Resistance to nutritional stress in ants: when being fat is advantageous

Adequate nutrition is fundamental to the maintenance of good health and optimum performance. However, in nature, animals might not have reliable access to a diet that precisely matches their nutritional needs. The quality of food and its variability in space and time, such as consuming nutritionally imbalanced diets or suffering periods of starvation, can be an important source of nutritional stress. The ability of animals to resist these prolonged periods of nutritional constraints constitutes a phenotypic trait of great interest for understanding the ecology and evolution of all living organisms.

Extensive studies from slime moulds to humans have elucidated the nutritional regulatory strategies and mechanisms employed by organisms to cope with food that are nutritionally imbalanced (reviewed in Simpson and Raubenheimer, 2012). When confined to an imbalanced diet, animals are confronted with a situation of conflict among their requirements for different nutrients, and need to balance over-ingesting those nutrients present in relative excess against under-ingesting others (Simpson and Raubenheimer, 2012). Over-ingesting or under-ingesting a particular nutrient might come at a cost. For example, over-ingesting carbohydrate when confined to food low in protein relative to carbohydrate leads to obesity, whereas under-ingesting carbohydrate leads to energy shortage (Simpson and Raubenheimer, 2012). Similarly, over-ingesting protein when confined to foods low in carbohydrate relative to protein leads to poor survival, whereas under-ingesting protein leads to poor reproductive performance (Lee et al., 2008; Simpson and Raubenheimer, 2012). The expected compromise adopted by the animal is one that minimizes the cost and maximizes fitness benefits (Simpson and Raubenheimer, 2012).

Numerous studies have examined the strategies used by animals to resist prolonged periods of food deprivation (reviewed in McCue, 2010). How an organism maintains itself under food deprivation in order to survive is a key problem, both from biological and ecological points of view. The two most common strategies used by animals in response to food shortage are reductions in energy expenditure and the mobilization of energy reserves (Arrese and Soulages, 2010; McCue, 2010). The strategies adopted by animals differ within a species, depending on factors such as physiological state, age and reproductive status. In Drosophila, individuals resist starvation as a function of the extent of their fat stores, and fat stores, in turn, vary with reproductive status and age (e.g. Chippindale et al., 1996; Rush et al., 2007; Goenaga et al., 2012, 2013; Lee et al., 2013; Lee and Jang, 2014; Lee, 2015).

Social insects offer unique opportunities for studying resistance to nutritional challenges compared with other animals. First, they show a remarkable reproductive division of labour between long-lived fertile queens and short-lived sterile workers, making them ideal for studying the direct effects of nutritional challenges, independent of reproductive effort (Hölldobler and Wilson, 1990; Chapuisat and Keller, 2002; Dussutour and Simpson, 2012). Second, there is a further division of labour among workers depending on genetic components, morphology, age and physiological status (Hölldobler and Wilson, 1990). In many species, for example, older and lean workers are the ones in charge of foraging for the entire colony, whereas younger and corpulent individuals stay in the nest (Seeley, 1995; Schulz et al., 1998; Markiewicz and O'Donnell, 2001; Tschinkel, 1998; Blanchard et al., 2000; Toth and Robinson, 2005; Toth et al., 2005; Smith et al., 2011; Tibbetts et al., 2011; Robinson et al., 2012; Mersch et al., 2013; Bernadou et al., 2015). Nevertheless, division of labour among workers remains highly flexible in social insects. Foragers in experimentally manipulated colonies might revert to inner-nest tasks – a behavioural shift sometimes associated with an increase in lipid reserves (Robinson et al., 1992; Nakata, 1995; Amdam et al., 2005; Baker et al., 2012; Kuszewska and Woyciechowski, 2013; Bernadou et al., 2015). Lastly, social insects function as long-lived ‘super-organisms’ that are often confronted with food shortage or food imbalance due to annual seasonal fluctuations in food availability (Cook et al., 2011). This feature of the natural history of social insects offers the intriguing possibility of disentangling the effect of age, task and fat reserves on resistance to nutritional stresses.

In this paper, we looked at the capacity of foragers and inner-nest workers to resist starvation and macronutrient imbalance. The experiments were performed using Lasius niger (Linnaeus 1758), which is an ant species whose division of labour and nutrition requirements have been well described (Lenoir, 1979; Dussutour and Simpson, 2012). We used the geometric framework for nutrition (Simpson and Raubenheimer, 2012) and experimental manipulation of colony demography to show how lifespan of foragers and inner-nest workers was affected by the concentration and ratio of protein to carbohydrate in available foods. Knowing that development from inner-nest task to foraging is associated with profound changes in fat content and that insects resist starvation as a function of the extent of their fat content, we expect that inner-nest workers will resist nutritional stress better than foragers. We also predict that change in fat content will follow behavioural task reversion and thus modify tolerance to nutritional challenges, and by extension, change mortality rate.

We used the black garden ant Lasius niger, which is a common Palaearctic ant species. Colonies are monogynous and can contain up to 20,000 workers. The workers are monomorphic (3 mm long) and show age-based polyethism (Lenoir, 1979). Twenty four colonies (mother colonies), each comprising up to 20,000 workers, were collected in Marquefave (south-west France). Each colony was housed in a plastic container (200×200×70 mm) whose walls were coated with Fluon. Inside each container, ants nested in 24 test tubes partially filled with water behind a cotton plug. The nests were regularly moistened and kept at room temperature (25°C) under a 12 h light:12 h dark photoperiod.

In the field, black garden ants scavenge for dead insects and collect honeydew from sap-feeding Homoptera. Accordingly, these ants are confronted with foods varying widely in their ratio of protein to carbohydrates. For the experiments described below, we used synthetic foods varying in both the ratio of protein to carbohydrate (P:C) and in the total concentration of protein and carbohydrate combined (P+C). The protein content (P) consisted of a mixture of casein, whey protein and white egg powder, whereas glucose was used as a digestible carbohydrate source (C). The quantity of egg powder was kept constant in each food in order to keep the quantity of fat and minerals identical. Each food contained 0.5% vitamins. The foods were presented in a 1% agar gel (further preparation details are given in Dussutour and Simpson, 2008a, 2012). Unless otherwise mentioned, ants were fed daily with a balanced diet (P:C 1:5, P+C 200 g l⁻¹) that was shown to maximize lifespan in queenless colonies of L. niger (Dussutour and Simpson, 2012).

In all experiments described below, we called foragers (F) ants that were collected exclusively in the foraging arena of the colonies and inner-nest workers (I) ants that were collected inside the nest, deeply buried in the cotton layer.

The first experiment was designed to establish whether there was a difference in longevity, fat content and foraging behaviour between foragers and inner nest workers when both the concentration and the macronutrient composition of the diet varied.

After being established in the lab and fed ad libitum with a balanced diet for 30 days, 8 mother colonies were used to construct 64 daughter colonies of 200 workers without brood. Daughter colonies were composed exclusively of F1 foragers (N=32) or I1 inner-nest workers (N=32). Each daughter colony was housed in an experimental nest consisting of a plastic box (diameter, 100 mm; height, 60 mm), the bottom of which was covered by a layer of moistened cotton, connected to a foraging arena (diameter, 150 mm; height, 60 mm) with walls coated with Fluon. The experiment started 24 h after formation of daughter colonies.

The daughter colonies were confined to one of four treatments: (1) a high-carbohydrate diet (P:C, 1:5; P+C, 200 g l⁻¹), i.e. a balanced diet for L. niger (Dussutour and Simpson, 2012); (2) a high-protein diet (P:C, 5:1; P+C, 200 g l⁻¹); (3) a diluted high-carbohydrate diet (P:C, 1:5; P+C, 40 g l⁻¹); or (4) total food (but not water) deprivation. Eight daughter colonies of F1 foragers and 8 daughter colonies of I1 inner-nest workers were followed for each treatment. Ants had ad libitum access to food that was replenished daily. Ants never collected all the food offered before it was renewed.

To assess mortality, the number of dead ants within each daughter colony was counted every day and removed, until all ants died. To measure ‘foraging activity’ and ‘feeding activity’, the total number of ants in the foraging arena and the number of ants feeding were counted once per day (2 h after replenishing the food) for 30 days.

We also measured daughter colony intake for 30 days. The food was placed in the foraging arena in two small containers (diameter, 15 mm; height, 5 mm). The ants had access to only one container; the second was used as a control for measuring and correcting for evaporation. To evaluate the colony's intake, the containers with the food were weighed to within 0.01 mg every day before they were placed in the foraging arena and again after they were removed.

Last, to measure changes in fat content during the experiment, we first killed 24 F1 foragers and 24 I1 inner-nest workers in each mother colony (384 ants in total) just before constructing the daughter colonies, and assessed their fat content. During the experiment, ants in the daughter colonies divided their workforce again between foragers (F2) and inner-nest workers (I2). Hence, 30 days after starting the four treatments, when possible, we sacrificed 3 F2 foragers and 3 I2 inner-nest workers in each daughter colony (384 ants in total). A chloroform extraction protocol was used to extract whole-body fat and fat content was estimated by measuring change in mass following extraction (Cook et al., 2010).

In L. niger, 50% of the workers exhibit a behavioural age-related transition (Lenoir, 1979). Thus, the differences in survival between inner-nest workers and foragers could be explained by a difference in age. To check the age hypothesis, we re-ran the starvation treatment as follows.

In the first step of the experiment, after being established in the lab, 8 mother colonies were fed ad libitum with a balanced diet for 30 days. Then, at day 30, we killed 36 F1 foragers and 36 I1 inner nest workers in each mother colony and assessed their fat content (576 ants in total).

In the second step, we constructed 16 daughter colonies of 1000 workers without brood and comprising exclusively F1 foragers (N=8) or I1 inner-nest workers (N=8) from the 8 mother colonies. The daughter colonies were housed in experimental nests and fed a balanced diet ad libitum for 30 days. To assess mortality, the number of dead ants within each daughter colony was counted every day and removed from the colony. At day 30, we ensured that daughter colonies had divided their workforce again between foragers (F2) and inner-nest workers (I2) by counting the number of ants in the foraging arena 2 h after replenishing the food (mean proportion of ants in the foraging arena ±0.95 CI, 0.11±0.02 and 0.09±0.02 for daughter colonies of F1 foragers and I1 inner-nest workers, respectively). Then, we killed randomly 9 F2 foragers and 9 I2 inner nest workers per daughter colony (288 ants in total) and assessed their fat content.

In the last step, we collected 100 F2 foragers and 100 I2 inner-nest workers from the 16 daughter colonies and housed them in new experimental nests, giving 32 granddaughter colonies. Thus we obtained groups of F2 foragers that were formerly either I1 inner-nest workers (N=8) or F1 foragers (N=8), and groups of I2 inner-nest workers that were formerly either I1 inner-nest workers (N=8) or F1 foragers (N=8). These 32 granddaughter colonies were then starved until all ants died. To assess mortality, the number of dead ants within each granddaughter colony was counted every day and removed from the colony.

Following the results obtained in the previous experiments, we investigated whether task reversal is automatically accompanied by change in fat content (Fig. 1).

In the first step of the experiment, we used 8 mother colonies. After being established in the lab and fed ad libitum for 30 days, we randomly killed 48 F1 foragers and 48 I1 inner nest workers per mother colony and assessed their fat content (768 ants in total, first month).

In the second step, 8 daughter colonies of 1000 F1 foragers and 8 daughter colonies of 1000 I1 inner-nest workers without brood were constructed from these 8 mother colonies and housed in experimental nests and fed ad libitum for 30 days. To assess mortality, the number of dead ants within each daughter colony was counted every day. At day 30, we ensured that daughter colonies divided their workforce again between F2 foragers and I2 inner-nest workers by counting the number of ants 2 h after replenishing the food (mean proportion of ants in the foraging arena ±0.95 CI, 0.13±0.02 and 0.10±0.01 for daughter colonies of F1 foragers and I1 inner-nest workers, respectively). Then, we collected 6 F2 foragers and 6 I2 inner-nest workers in each daughter colony and measured their fat content (second month, first caste reversion; 192 workers in total).

In a third step, 16 granddaughter colonies of 100 F2 foragers and 16 granddaughter colonies of 100 I2 inner-nest workers were constructed from the 16 daughter colonies and fed ad libitum for 30 days. To assess mortality, the number of dead ants within each granddaughter colony was counted every day. At day 30, we ensured that colonies divided again between F3 foragers and I3 inner-nest workers by counting the number of ants in the foraging arena 2 h after replenishing the food (mean proportion of ants in the foraging arena ±0.95 CI, 0.14±0.03 and 0.10±0.02 for granddaughter colonies of F2 foragers and I2 inner-nest workers, respectively). Then, we collected 6 F3 foragers and 6 I3 inner-nest workers in each granddaughter colony and measured their fat content (third month, second caste reversion; 384 ants in total).

All statistical tests were conducted with SPSS (v21.0). In all experiments, longevity data were compared using Cox regression analysis, with treatment (diet fed to the ants in experiment 1 only), former worker caste, current worker caste (experiments 2 and 3 only) as factors and mother colony as a clustered term (nested factor). In experiment 1, we used generalized linear mixed-models to test for the effect of treatment, former worker caste, time and mother colonies on (1) the amount of food collected per individual, (2) the proportion of ants in the foraging arena and (3) the proportion of foragers feeding. Time was considered as a repeated factor and colony as a random factor. For all experiments, we used a general linear model to test for the effect of treatment (experiment 1), former worker caste and current worker caste on the proportion of fat, with colony as a random factor. In experiment 1, the relationship between carbohydrate intake and fat content across current caste and former caste was examined using multiple regression analysis.

For both F1 and I1 worker castes, we observed a decrease in ant longevity when the nutrient concentration was decreased (hazard ratio, HR=0.35, P<0.001), when the ratio of protein to carbohydrate was increased (HR=0.02, P<0.001) and as expected when the ants were starved (HR=0.02, P<0.001, Fig. 2).

Overall, F1 foragers lived less long than I1 inner-nest workers, except on the less-diluted high-carbohydrate food, where F1 foragers (median lifespan, 289 days; maximum lifespan, 409 days) lived as long as I1 inner-nest workers (median lifespan, 295 days; max. lifespan, 436 days; HR=1.18, P=0.126, Fig. 2A). When F1 foragers were fed the most diluted high-carbohydrate food (median lifespan, 205 days; max. lifespan, 300 days), a substantial decrease in survival probability occurred early in the experiment and lasted for 20 days. The survival probability thereafter decreased more slowly, following the time course observed for I1 inner-nest workers (median lifespan, 202 days; max. lifespan, 369 days; HR=1.78, P<0.001, Fig. 2B). I1 inner-nest workers fed the high-protein diet (median lifespan, 40 days; max. lifespan, 118 days) lived twice as long as F1 foragers (median lifespan, 21 days; max. lifespan, 72 days; HR=4.94, P<0.001, Fig. 2C). I1 inner-nest workers (median lifespan, 45 days; max. lifespan, 112 days) lived five times longer than F1 foragers (median lifespan, 10 days; max. lifespan, 59 days) when starved (HR=14.08, P<0.001, Fig. 2D). Interestingly, F1 foragers fed the high-protein diet lived twice as long as F1 starved foragers (HR=0.33, P<0.001, Fig. 2B,C), whereas I1 inner-nest workers fed the high-protein diet lived as long as I1 starved inner-nest workers (HR=1.16, P=0.263, Fig. 2C,D).

When the high-carbohydrate diet was diluted, all colonies had overall more ants in the foraging arena (Fig. 3) and a higher proportion of workers were engaged in feeding (Fig. 4A), thereby increasing food collection to the colony (Fig. 4B) and compensating for nutrient dilution (Tables S1-S3). Similarly, when colonies were exposed to the high-protein diet, more ants were observed in the foraging arena (Fig. 3), more workers were engaged in feeding (Fig. 4A) and food collection increased (Fig. 4B), presumably in an effort to maintain a constant carbohydrate intake, but at the cost of incidentally increasing protein supply to the nest (Tables S1-S3). Last, all colonies exhibited a high rate of ant mobilization in the foraging arena when starved (Table S1, Fig. 3).

I1 inner-nest workers showed a very low level of foraging activity, feeding activity and food collecting in comparison to F1 foragers for all treatments (Figs 3 and 4). On the less-diluted high-carbohydrate diet, the first I1 inner-nest workers feeding were observed after about 10 days from the start of the experiment (Fig. 4A). The I1 inner-nest workers reached the same level of foraging activity, feeding activity and food intake as the F1 foragers after 12 days on the less-diluted high-carbohydrate diet (Tables S1-S3, Figs 3 and 4); it took them 21 days when the high-carbohydrate diet was highly diluted (Tables S1-S3, Figs 3 and 4). When I1 inner-nest workers were fed the high-protein diet they never reached the same level of foraging activity, feeding activity and food intake as F1 foragers over the observation period (Tables S1-S3, Figs 3 and 4). Under starvation, I1 inner nest workers were almost never observed in the foraging arena (Table S1, Fig. 3).

Prior to the start of the experiment, we confirmed that F1 foragers were leaner than I1 inner-nest workers (F_1,119=265.44, P<0.001, Fig. 5). After being fed for 1 month with the less diluted high-carbohydrate diet, the most diluted carbohydrate diet or the high-protein diet, the pattern of ‘lean forager (F2)–corpulent inner-nest worker (I2)’ emerged among both F1 foragers and I1 inner-nest workers (Table S4, Fig. 5). The pattern ‘lean forager (F2)–corpulent inner-nest worker (I2)’ did not occur when ants were starved (Table S4, Fig. 5).

For both F1 and I1 worker castes, we observed a decrease in fat content when the nutrient concentration was decreased, when the ratio of protein to carbohydrate was increased and when the ants were starved (Table S5, Fig. 5).

The regression model of the fat content on the carbohydrate intake across current and former caste was significant (F_7,117=28.06, P<0.001, Fig. 6) and accounted for 64.1% of the variance. Fat content increased significantly with carbohydrate intake (Table S6). As a confirmation of the preceding analysis, the model indicates that fat content was significantly affected by former caste and current caste (Table S6): for the same carbohydrate intake, the fat content was higher for I1 inner-nest workers than for F1 foragers, and higher for I2 inner-nest workers than for F2 foragers. Moreover, there was a significant interaction effect between carbohydrate intake and current caste (Table S6). Fat content increased more rapidly for I2 inner-nest workers than for F2 foragers. Examination of the standardised regression coefficients of the multiple regression model shows, however, that fat content was almost entirely determined by carbohydrate intake (Table S6).

A second explanation for the difference in survival between F1 foragers and I1 inner-nest workers could be difference in chronological age due to age-related division of labour. Former F1 foragers might be older than I1 inner-nest workers.

Again, prior to the start of the experiment, we observed that F1 foragers were leaner than I1 inner-nest workers in the mother colonies (F_1,176=300.5, P<0.001, Fig. 7A).

After being separated for 1 month and fed ad libitum, daughter colonies of F1 foragers and daughter colonies of I1 inner-nest workers showed once again the pattern of ‘lean forager (F2)–corpulent inner-nest worker (I2)’ (current caste: F_1,181=159.60 P<0.001, Fig. 7A). Nevertheless, I1 inner-nest workers hold still a slightly higher fat content than F1 foragers (former caste: F_1,181=7.34 P=0.007, Fig. 7A).

When daughter colonies were fed ad libitum for a month, mortality rate was slightly higher for F1 foragers than I1 inner-nest workers (HR=2.03, P=0.044, Fig. S1A). When granddaughter colonies were starved, I2 inner-nest workers lived longer than the F2 foragers (current caste: HR=18.03, P<0.001, Fig. 7B). F2 foragers that were once I1 inner-nest workers (median lifespan, 11 days; max. lifespan, 52 days) lived as long as F2 foragers that were once F1 foragers (median lifespan, 10 days; max. lifespan, 44 days). I2 inner-nest workers that were once F1 foragers (median lifespan: 32 days; max. lifespan, 95 days) lived for a shorter time than I2 inner-nest workers that were once I1 inner-nest workers (median lifespan, 50 days; max. lifespan, 116 days; former caste: HR=1.50, P=0.002, Fig. 7B).

Fat content was correlated with worker caste (current caste: F_1,371=116.37 P<0.001) and followed caste reversion (time: F_1,371=4.67 P=0.010; time×current caste F_1,371=2.87 P=0.059, Fig. 8). Hence, ants demonstrated the capacity to revert from a low-storage (F1 forager) to a high-storage (I2 inner-nest worker) state and back again (F3 forager). Similarly, they were able to revert from high-storage (I1 inner-nest worker) to low-storage (F2 forager) and back to a high-storage state (I3 inner-nest worker).

Survival probability was higher for I1 inner-nest workers than for F1 foragers in daughter colonies (current caste: HR=1.76, P=0.003, Fig. S1B) and higher for I2 inner-nest workers than for F2 foragers in granddaughter colonies (current caste: HR=2.08, P<0.001, former caste: HR=1.11, P=0.835, Fig. S1C).

We found that worker mortality was highest on diets with high P:C ratios, and that both foraging and mortality rates were inversely correlated with worker fat reserves. This finding suggests that a worker's physiological condition might be tightly linked to its behaviour and performance. Inner-nest workers were more resistant to nutritional challenges than foragers. Our study also highlighted the extent of behavioural and physiological flexibility.

Three mechanisms can be proposed to explain why inner-nest workers were more resistant to both starvation and suboptimal nutritional conditions. First, this difference could be due to differences in energy reserves. In insects, starvation resistance is positively correlated with proportion of body fat (e.g. Ballard et al., 2008; Goenaga et al., 2013). Ants that possess greater fat reserves might also survive longer when facing adverse conditions. Measures of fat content in our study confirmed the pattern of ‘lean forager–corpulent inner-nest worker’, which is common in social insects (Seeley, 1995; Schulz et al., 1998; Markiewicz and O'Donnell, 2001; Tschinkel, 1998; Blanchard et al., 2000; Toth and Robinson, 2005; Toth et al., 2005; Smith et al., 2011; Tibbetts et al., 2011; Robinson et al., 2012; Mersch et al., 2013; Bernadou et al., 2015) and suggests that inner-nest workers survived longer due to their high fat reserves. When fed an imbalanced diet, foragers that remained foragers lost a considerable amount of fat and foragers that reverted to inner-nest tasks did not gain any fat. Thus, as a whole, after 1 month, colonies of foragers still had less reserve than colonies of inner-nest workers and survived less long. On the contrary, when fed a balanced diet, foragers that took on inner-nest tasks gained fat and reached the same level of reserve as ‘original’ inner-nest workers. Thus, after 1 month, the amount of reserve between colonies of foragers and colonies of inner-nest workers became identical. As a consequence, under optimal conditions, groups of foragers lived as long as groups of inner-nest workers.

Interestingly, the mortality pattern observed under starvation conditions in inner-nest workers, with two waves of increased mortality, is related to the bimodal distribution of abdomen size with ants storing either small or large amounts of reserve (Fig. S2). Our results show that food distribution among workers is heterogeneous, with a minority of foragers storing small amounts of reserves and a majority of inner-nest workers holding large amounts of reserves. These inter-individual differences in storage capacities/levels are sometimes extreme in ants that have a small fraction of specialized workers known as crop repletes that store large quantities of food (e.g. Burgett and Young, 1994; Børgesen, 2000). Our results suggest that a heterogeneous distribution of reserves might be more beneficial for the colony than an homogeneous one. Indeed, the few workers that hold the majority of the colony reserves will survive longer periods of adverse nutritional conditions and might then ensure colony success when nutritional conditions improve (Fig. S3).

A second mechanism to increase resistance to adverse nutritional conditions may be to reduce the rates at which reserves are utilized to save metabolic energy. For example, Drosophila has been reported to show lower locomotor activity under prolonged nutritionally adverse conditions (Rion and Kawecki, 2007; Slocumb et al., 2015). Inner-nest workers did indeed show a reduction in foraging activity during the first 3 weeks of the experiment, supporting this hypothesis in our study. Another strategy, which is often employed earlier in response to food deprivation or imbalance, would be to increase locomotor activity to maximize exploration and therefore chances to find food (McCue, 2010). This latter strategy seemed to be the one adopted by groups of foragers in the present study. Transitions between the two strategies (move more to enhance chances of finding better feeding conditions versus move less and conserve reserves) might depend on the level of lipid reserves and by inference on the extent of starvation resistance.

A third explanation for the difference in survival observed between foragers and inner-nest workers could be a difference in chronological age due to age-related division of labour shown by many social insect species, including L. niger, whereby each worker progresses through a set of behavioural changes over her life from inner-nest workers to forager (Lenoir, 1979; Hölldobler and Wilson, 1990). However, we showed that the inner-nest worker-to-forager transition can be reversed by manipulation of the demographic structure of a colony, inducing foragers to return to nest duties, as has been demonstrated in honeybees (Robinson, 1992; Huang and Robinson, 1996; Rueppell et al., 2007). However, unlike honeybees, our ants were able to regain their lipid stores when reverting to inner-nest activities. Thus, ants possessed a means to revert from a low-fat (forager) to a high-fat (inner-nest worker) storage state in response to diet quality. Although our results confirmed that the link between age and task is flexible in ants (e.g. Sendova-Franks and Franks, 1993; Blanchard et al., 2000; Dolezal et al., 2012; Bernadou et al., 2015), not unexpectedly there was a slight increase in mortality rate with chronological age, with foragers that reverted to inner-nest workers dying somewhat sooner than ‘true’ inner-nest workers. This difference could be due to the fact that fat reserves were lower in foragers that reverted to inner-nest workers than in inner-nest workers that remained inner-nest workers.

In groups of foragers, we observed a rapid task re-allocation between foragers and inner-nest workers after only a single day, whereas in groups of inner-nest workers, all workers remained in the nest for 2 weeks before allocating ants to foraging duties. Carbohydrate scarcity in either the high-protein diet or highly diluted diet has been observed to trigger increased foraging activities in ants in order to achieve a target intake of carbohydrate (Dussutour and Simpson, 2008b, 2009, 2012; Cook et al., 2010). However, this compensatory feeding behaviour does not occur immediately but is observed after a certain period of carbohydrate deprivation (Dussutour and Simpson, 2008b). In our experiment, foragers regulated carbohydrate intake after only a few days. They consumed most and recruited more congeners when confined to imbalanced (high P:C, 5:1) and diluted diets (1:5, 40 g l⁻¹) than nutritionally balanced diets (1:5, 200 g l⁻¹). A similar behaviour was also observed in inner-nest workers but later as the weeks progressed. We suggest that carbohydrate deprivation was detected later in colonies of inner-nest workers than colonies of foragers as a result of their larger energy reserves. One consequence of increasing foraging on the high P:C (5:1) diet to gain limiting carbohydrate and by extension to increase fat reserves, was incidentally increasing collection of protein, with associated fitness costs (Simpson and Raubenheimer, 2012). We confirmed once again that restriction to high-protein, low-carbohydrate diets decreased lifespan in ants (Dussutour and Simpson, 2009, 2012; Bazazi et al., 2016) as in other species of insects and mammals (Lee et al., 2008; Simpson and Raubenheimer, 2012; Le Couteur et al., 2016).

In our study, foragers were the first to die when faced with adverse nutritional conditions (Fig. S3). This might have drastic consequences on colony survival. As shown in our study, when foragers are removed from the colony, ants make the transition from inner-nest workers to become foragers. By doing so, inner-nest workers cease their nest-related activity. Under natural conditions, this behaviour might translate to reduced brood care, less defence or poorer hygiene. Some recent theoretical papers (Khoury et al., 2011, 2013; Russell et al., 2013) investigating colony collapse disorder (CCD), strongly suggest that high mortality rate in honeybee foragers increases the pressure on colony population, affecting its growth and, by extension, could be responsible for colony decline. Their model predicts that above a certain forager death rate, colony population would decrease dramatically and colony failure would become inevitable. As suggested in our previous study, impoverishment of natural resources and/or common use of artificial pollen enriched in protein might be one major factor behind CCD (Dussutour and Simpson, 2012). Here, we demonstrate clearly that imbalanced diets are deadly for foragers in comparison to inner-nest workers, increasing their death rate more than tenfold. In CCD-affected colonies, researchers reported reduced numbers of adult bees, queens and many frames of brood, suggesting rapid depopulation of adults. Knowing who dies first in CCD-affected honeybee colonies would allow a better understanding of the major threat that is CCD.

We thank two anonymous referees and the editor for their helpful comments on the manuscript. We also thank Abel Bernadou and Gerard Latil for technical assistance.