M, Fefferman NH, Starks PT (2009) Genetic, and home range marking. J Chem Ecol 35:913–921 individual, and group facilitation of disease resistance in insect Merkle T, Knaden M, Wehner R (2006) Uncertainty about nest position societies. Annu Rev Entomol 54:405–423 influences systematic search strategies in desert ants. J Exp Biol Zeh JA, Zen AD, Zeh DW (1999) Dump material as an effective small-scale 209:3545–3549 deterrent to herbivory by Atta cephalotes.Biotropica31:368–371