Collection and rearing of Diacamma cf. indicum
We collected D. cf. indicum colonies from the main island of Okinawa during July and September 2020. A colony consists of the gamergate and 50–300 workers (39, 40). In this species, all emerged workers possess a pair of tiny appendages, called gemmae, on the thorax (38). The gamergate bites off the gemmae of newly emerged individuals, causing them to irreversibly differentiate into the worker caste.
We subjected eight colonies with gamergates to behavioural experiments (colony A: 76 workers, colony B: 195 workers, colony C: 175 workers, colony D: 134 workers, colony E: 199 workers, colony F: 181 workers, colony G: 129 workers, colony H: 130 workers). Each worker was marked with a unique colour pattern using enamel pens. Colonies were then maintained in plastic containers (240 mm × 180 mm × 85 mm) including an artificial nest (148 mm × 84 mm × 32 mm) until experiments began. Throughout the experiments, ants were fed standard artificial diets (1:1 ratio of protein to carbohydrate) (63) and fresh mealworms three times per week, given water daily, and kept in the laboratory at 25 ± 1℃ with a 14:10 light:dark photoperiod.
Behavioural observations
Colonies were transferred into experimental nests (148 mm × 84 mm × 32 mm or 221 mm × 141 mm × 37 mm) connected to a foraging area (240 mm × 180 mm × 85 mm) at least 1 day before beginning behavioural observations. Scan sampling was employed to categorize worker tasks, with digital cameras (DSC-WX350, SONY, Tokyo, Japan; iPhone XR, Apple, Cupertino, CA, USA) taking photographs of the artificial nest and foraging area 10 times per day, with intervals of at least 30 min between observations. Worker location was recorded based on these photographs. Workers were categorized into three types: physical contact with brood items (nursing), outside the nests (foraging), or neither. After recording states for all workers in each observation period, we classified individuals as nurse or forager according to our previous study (15): We defined workers performing only nursing at least once during the observation period as nurses (Fig. 1a), and those that performed only foraging at least once during the observation period as foragers (Fig. 1a). Workers that engaged in neither or both tasks were classified as an undefined worker and excluded from further analyses (Fig. 1a).
Reintroducing experiments
A previous model (33) suggests a series of manipulation experiments to test the change of threshold by task experience as follows: Nurses are removed from a colony, and then the reverted nurses are induced in the colony; the removed nurses prevented from performing the nurse task are re-introduced to the original colony; the researchers evaluate whether the reverted nurses continue the nurse task or not. We summarized the interpretations in results according to the prediction of the model (Fig. 1c). In addition to the above observation, we examined the task choice of all workers in the reintroduced colonies.
First, to define worker tasks in an unmanipulated (original) colony, we conducted 40 behavioural observations over four consecutive days (10 times per day). Based on the 40 observations, we categorised the workers into three types (Fig. 1a). Within 1 day after observations ended, the original colony was split into a nurse-biased (only nurses) and a forager-biased (only foragers) colony (Fig. 1b). Both sub-colonies contained the same number of eggs and larvae. Gamergates were added to the forager-biased colonies only. After the manipulation, we performed four days of scan observation daily (40 behavioural observations over four consecutive days) to determine worker tasks during DOL reconstruction (Fig. 1b). Based on the above observations, we categorised workers into three types using the 40-observation period, and finally obtained 391 foragers, 490 nurses, and 338 undefined workers (both tasks: 124 workers; non task: 214 workers). Workers that engaged in neither or both tasks were classified as an undefined worker and excluded from further analyses (Fig. 1a). Note that in this species, long-time isolation from the gamergate causes several physiological changes in workers, leading to behavioural changes (e.g., 49). Thus, we added gamergates to minimize the impact on the physiological changes of workers.
Next, to test the effect of experience on task choice, we moved some workers from the nurse-biased colony to the forager-biased colony (Fig. 1b). Workers selected for the move were those that remained nurses (hereafter, stable nurse) and those that became foragers (hereafter, precocious forager) or only the precocious foragers. From the forager-biased colony, we also removed the same number of nurses (hereafter, reverted nurse) and foragers (hereafter, stable forager) as the number of collected workers in the nurse-biased colony. This change allows all workers in the reintroduced colony to select any task. To set up this experiment, we transferred eggs (mean ± SD: 21.1 ± 4.1), larvae (mean ± SD: 4.8 ± 2.5), and the gamergate from the forager-biased colony to a new artificial nest. All selected workers from both sub-colonies were then released into the foraging area connected with this new nest. In this experiment, we prepared two types of treatments. First, using colonies A, B, C, and D, precocious foragers (mean ± SD: 5.3 ± 3.0) were reintroduced into the forager-biased colony (treatment1; Fig. 1b). In this treatment, the task-choices of the stable nurses in the reintroduced colonies are interpreted as the control, because the threshold of the stable nurse would be expected to stabilise. Next, to confirm the task choice of stable nurses, we reintroduced precocious foragers (mean ± SD: 4 ± 1.4) and stable nurses (mean ± SD: 4 ± 1.4) into the forager-biased colony using colonies E, F, G, and H, (treatment2; Fig. 1b). In both treatments, the task choice of stable foragers can also be regarded as control. We observed the reintroduced colonies on days 1–4, 11–12, and 18–19 (10 times per day). In these experiments, we categorised worker types daily based on 10 observations. In addition, to validate the effect of task experience on threshold change, we categorised all workers in the reintroduced colonies based on 80 observation periods. If task experience has no effect on the response threshold of workers, we predicted that task-shifted workers in the sub-colonies would return to the tasks they had in their original colony (Fig. 1c). If individual experience sufficiently reduces response thresholds, we predicted that task-shifted workers would continue their current tasks (Fig. 1c). Note that for one colony (colony G), no workers were removed because the number of workers in the forager-biased colony was too low, and another colony (colony A) was excluded from the reintroducing experiment because all reverted nurses died within four days in the reintroduced colony.
Statistical analysis
Effect of behavioural propensity on task choice
Our previous study showed that forager frequency in the original colony was correlated with the tendency to task shift (15). To confirm this result, we examined how proportion of time (to total observation time) spent in contact with the brood (nurse-task) or outside the nest (forager-task) in the original colony was related to worker tasks in forager-biased and nurse-biased colonies. We used the R package glmmTMB (64) to construct generalized linear mixed models (GLMMs) with a binomial error distribution. As the response variable (task), nurses and foragers in the sub-colonies were binarised to 0 and 1, respectively. The proportion of foragers or nurses across the total observation period (40 observations) in the original colony was the fixed effect, and colony ID was the random effect.
Effect of task experience on task choice
To further understand the effect of individual experience on task choice, we focused on reverted nurses after reintroducing workers from the nurse-biased colony to the forager-biased colony. We first used a GLMM with a binomial distribution to compare the proportion of reverted nurses that performed the nurse or forager task in treatment 1 with those in treatment 2. The number of elapsed days, treatment, and their interaction were set as fixed effects. We set the proportion of individuals performing focal tasks (forager or nurse) in each sub colony as a response variable. Colony ID was set as a random effect. Because none of the fixed effects significantly influenced either task in reverted nurses (see Results), we included all days and treatments in our GLMM to compare task selection (forager or nurse) among reverted nurses. We set task as a fixed effect, performing/not performing the focal task as the response variable, and colony ID as a random effect. This analysis included only reverted nurses that survived until day 19 after reintroduction (treatment 1, n = 35; treatment 2, n = 35).
In this study, P values of fixed effects were evaluated with the Anova function (Type II ANOVA) of the car package. Effect sizes of the models were determined with conditional R2 values (65) using the r.squaredGLMM function of the MuMIn package (66). All constructed models were checked for overdispersion using the testDispersion function of the DHARMa package (67). Statistical analyses were performed in R version 4.0.2 (68).