Insect societies, which are at a high risk of disease outbreaks, have evolved sanitary strategies that contribute to their social immunity. Here, we investigated in the red ant Myrmica rubra, how the discarding of nestmate cadavers is socially organized depending on the associated pathogenicity. We examined whether necrophoresis is carried out by a specific functional group of workers or by any nestmates that may become short-term specialists. By observing the behavioural profiles of tagged individuals, we assigned half of the colony members to functional groups (foragers, intermittent-foragers, domestics, nurses and inactives). Following the introduction of uninfected or sporulating corpses into the nest, intermittent-foragers were the functional group most involved in necrophoresis, as they touched, moved, and discarded more cadavers. Interestingly, sporulating corpses induced a more generalized response in workers from all functional groups, thereby accelerating their rejection from the nest. The individuals contacting corpses were also prophylactically engaged in more grooming behaviour, suggesting the existence of hygienist workers within ant colonies. These findings raise questions about a trade-off existing between concentrating health risks on a few workers who are highly specialized in necrophoresis and exposing a larger population of nestmates who cooperate to speed up nest sanitisation.

2.1   Biological models

The ant species, Myrmica rubra is commonly used as a biological model in studies about social immunity [16,17, 40]. Myrmica rubra is a common ant species in temperate areas of Europe and is considered an invasive ant in North America. This polygynous species usually lives in semi-open forests and grasslands [41], with nest populations ranging from 1 to 20 queens and 100 to 2500 workers in Europe [42]. Its nests are dug in various substrates, such as soil under stones, rotting wood, or roots of nettles and bramble bushes ([42], personal observations). Although ants share confined nest sites with genetically related nestmates, disease outbreaks leading to the death of the whole colony are rarely observed [43]. However, as an omnivorous species that frequently forages on prey, M. rubra workers are exposed to insect remains, which can be a source of entomopathogens. The ant genus Myrmica is actually exposed to a wide variety of parasites [44] and is commonly confronted in naturae to the generalist entomopathogenic fungus Beauveria bassiana [44,45]. Infection by this fungal pathogen starts with the attachment of conidia, i.e. asexual propagules, on their cuticle, and by the release of an enzymatic cocktail allowing a germ tube to penetrate the cuticle and enter the insect’s body [46]. The fungal mycelium then grows and invades the hemocoel, leading to the death of the host [47]. Ultimately, the fungus works its way outside the insect body to produce sporulating structures bearing conidia propagules [46]. At the last stage of infection, conidia are produced over the host cadaver, resulting in the white fluffy appearance of B.bassiana sporulating corpses, also called ‘white muscardine’ [46].

2.2   Collection and rearing of ant colonies

Five M. rubra colonies were collected during the summer of 2021 in woods located at Aiseau-Presles (Province of Hainaut: N 50°25.657’; E 04°35.674’) and Sambreville (Province of Namur: N 50°25.210’; E 04°37.878’). Nests were dug out from the litter and birch wood logs. Ant colonies were maintained in a room with constant temperature and humidity (21 ± 1C° and 50 ± 5%), and 12h :12h. Dark:Light regime and darkened test tubes were offered as nesting sites.

From these colonies, we made five experimental colonies consisting of two queens, around 225 workers (1/3 foragers and 2/3 internal workers), and 60 larvae at the second and third instars. Each experimental colony was derived from a different mother colony. Foragers were sampled from among ants walking outside the nests, and internal workers were removed from nesting test tubes. These ants were transferred to an experimental tray (30 cm × 22 cm × 7.5 cm), the bottom of which was covered by a 1 cm plaster layer, and the edges were coated with Fluon^TM to prevent ants from escaping. A hole was drilled in the tray, vertical to the experimental nest. A cotton wick was inserted into the hole and placed over a sponge soaked daily to evenly moisten the plaster below the nest.

The experimental nest was circular (3.45 cm radius) with a single chamber and a single entrance (5 × 5 × 2 mm). It was made of two square Plexiglas plates (8 cm on each side and 2 mm in thickness). The gap between these two plates created a 2 mm-high space between which ants settled in a monolayer, thus facilitating behavioural observations (Fig. 1). The nests were designed in vector drawings using the InkScape software and cut out using a laser cutter (HIGH-Z S-1000/T CNC ROUTER). The nest roof plate had a central circular hole (1.1 cm diameter) that could be alternately closed by one of the following two removable parts. The first removable part (red in Fig. 1.a) was a column passing through the roof down to the plaster ground (1.1 cm radius * 4.1 mm high). This created a space in the centre of the nest where workers could not settle and where corpses could be further dropped inside the nest with minimal disturbance to nestmates (see experimental procedure). The second removable part (blue in Fig. 1.d) was a transparent circular plug that served as a roof cover during the four hours of observations following the insertion of corpses. This plug allowed for the passage of ants underneath and provided them with access to the inserted items. Finally, the nest was surrounded by a plexiglass cage coated with Fluon^TM (polytetrafluoroethylene) to prevent the ants from climbing above the nest and disturbing behavioural observations. Experimental colonies were provided with water, a 0.3M sucrose solution and ten dead fruit flies (Drosophila melanogaster) were placed daily in the area at 9:30 am. Fruit flies were chosen as the protein source because they were small enough to be retrieved by ants through the nest entrance and large enough to observe prey consumption by ants inside the nest.

2.3   Experimental procedure

2.3.1        Planning

The experiment lasted for three weeks for each tested colony (Fig. 2). It began with two days of tagging, followed by two days of settlement and habituation of the ants to the experimental setup. It was organised into three distinct experimental sessions. During the first session, the behaviour of tagged ant individuals inside the nest was quantified for three days to further assign them to a functional group. The second session consisted of introducing uninfected corpses and, two days later, sporulating dead nestmates. This allowed us to quantify the dynamics of corpse removal. Particular attention was paid to the interaction of each tagged ant with corpses to assess whether their contribution to necrophoresis differed according to their functional group. During the third session, we monitored daily mortality and the causes of death of workers by placing their cadavers under sporulation conditions. This monitoring started on day 5, at the same time as session 1, and continued until day 21 (Fig. 2).

2.3.2        Tagging

Half of the worker population in each experimental colony was tagged individually. The tags were made of waterproof "toughprint" paper (0.9 mm² square) marked with four tiny 0.45 mm * 0.45 mm coloured squares. One square was black and the other three had a combination of different colours, selected to maximise contrast on video recordings (blue, green, red, white, and yellow, Fig. 1). This allowed us to identify the ant individual, regardless of its position, simply reading the colours clockwise, starting with the black square.

Before being tagged, the ants were briefly CO2-anaesthesized for 45 seconds. To restrict their movements, the ant head and thorax were delicately placed in slits made of a professional make-up sponge (Cala^TM). A droplet of glue droplet (Loctite^TM superglue) was applied to the first abdominal segment using a small guitar string (D'Addario^TM NYS007), and the coloured tag was placed using a toothpick (See SI). While being removed from the slit, the tagged ants were briefly anesthetized again for one minute and then isolated in Petri dishes until the glue dried completely. Four hours later, the tagged ants were placed in the experimental tray along with their conspecifics. As approximately 20% of the tagging attempts failed, we tagged 125 workers to ensure that we had approximately 100 tagged workers per colony at the beginning of the experiment (see SI for more details). The ants were randomly selected for tagging in a ratio of two-thirds internal to one-third external workers. We excluded callows or any individual showing signs of injury. In total, 491 workers were tagged successfully. As shown in previous studies, tagging had no significant effect on a worker's behaviour [48-50], except for a temporary increase in self-grooming behaviours that lasted no longer than 24 hours.

2.3.3        Session 1: Behavioural profile and assignment to a functional group

For each tagged individual, we quantified its location (outside VS inside the nest) as well as its behaviours inside the nest by observing it three times (10:30, 13:30, and 16:30) for 3 min over three successive days (Fig. 1). The recorded behaviours are listed in Table 1. On video recordings, we quantified the proportion of time each tagged ant was engaged in a given behaviour over the total duration during which this ant was observed inside the nest. The time that a tagged worker spent outside the nest was considered as foraging. Inside the nest, if a behaviour stopped for less than five seconds, it was considered uninterrupted. If an ant was motionless and did not engage in any activity for at least five seconds, it was recorded as inactive (based on the same criteria as in [51] study). Any instance of grooming behaviour performed by a worker alone, cleaning any part of its body such as antennae or abdominal tip, was classified as ‘self-grooming’. Similarly, when a worker actively groomed a nestmate, it was reported as ‘allo-grooming’. Once the data were collected, we compiled a behavioural profile for each tagged worker. These proportion of time spent on each task were used in a cluster analysis to assign each worker to one of the functional groups (see the Statistical Analysis section for further details).

2.3.4 Session 2: Necrophoresis

We compared the behaviour of tagged individuals exposed to either 10 uninfected or 10 sporulating corpses. The 10 uninfected cadavers were nestmates killed by freezing one day before their corpses were introduced into the nest (Fig. 1). Indeed, it takes 24 hours for the cuticular profile of dead workers to become enriched in oleic and linoleic acids, which allows their recognition as cadavers and triggers necrophoresis [52,53]. The ten sporulating cadavers were nestmates that died from infection by the entomopathogenic fungus B.bassiana (Naturalis-L^TM). These sporulating corpses were produced prior to the start of the experiment. To do so, we topically applied a droplet of Naturalis-L^TM insecticide (± 4 µL) to the abdomen of 30 nestmates that were isolated in a Petri dish with a wet cotton ball until their death. On the day of their death, ant cadavers were washed with ethanol and distilled water according to Lacey's method [54] to avoid contamination by microorganisms other than B.bassiana fungus. Following this treatment, the cadavers were placed on moist filter paper in a Petri dish sealed with parafilm and placed in a temperature-controlled cabinet at 25°C until the fungus sporulated over the ant body.

As for the course of the necrophoresis experiment, ten uninfected cadavers were introduced on day 8 in the centre of the experimental nest (Fig. 1). After delicately removing the column from the nest (Fig. 1, in red), in a fast and smooth movement, we dropped the cadavers through the hole and immediately closed them with a circular plug (Fig. 1, in blue). As the video recordings were launched before the introduction of corpses, we could observe the first workers who came into contact with corpses. Workers’ behaviour was filmed continuously for four hours (10:00 to 14:00) and thereafter for three minutes every half hour for another four hours. After the last cadaver was discarded outside the nest, the roof plug was gently replaced by the column piece in order to create an empty space for the second insertion of sporulating bodies. After two days of rest (Fig. 2), 10 sporulating cadavers were introduced on day 11 following the same procedure as described above. For both uninfected and sporulating corpses, we measured the time elapsed before they were removed from the nest, the identity of the workers that interacted with them, and the time, number, and type of interactions that each tagged worker had with the corpses. Any interaction lasting more than two seconds was recorded as contact, regardless of whether the ant touched the corpse with its antennae or mandibles. Interactions lasting less than two seconds were not considered. If the worker grabbed the corpse with its mandibles and moved it from its location, the interaction was considered as a displacement. The last displacement of a corpse that carried the cadaver outside the nest was considered as a discarding.

2.3.5 Session 3: Mortality and sporulation

From days 5 to 21, the bodies of the dead workers were removed from the tray each morning. For each cadaver, we identified the dead worker by its tag (if present), and we determined the cause of death by placing it in sporulation following the Lacey method, as described above [54]. The sporulation status of the corpses was monitored for two weeks by placing them in a dark, thermoregulated cabinet at 20°C.

2.4. Statistical analyses

All statistical analyses were conducted using RStudio software (RStudio Desktop v1.3.1056). The graphs and tables were created using Excel and R (ggplot2 and flexplot functions). The R code used to perform these analyses is provided in the supplementary materials for the reader's reference.

2.4.1 Behavioural profile and assignment to a functional group

Firstly, in order to determine the functional groups, behavioural data from all the tagged individuals from the five colonies were pooled to perform a Principal Component Analysis (stats package, princomp function). Based on the results of this analysis and data from the existing literature ([6], [55-60]), five different functional groups were identified: Foragers, Intermittent-foragers, Domestics, Nurses and Inactives (see the Results section).

Secondly, in order to assign each tagged individual to one of the five functional groups, we used the dataset from its own colony, as this is the relevant level of study when examining work organisation. As brood care was observed to be a relatively rare behaviour, workers that spent at least 5% of their time taking care of brood were defined as nurses All other tagged workers were assigned to one of the four other functional groups by means of a centroid cluster analysis performed on their colony dataset (factoextra package, kmeans function). The procedure resulted in a well-balanced proportion of each of the five functional groups in each of the five colonies (for further details, refer to the SI).

To test for possible changes in the behavioural profiles of tagged workers over the three days of observation, a Friedman test (friedmann.test function) was conducted, followed by post hoc analysis using pairwise Wilcoxon-Mann-Whitney tests (pairwise.wilcox.test function). We also conducted Kruskal-Wallis tests to compare behavioural profiles across colonies (kruskal.test function), followed by post-hoc Nemenyi tests (kwAllPairsNemenyiTest function). Further details are provided in the supplementary (SI Fig. 3).

2.4.2 Necrophoresis

The number of contacts with cadavers was compared across functional groups with a Kruskal-Wallis test followed, if significant, by pairwise comparisons using the Nemenyi Test. Owing to the many tied values, we used a correction procedure by randomizing the rank of those ties 50 times and keeping only the highest p-value obtained (rank function, ties.method = random). The proportion of tagged individuals contacting corpses or not was tested in each functional group using a Chi² test followed by pairwise Chi² independence tests with “False Discovery Rate” correction as post hoc tests (functions: chisq.test and pairwiseNominalIndependence, method = fdr). The same statistical analyses were applied to the proportion of tagged individuals in each functional group that displaced corpses (functions: chisq.test and pairwiseNominalIndependence, parameters; chisq = T, method = fdr). Since the same tagged workers were exposed to non-infected and sporulating corpses, the proportion of individuals contacting/displacing cadavers were compared between those two conditions using McNemar test (mcnemar.test function).

2.4.3 Mortality and Prophylactic grooming

The survival curves were compared across functional groups using Kaplan-Meier tests (survfit function). The same methodology was employed to control for the colonial effect on survival curves, with no significant differences observed between the five colonies (SI Fig. 5, Kaplan-Meier, n.s.). To assess whether different levels of exposure to the corpses had an impact on the mortality of workers, we performed Chi² test followed whenever needed by pairwise Chi² independence tests as post hoc (functions: chisq.test and pairwiseNominalIndependence, method = fdr).

Wilcoxon tests were also conducted to determine whether workers who contacted corpses exhibited higher levels of self-grooming or mobility, prior to the introduction of cadavers, than workers that did not contact them (wilcox.test function). Finally, we tested the potential correlation between the time a worker spends grooming itself and their mobility. To this end, we employed the Pearson correlation (ggscatter function, cor.method = "pearson").