Regulation of dietary intake of protein and lipid by nurse-age adult worker honeybees

ABSTRACT Essential macronutrients are critical to the fitness and survival of animals. Many studies have shown that animals regulate the amount of protein and carbohydrate they eat for optimal performance. Regulation of dietary fat is important but less often studied. Honeybees collect and consume floral pollen to obtain protein and fat but how they achieve the optimal balance of these two macronutrients is presently unknown. Here, using chemically defined diets composed of essential amino acids and lipids (lecithin), we show that adult worker honeybees actively regulate their intake of lipids around optimal values relative to the amount of protein in their diet. We found that broodless, nurse-age worker honeybees consume foods to achieve a ratio between 1:2 and 1:3 for essential amino acids to lipid or ∼1.25:1 protein to fat. Bees fed diets relatively high in fat gained abdominal fat and had enlarged hypopharyngeal glands. In most cases, eating diets high in fat did not result in increased mortality. Importantly, we also discovered that the total quantity of food the bees ate increased when they were given a choice of two diets relatively high in fat, implying that dietary fat influences bee nutritional state in a way that, in turn, influences behaviour. We speculate that dietary fat plays a critical role in maintaining workers in the nurse-like behavioural state independently of the influence of queen pheromone.


Introduction
Protein and fat are essential macronutrients required in most animals' diets. Protein provides essential amino acids whereas oils or fats provide the essential fatty acids, linoleic acid and α-linolenic acid. To satisfy their nutritional needs for growth, maintenance, and reproduction, animals have evolved mechanisms for the detection and regulation of the procurement of these nutrients. The regulation of macronutrient intake has been formalized in a modelling framework called the Geometric Framework (GF) for nutrition.
The GF is designed to measure the strategies animals employ when regulating two types of essential nutrients simultaneously (Simpson and Simpson, 1990;Raubenheimer and Simpson, 1993). It specifically employs the use of diets of specific ratios of macronutrients.
Animals are given one of two types of conditions: they are either confined to a diet of a defined ratio or they are given access to at least two different foods of specific ratios of two macronutrients. Using chemically-defined diets in this way makes it possible to study the trade-offs made when animals are confined to eat substandard diets. If an animal is given a choice of two diets (i.e. a 'paired diet' design), it is possible to identify how animals selfselect a presumed optimal quantity of each macronutrient through the measurement of the consumption of both diets.
Most research employing the GF has focused on how animals acquire carbohydrates and protein, in part because these nutrients are required in the largest amounts by herbivores. A few insightful studies employing the GF in predatory arthropods have revealed that dietary lipids are also actively regulated (Jensen et al., 2011;Jensen et al., 2012;Mayntz et al., 2005). For example, predatory ground beetles regulate their intake of protein relative to lipid in a way that suggests both nutrients are costly when overeaten (Jensen et al., 2012). Egg production in ground beetles is also optimized around a 2:1 protein to lipid ratio (Jensen et al., 2012). The extent to which protein and lipid intake is regulated, however, depends on the species in question. The orb web spider, Argiope keyserlingi, does not appear to actively regulate its intake of protein to lipid, though the composition of the food affects its own body composition (Hawley et al., 2014).
Most plant material has very little fat (<1% (Ivanova et al., 2012)). For this reason, herbivorous insects might be fat limited in diet, especially if they require the essential fatty acids, -linolenic and linoleic acid. Animals assimilate fat to their bodies through multiple mechanisms. Non-essential fatty acids can be created from excess carbohydrate (Arrese and Soulages, 2010), or provided by gut endosymbionts (Temara et al., 1991;Pranal et al., 1996;Douglas, 1998;Douglas, 2009). Bees and other pollinators acquire dietary protein, lipid, and carbohydrate (starch) from floral pollen (Wright et al., 2018). Floral pollen is similar to insect prey in its composition, as it is largely made up of protein and lipid (Roulston and Cane, 2000;Human and Nicolson, 2006). Foraging-age bumblebees (B. terrestris and B. impatiens) regulate their intake of protein and lipid to ratios between 14:1-12:1 when fed holidic diets (14:1 for B. impatiens and 12:1 for B. terrestris) (Vaudo et al., 2016). The regulation of fat, therefore, may be easier to study in animals adapted to a diet high in fat (Wilder et al., 2013).
To date, few studies have tested macronutrient regulation in eusocial insects and these have focused mainly on protein and carbohydrates. Eusocial workers in ants, honeybees, and bumblebees actively regulate their intake of protein (or EAA) because it is costly; over ingestion increases their risk of mortality for ants, honeybees and bumblebees (Altaye, et al., 2010;Pirk et al., 2010;Dussutour and Simpson, 2012;Paoli et al., 2014;Stabler et al., 2015). Additionally, the regulation of macronutrients by adult workers depends on their caste and age. For example, nurse-age adult worker honeybees regulate their intake of protein (P): carbohydrate (C) to a ratio of 1:50 when they are fed with diets composed of sugar and EAA; however, as they age, their requirement for dietary carbohydrate increases relative to EAA such that their optimum approaches a ratio of 1:250 P:C (Paoli et al., 2014). Likewise, foraging-age adult worker bumblebees (Bombus terrestris) regulate their intake of protein to carbohydrate to a ratio of 1:149 when fed with diets composed of protein (casein) and carbohydrate (sucrose). However, when fed with diets composed of free amino acids, they alter their intake towards a carbohydrate-biased ratio of 1:560 EAA:C (Stabler et al., 2015).
Eusocial bees collect and store pollen for use by the whole colony. The quantities of protein and fat in floral pollen vary greatly depending on floral origin. Protein ranges between 2-60% and lipid between 2-20% of pollen (Roulston and Cane, 2000). A study of bee-pollinated plant species' pollen reported P:L ranges of 13:1 to 1:5 P:L (Vaudo et al., 2020). In a honeybee colony, the nurse-age bees consume pollen (stored as bee bread, pollen mixed with honey) and create glandular secretions such as royal jelly to provision developing larvae with sufficient nutrition to reach the stage of pupation. A reasonable hypothesis would be that nurse-age honeybees, the main consumers of bee bread and pollen, regulate their intake of fats relative to protein in proportions that are similar to royal jelly they produce (i.e. ~2:1, see Wright et al., 2018), but this has never been tested. Here, we have performed a series of experiments using a chemically defined diet to identify the optimal protein and fat requirements of nurse-age adult worker honeybees. These experiments were designed to test how adult worker honeybees regulate their intake of essential amino acids relative to lipid (EAA:L). To identify the costs of consuming nonoptimal ratios of EAA:L, honeybees were confined to a diet made of a specific ratio of EAA:L.

Journal of Experimental Biology • Accepted manuscript
To identify the range of ratios of EAA:L that were optimal for nurse bees, honeybees had a choice of foods containing a specific ratio of EAA:L vs a food made of EAA without L or they were given a choice of two foods with different ratios of EAA:L.

Animals
Frames of sealed brood were collected from 6 -10 (depending on the experiment) honeybee (Apis mellifera carnica L.) colonies kept at Newcastle University and Northumberland, UK. Brood frames were suspended in ventilated boxes and kept in an incubator at 34°C and 60% relative humidity (RH). Newly emerged female workers were brushed from the frames daily and combined to form a mixed population in order to give a stronger representation of a population of bees across the treatment groups (Paoli et al., 2014). Cohorts of ~30 bees from the mixed population were then housed in ventilated Perspex feeding cages (11 × 6 × 20 cm). Feeding cages had 6 holes for feeding tubes, with 3 holes positioned at each end of the cage. Feeding tubes were modified 2 ml microcentrifuge tubes with 4 × 2 mm holes drilled along their length. The number of feeding tubes provided was experiment dependent and any blank holes were blocked with strong tape. Liquid diets were added to the feeding tubes, and the feeding cages with bees were placed in an incubator at 34°C and 60% RH. Bees were housed in feeding cages for 10 days and any dead bees were counted and removed daily.

Chemically-defined diets
Diets were designed so that the ratio between total EAA and L could be varied using liquid diets as in Vaudo et al. (2016). Three experiments were designed to test how honeybees regulate the consumption of essential amino acids and lipid. Diets were composed of the 10 essential amino acids (EAAs), sucrose and lipid (L) (see Vaudo et al., 2016). Amino acids were present at equimolar concentrations and each amino acid contributed 0.004 M to make a final 'total' amino acid concentration of 0.04 M (as in Paoli et al., 2014). Reagent grade L-amino acids were sourced from Alfa Aesar. Sucrose was provided at 1 M, making an EAA to carbohydrate ratio of 1:25 EAA:C M/M (Table S1); this ratio was chosen based on previous work showing the intake targets achieved by newly-emerged bees using the same diets (Paoli et al., 2014). The EAA:C ratio was constant in all diets, but the EAA:lipid (EAA:L) ratio depended on the treatment group. The concentration of EAA in the diet was 0.6% of the solution. The lipid source was soy lecithin (Optima Health & Nutrition, Bradford, UK) as in Vaudo et al., (2016). Consumption of diets was measured by weighing feeding tubes prior to placing them in the feeding cages and then again after 24 h. Tubes were weighed and replaced every 24 h for 10 days for all experiments. To control for potential changes in tube weight because of water loss from evaporation, control boxes without bees were run in parallel. Daily consumption of each diet was then adjusted for its specific evaporation value.
To relate the EAAs in diet to a specific amount of protein, we also conducted a series of control experiments. Casein is often used as the protein source in nutrition studies with insects and has been used with bees (Altaye et al., 2010;Pirk et al., 2010;Stabler et al., 2015). Casein does not have the EAAs in the same proportions of our equimolar diet. To be able to generalize our data to a 'real world' situation where bees were eating protein, we tested if the ratio of EAAs in our 'equimolar' diet influenced the quantity of food eaten by honeybees. To do this, we compared their food consumption to bees fed with a diet of EAAs with the proportions found in casein. Bees had access to two feeding tubes. One diet contained amino acids in 1M sucrose and the second contained 1M sucrose only. One cohort of bees was fed an equimolar EAA diet and a second cohort was fed the amino acid profile of casein (Table S2). Bees ate 3.9 times more AAs when feeding on casein AAs compared to an equimolar diet (Table S2). When combined with data from bumblebees, this suggests that the proportion of EAAs in diet can influence the total amount of EAA eaten.
We use these data to infer the quantity of a 'protein' that might be consumed when bees eat pollen (Stabler et al., 2015).

Experiment 1: Confined rails
The first experiment designed was a confined assay in which bees had access to one treatment solution containing EAAs, L and sucrose and a second solution containing only 1M sucrose. EAA and C were held constant in all treatment diets but the ratio of EAA:L was modified for 8 treatment diets by varying lipid content. The treatments were modified to include 1:0, 25:1, 10:1, 5:1, 1:1, 1:5, 1:10 and 1:12.5 EAA:L (Table S1). Each cohort of 30 bees had access to one of the specified EAA:L treatments and 1M sucrose (N=10 cohorts of 30 bees per treatment).

Experiment 2: Amino acid-paired rails
The next experiment was designed to identify the range of values for the EAA:L ratio that we identified as a putative intake target in Experiment 1. We achieved this by giving honeybees access to three diets: (1) one EAA:L treatment diet (25:1, 10:1, 5:1, 1:1, 1:5, 1:10 or 1:12.5); (2) a 1:25 EAA:C diet (no lipid); (3) and a 1 M sucrose diet. The concentration of the EAA and Journal of Experimental Biology • Accepted manuscript C components of diet were held constant, and L was varied as in Experiment 1. (N=10 cohorts of 30 bees per treatment).

Experiment 3: Lipid-paired rails
In this experiment, cohorts of honeybees were fed with two diets that differed in their EAA:L ratio. The logic of this experiment was to test whether bees regulated their intake of EAA:L to a specific ratio (Raubenheimer and Simpson, 1999). We used information from Experiments 1 and 2 to design the rails, as we predicted that the intake target for EAA:L was between 1:1 and 1:5 EAA:L. The nutrient space was limited such that all treatments were given one high lipid diet (1:5 EAA:L) and an additional diet of one of the following: 25:1, 10:1, 5:1 and 1:1. As in Experiment 1, all diets contained 1 M sucrose with an EAA:C ratio of 1:25 (M/M). A separate 1 M sucrose diet was also provided (N=10 cohorts of 30 bees per treatment).

Hypopharyngeal gland size
To assess how EAA:L diet affects hypopharyngeal gland development (HPG), the size of hypopharyngeal glands was measured from a subsample of bees from the confined rails experiment. To measure HPG size, the heads of 10 bees from each treatment were dissected. A needle was used to remove the dorsal plane of the head, and HPGs were removed with fine forceps. After being removed from the head, HPGs were added to 50 µl phosphate buffered saline (PBS). As an indicator of HPG size, the diameter of the shorter axis of 10 neighbouring acini per gland were measured under a microscope (Leica M125, Leica) and the average value for each bee was used as a unit of replication (Lass and Crailsheim, 1996;Babendreier et al., 2005). Images were taken of each acini and ImageJ Journal of Experimental Biology • Accepted manuscript software (version 1.51j8) (Schneider et al., 2012) was used to measure the diameter against a reference scale (Corby-Harris and Snyder, 2018).

Abdominal fat content
Ten bees from each cohort of the confined rails (Experiment 1) and amino acid-paired rails (Experiment 2) experiments were used to measure abdominal fat content. Bees were frozen at -80°C at the end of each experiment and were then dissected on ice. The abdomen was removed from the body and the gut was dissected out while the bee was frozen to prevent any fat in the gut contaminating the sample. A glass scintillation vial was pre-weighed, and the abdomen samples were added to it. Vials containing abdomens were placed into an oven at 60°C and samples were dried to constant mass. Once dry weight was recorded, 5 ml of petroleum ether was added to each vial and the lid was replaced. Samples were left for 5 d at RT and then the petroleum ether was removed from the samples which were then dried to constant mass. Fat content was recorded as the difference in pre and post petroleum ether extraction weight (Ellers, 1995;Strohm, 2000;O'Neill et al., 2011).

Statistical analyses
Analyses were carried out using IBM SPSS V.24 (IBM Corp. Released 2016. IBM SPSS Statistics for Macintosh, Version 24.0. Armonk, NY: IBM Corp). Average daily volume consumption was compared using 2-way analysis of variance (ANOVA) with treatment and solution as main effects and with Šidák's post hocs for multiple comparisons. The intake of each nutrient was calculated (mg) and presented as intake per bee. Consumption of nutrients (mg) was compared using multiple analysis of variance (MANOVA) with total EAA, lipid and carbohydrate intake as dependent variables. Intake targets were identified using Šidák's post hocs. To assess the differences in daily consumption of nutrients, repeated measures ANOVA was applied to daily consumption values of EAA, carbohydrate and lipid, with EAA:L treatment as the main effect. Survival data were analysed using Cox regression (Cox proportional hazards model). In the confined assay, the 1:0 (no lipid) diet was used as the indicator variable. Survival in the protein-paired and lipid-paired assays were also analysed with Cox regression using the most dilute lipid diet as the indicator variable. The effect of EAA:L treatment on abdominal fat content was compared using a one-way ANOVA.
HPG acini diameter was applied as the dependent variable in a one-way ANOVA with EAA:L treatment as the main effect.

Adult worker honeybees compromise their intake of EAA when diet is too lipid-rich
When bees were confined to a diet with a specific ratio of EAA:L, they regulated their intake of macronutrients in a way that suggested a 'rule of compromise' (Fig. 1A, Raubenheimer and Simpson, 1997). This rule of compromise was observed for only a subset of the diets (1:1-1:12.5) (Fig. 1B). For these diets, honeybees regulated their intake such that they ate no less than 0.3 mg/bee EAA (diets 1:12.5 and 1:10) and no less than 0.66 mg/bee of lipids (diets 1:1) ( Table 1, Fig. 1B). The point of change in plots of this kind has been used previously to identify the putative optimal ratio of macronutrients (Raubenheimer and Simpson, 1997). In our data, the predicted optimal value for EAA:L was ~1:2. Because they were confined to the diets, bees fed with this range of EAA:L actively adjusted their intake of EAAs; these bees consumed ~1.9× less amino acids than the other treatments (Table 1 Table 2, Šidák's post hoc, P > 0.05). The C:L ratio depended on the diet; the total amount of carbohydrates was the same, but lipid varied as a function of diet (Table 1, MANOVA Table   2, Fig. S1B).

Honeybees regulate their intake of EAA:L to a region between 1:2 and 1:3
To verify the predicted optimal intake target from the confined rails experiments, we designed an experiment in which bees could freely regulate their intake of carbohydrate and EAA but were restricted to a specific ratio of EAA:L. This allowed the bees to independently regulate their intake of EAA:C. Honeybees consumed the diets in a way that depended on the EAA:L ratio (Fig. 3A,B Fig. S3). They also regulated their intake of EAA:C in a relatively narrow range of values (1:82-1:104) (Table 1, Table 3, Fig. S3A).
As before when bees were confined to a specific EAA:L ratio, honeybees fed with diets in the range between 1:1 and 1:12.5 EAA:L exhibited active regulation of the EAA:L (Fig. 3A, Table 3). However, the EAA:L ratio achieved depended on the diet (Table 1, Table   3). Honeybees fed with the 1:5 and 1:10 EAA:L diets were able to regulate their intake so that they maintained a ratio of 1:2 of EAA:L (Table 3 (Table 1, Table 3, Fig. 3). The bees in the relatively high lipid diets regulated their intake of L:C within a range of 1:40-1:160 (Table 3, Fig. S3B).
When bees were provided with a diet with more EAAs than 1:1 EAA:L (5:1, 10:1 and 25:1), the EAA:L ratio achieved was very near to a predicted EAA:L if the bees ate from the tubes without active regulation of lipids (Table 3). These bees regulated their intake of EAA:C along a narrow range from 1:92-1:99, indicating active regulation of these two macronutrients ( Table 3). The ratio of L:C, however, varied over a much wider range from 1:786-1:1706 and was near to the ratio predicted by random feeding (Table 3).
The bees fed with the relatively higher lipid ratios exhibited a higher risk of mortality ( Fig. 4, Coxreg, 6 2 = 27.64, P < 0.001). In fact, bees feeding on 1:10 treatment had 1.85x greater risk of mortality over 10 days than bees on the most dilute lipid diet (25:1) (P = 0.019). Bees feeding on all other diets had similar risk of mortality (Fig. 4).

Honeybees fed with proportionally higher fat diets consume more food
The previous experiments indicated that bees regulated their intake towards a ratio of ~1:2 EAA:L. We performed one final experiment to test whether bees could indeed regulate to a predicted intake target when given a choice of two diets containing different ratios of EAA:L. Unexpectedly, the EAA:L treatment pair significantly influenced the total quantity of food consumed over the course of the experiment (Fig 6, Fig. 7, Table 1, MANOVA Table 2, two-way ANOVA, treatment×solution, F6,108 = 5.114, P < 0.001). Bees provided with the more concentrated EAA:L diets consumed more total food overall (5:1+1:5 and 1:1+1:5) than bees fed the other two diet pairs (Fig. 6, Fig. 7, Table 1 Table 3). None of the diet pairs significantly affected the rate of mortality over the 10 days of the experiment (Fig. S4, Coxreg, 3 2 = 2.33, P = 0.51).

Characterising relative protein from EAA consumption
When bumblebees feed on solutions containing EAA, they consume around 3 times less of the amino acid diet than they do from a diet containing protein (sodium caseinate) (Stabler et al., 2015). In order to estimate how the results found in our current study may be translated from EAA:L to P:L regulation, we carried out an experiment in which bees were provided with one of two treatment diets; one that contained equimolar essential amino acids, and a second with the amino acid profile of casein protein. In doing so we found that bees ate 3.9 times more amino acids on the casein profile treatment than on the equimolar treatment (Table S2), a similar finding to that of B. terrestris (Stabler et al., 2015).

Discussion
Broodless nurse-aged honeybees adjusted their intake of diets in our experiments to reach an EAA:L ratio between 1:2 and 1:3. Using an adjustment, we estimate that bees fed with protein rather than free amino acids regulate their food intake to an equivalent ratio of ~1.25:1 P:L (see Table S2). Thus, the experiments presented here show that nurse bees without brood bias their intake of EAA:L towards fats when compared to the ratio required to produce royal jelly. Our experiments also showed that the context of the dietary choices (i.e. high lipid vs low lipid pair) affected the total amount of food eaten. When given a choice of two diets with a relatively high lipid content, honeybees increased their total food intake by 25% compared to a choice among the lower lipid diets (Fig. 7). Bees avoided overeating dietary lipid: when they were confined to a high lipid diet (e.g. 1:12.5), they consumed the minimum food needed to meet their requirements for EAA in order to limit lipid intake. Unlike our previous work with bumblebees, we did not observe a strong cost to overeating lipid in these experiments. There was an increased risk of mortality in one treatment diet (experiment 2, 1:10 EAA:L) but dietary lipid was not associated with increased risk of mortality in the other treatments or in any of the treatments used in experiments 1 and 3.
Our original prediction was that bees would regulate their intake of lipid relative to protein in proportions that are similar to royal jelly they produce (i.e. ~2:1, see Wright et al., 2018). This ratio is quite similar to the measurements made of protein and lipid in larval honeybee tissue (2.4:1 P:L) (Ghosh et al., 2016). However, chemical analyses of adult bee tissues indicate that adult nurses have a P:L ratio skewed towards protein (7.4:1) (Ghosh et al., 2016). In our experiments, newly-emerged bees self-select diets that are ~1.25:1 (P:L) (we approximate this from our EAA:L diets, see Table S2), and thus, closer in ratio to what is required for royal jelly production. A previous study that used a mixture of honeybeecollected pollen reported that newly-emerged honeybees did not exhibit a preference for diets based on their P:L content (Corby-Harris et al., 2018). However, the range of diets provided to these bees excluded the values that we identified as optimal, and, therefore, were likely to have been too narrow to draw this conclusion. If the bees had fed randomly in the experiments of Corby-Harris et al. (2018), they would have achieved P:L intakes in a narrow range between 3:1 and 4.5:1. Our methods permitted us to explore a much larger 'nutritional space', making it possible to identify the range of values they actively self-select.
A foraging honeybee colony will collect pollen from a diverse range of flowering plant species at any given time, depending on the local environment. Pollen nutrient

Journal of Experimental Biology • Accepted manuscript
composition varies considerably among plant species (Roulston and Cane, 2000). The P:L ratio in floral pollen has large range; in some cases, lipid content surpasses protein (Roulston and Cane, 2000;Vaudo et al., 2020). In a honeybee colony, the foraging honeybees do not consume floral pollen, they only collect it; the nurse-age bees eat the pollen and make glandular secretions to feed larvae and the queen. A recent analysis of the protein and lipid of honeybee collected pollen identified that they collect pollen with an average P:L composition of 1.5:1 (Vaudo et al., 2020). This ratio, however, is quite different to the reported ratios of honeybee collected pollen from a study in Israel (Avni et al., 2014). When honeybees collect pollen, it is mixed with nectar and stored in cells in the colony. Stored pollen (bee bread) comes from several different floral origins. Bee bread, therefore, creates a nutrient profile for P:L that is less varied than that of individual plant species (10-30% protein and 3-8% lipid, Wright et al., 2018). Using these values, at its greatest extremes, the P:L ratios of bee bread vary from 10:1 to 1.25:1. Interestingly, these values fall across the range predicted by the studies of Vaudo et al. (2020) and Avni et al. (2014).
Bumblebees, by contrast, appear to require diets much higher in protein than in fat.
In a previous experiment, we measured the regulation of P:L of two species of bumblebee (Bombus impatiens and B. terrestris) (Vaudo et al., 2016). Foragers of these species regulate their intake of P:L to between 12:1 and 14:1 on their own, but to a ratio of 3:1 when they are foraging for a whole colony (Kraus et al., 2019). In addition, the P:L ratio of mixed plant species pollen collected by B. impatiens reportedly has a ratio near to 4:1 P:L (Vaudo et al., 2020). These ratios are considerably less concentrated in fat than we find for broodless, nurse aged honeybees, however, bumblebees do not produce royal jelly, as honeybees do and so their requirement for dietary lipid may be less. A closer examination of the P:L ratios found in the tissues of different castes in bumblebee colonies may reveal that the ratio of P:L required depends on bumblebee caste and sex.
Though we did not investigate it here, the reproductive and behavioural castes in a honeybee colony are likely to have different dietary needs for protein and lipid. It is also possible that our estimation of the preferred P:L ratio was biased more towards lipid because we worked with bees in a queenless, broodless setting in the lab. When bees are in a colony setting, they are exposed to many other in-hive factors that can influence their behaviour. Here, we have removed these stimuli in order to explore the nutrient balancing behaviour of the adult worker in a controlled setting (Paoli et al., 2014;Stabler et al., 2015;Vaudo et al., 2016). We expect that the EAA:L regulation reported here indicates the requirements of individual adult workers and that this may be different to when they are exposed to brood and queen pheromones. In our previous work, for example, we know that nurse-age honeybees regulate their intake of EAA:C towards diets that are considerably more concentrated in EAA than the ratio selected by foragers (Paoli et al., 2014). Thus, the differences in the P:L ratios identified as optimal for nurse-age honeybees in our study and in our previous work with bumblebees is likely due both to age and species differences in the relative demand for protein and lipid. We predict that when honeybees are consuming food to rear brood, they need proportionally more fat in their diet. This is because nurseage honeybees produce glandular secretions (e.g. royal jelly) that have an estimated P:L ratio of 2:1 (Howe et al., 1985;Wright et al. 2018).
The present experiments and our previous work indicate that honeybees have a greater cost to over or under ingesting protein than fat. The main purpose of eating protein for a sterile adult worker is to acquire the essential amino acids needed for somatic maintenance and growth and to produce glandular secretions to feed to brood and the queen (De Groot, 1953;Crailsheim, 1990). Imbalance of protein intake increases the risk of mortality for nurse and foraging-aged bees (Paoli et al., 2014). There are also other sublethal consequences including impaired royal jelly production, facilitated through reduced HPG development, varying degrees of ovarian activation (increased on higher quality proteins) and by the level of mobilised protein and amino acids found in the haemolymph, depending on the protein source (Altaye et al., 2010;Stabler et al., 2015).
Though we show here clear behavioural regulation of dietary lipid, the consequences of imbalanced P:L intake are less severe than the imbalanced intake of protein or carbohydrate. In our experiments, nurse-aged bees could consume diets as high as 1:12.5 EAA:L and still survive. The main consequence of diets high in fat for nurse honeybees is that they exhibit an increase in body fat.
Body fat acquired in diet could affect the progression of honeybees from nurse-age to forager. Previous work has shown that queen mandibular pheromone (QMP) plays an important role in controlling the behavioural caste trajectories of workers in the colony (Pankiw, 2004;Fischer and Grozinger, 2008). In the presence of the queen, nurse-age bees remain nurse-like (Jay, 1970(Jay, , 1972Jay and Jay, 1976;Kaatz et al., 1992). When they are removed from the influence of QMP, they begin to transform from nurse to forager (Pankiw et al., 1998). This transition to forager is accompanied by radical changes in their physiology that include the loss of body fat and reduced HPG size (Crailsheim, 1986;Crailsheim, 1992;Toth and Robinson, 2005;Fischer and Grozinger, 2008) and is found in other social hymenopteran insects (Blanchard et al., 2000;Toth and Robinson, 2005).
Though pheromones are important in this behavioural caste transition, bees that cannot acquire fat from diet precociously become foragers (Toth et al., 2005). These experiments indicate that body fat plays a role in the regulation of state in honeybees. Our data support this and show that fat in diet also influences how much food bees eat. When the diet choices we gave to bees were relatively high in fat, bees ate more food and also changed the ratio of P:L towards more L. Fat in diet is likely to stimulate the production of The proportion of essential fatty acids in pollen varies with pollen source (Manning, 2001) but also has profound effects on honeybee cognition which directly affects foraging behaviour (Arien et al., 2015;2018). Here, we have tested one dietary lipid source with a fixed proportion of the essential fatty acids (soy lecithin, 8:1 linoleic: α-linolenic acid); a ratio exceeding that known to cause impaired cognition with young honeybees. Though high dietary omega 6 fatty acids may induce physiological mechanisms to extend lifespan (O'Rourke et al., 2013;Chen et al., 2019) this has not been tested in honeybees. Due to the potential physiological impacts of essential fatty acid imbalance, it is possible that the ratio of the essential fatty acids also influences the regulation of EEA:L. Wright et al. (2018) predict that fat is the most variable component of the diet of adult worker honeybees, because it is also the most variable component of royal jelly. We have shown that broodless adult honeybees actively regulate their intake of protein and lipid. Foraging worker bees are likely to collect pollen with a narrow range of P:L ratios, but nurse bees still must consume food in the correct proportions to produce royal jelly. Future studies which incorporate nutrient balancing in whole colonies will elucidate how the P:L ratio in food influences the quality of royal jelly and other glandular secretions, and how this in turn might influence brood production and adult worker size and development.   Table 3. Predicted and observed endpoint ratios of nutrient consumption for lipid-and protein-paired rails. Note that confined rails are not included as bees could not adjust the ratio that they consumed.