Fat body lipogenic capacity in honey bee workers is affected by age, social role and dietary protein Free

Metabolic homeostasis is essential for life. All organisms need to balance reactions that consume energy (anabolism) with those that produce energy (catabolism). When energy is abundant, cells generally store energy in the form of fatty acids (Renne and Hariri, 2021), which can be taken up from the extracellular environment or synthesized in de novo lipogenesis by fatty acid synthase (FAS), linked together into larger lipid molecules such as triacylglycerol (TAG) and sequestered in storage organelles called lipid droplets. In times of energetic stress, stored energy can be released from lipids via lipolysis by lipase enzymes. With the evolution of higher levels of biological organization above the cell, regulation of metabolic homeostasis has shifted to support the higher unit, such as the tissue, organ or multicellular organism organization (Torday, 2015). This regulatory shift has resulted in the evolution of metabolic division of labor, where anabolic and catabolic tasks can be decoupled (Negroni and LeBoeuf, 2023). For example, in metazoans, different organs with specialized functions coordinate lipogenic and lipolytic tasks to maintain organismal homeostasis (Zhao and Karpac, 2020).

Some tissues thus incur a metabolic cost to provide for the energetic needs of other tissues, which can only be explained by fitness benefits at a level of organization above tissues (the individual organism) (Negroni and LeBoeuf, 2023).

This trend is similarly seen in eusociality, a level of biological organization above the individual organism, where individuals live together in colonies with cooperative brood care, overlap of generations and reproductive division of labor (Szathmáry and Smith, 1995). A well-studied example is found in the division of labor between queens and female workers in the western honey bee, Apis mellifera. Queens, which are responsible for most or all of the reproductive output of the colony, develop in response to high nutrition (quality and amount) during the larval period (Patel et al., 2007), and display adult physiology consistent with high nutritional status, such as abundant lipid stores (Strachecka et al., 2021). The female workers, which are typically functionally sterile, develop in response to less nutrition (quality and amount) during the larval period and have lower adult lipid stores. Reproductive division of labor thus has as its basis the net transfer of energy from workers to queens, with individual workers incurring an energetic cost in order to increase not their own fitness but the collective fitness of the colony (Negroni and LeBoeuf, 2023). This system is argued to be a key driver of the evolutionary success of eusocial species (Bernadou et al., 2021).

Because they forgo direct reproduction, workers instead contribute to colony fitness through the performance of social tasks. Honey bees have evolved a system of temporal polyethism, in which behavior correlates with chronological age: in the first phase of life, workers perform a series of (inside) nest tasks, most notably nursing, in which bees produce proteinaceous jelly in specialized, highly developed hypopharyngeal glands (HGs) in the head to feed to developing bees and other members of the colony. Such tasks that are performed inside the protected nest carry low extrinsic mortality risk. Next, in response to colony needs and their own physiological status, workers transition from inside tasks to tasks that involve outside activities such as foraging. Foraging involves making frequent flights to collect food and water resources to nourish the colony (Robinson, 1992). Flight activity is associated with high extrinsic mortality risk (Prado et al., 2020).

The behavioral transition from nest tasks to foraging is accompanied by many physiological changes in workers. Mechanistically, one of these events occurs in the fat body, the major site of lipid synthesis and storage in insects. Nurses maintain high lipid stores in the fat body, but undergo dramatic loss of lipid stores prior to the transition to foraging behavior and remain lean for the remainder of their lives (Toth and Robinson, 2005). Loss of lipids is thought to be functionally important, minimizing energetic loss to the environment from the high attrition rate of foragers (Prado et al., 2020) and increasing flight performance by lowering body weight (Vance et al., 2009). The changed physiology also makes evolutionarily sense, as foragers are stripped of transferable resources that (otherwise) would be lost to the colony when foragers die in the field (Amdam and Page, 2005), increasing colony-level fitness.

Studies have shown that this lipid loss is accompanied by broad transcriptional (Ament et al., 2011) and proteomic changes (Chan et al., 2011) in the fat body but specific metabolic mechanisms remain unclear. There are a number of possibilities: lipids could be metabolically consumed to fuel the high flight activity of foragers, but this is unlikely because lipids are already low on the first day of foraging and do not correlate with foraging activity metrics (Toth and Robinson, 2005). A reasonable hypothesis is that lipid loss is induced by dietary changes, as nurses and foragers have very different food intake. Nurses consume large amounts of pollen, which is honey bees' primary source of protein and lipids, while foragers consume little or no pollen (Crailsheim et al., 1992). Pollen deprivation during the first week of life prematurely induces forager-like traits, such as degraded HGs, increased fat body lipolysis (Corby-Harris et al., 2019) and a shift in fat body transcriptome from protein to carbohydrate metabolism (Martelli et al., 2022). Experiments using pollen do not allow us to determine the specific nutrients responsible, but the simplest explanation is thus that foragers lose abdominal lipids because they consume much less lipid than nurses. Recent work using artificial diets shows that nurse-age bees regulate consumption to avoid over- or under-consuming lipid, aiming for an intake target of 1.25:1 protein to fat, indicating a substantial need for dietary lipids (Stabler et al., 2020). High fat diets also promote increased consumption and larger abdominal fat stores and HGs. Increasing the percentage of dietary lipid results in increased success in brood rearing in nurses (Arien et al., 2020). However, decreased lipid consumption alone is unlikely to fully explain abdominal lipid loss, as evidence indicates that de novo lipogenesis contributes substantially to lipid stores when dietary lipids are deficient (Toth et al., 2005), and this does not explain why foragers, which maintain higher hemolymph sugar levels than nurses (Mayack et al., 2019), do not convert these sugars into lipid stores, especially given that high sugar diets and resulting high hemolymph sugar levels are known to promote lipid storage in related insects such as Drosophila (Baenas and Wagner, 2022).

In addition to lipids, consumption of proteins differs between nurses and foragers. As foragers consume little or no pollen (Crailsheim et al., 1992), jelly received from nurses must be their primary source of protein (Crailsheim, 1992). The amount of jelly given to adult workers does significantly decrease with age, and foragers have 3- to 10-fold lower proteolytic capacity in the gut than nurses, suggesting that foragers receive and digest less total protein than younger workers (Crailsheim, 1992). This is further supported by work using artificial diets: experiments using paired diets in a geometric framework showed that the intake target of young, queenless bees shifted from a ratio of essential amino acids to carbohydrates (EAA:C) of 1:50 towards 1:75 over a 2 week period, accompanied by a reduced lifespan on diets high in EAAs (Paoli et al., 2014). Foragers required a diet proportionally much lower in protein (1:250) and also had low survival on diets high in EAAs (Stabler et al., 2020). Protein consumption is clearly important to nurse bees, but the relative contributions of dietary protein and lipid to fat body lipid metabolism remain unclear.

Based on these previous findings, we hypothesized that worker honey bees experience a developmental shift in the balance between lipid synthesis and breakdown during the nurse–forager transition. To test for the presence of this shift, we looked for changes in the fat body activity levels of two key enzymes, FAS and lipase, during a time frame around the transition from in-nest tasks to foraging. We then characterized the relevant contributions of feeding status and worker class by starving and feeding nurses and foragers. Finally, to test whether a difference in macronutrient consumption could account for the differences we found in FAS activity, we used artificial diets to manipulate the availability of protein and lipid to workers during the first 8 days of adult life.

Frames of sealed brood were collected from three different honey bee (Apis mellifera ligustica Spinola 1806) colonies kept at Arizona State University (Tempe, AZ, USA). Brood frames were kept overnight in ventilated mesh boxes in an incubator kept at 33°C and humidified by placing an open container of water at the bottom of the incubator. Newly emerged workers were brushed from the three frames the next day and combined to form a mixed population to increase genetic diversity of the experimental groups. Bees were then used in subsequent experiments.

To determine the effect of age on lipid metabolism, newly emerged bees were marked and released into colonies, and collected at 3, 8 and 13 days of age. On the day of sampling, workers were anesthetized on ice, euthanized and dissected. Bees of all age groups were collected systematically on a sampling day. For measurements of abdominal FAS and lipase activity and total protein, 8 bees were collected from each age group and pooled in groups of 2 as described below. A separate set of 8–10 bees per age group was collected to measure total abdominal lipids with an additional group of 10, 21 day old bees in one colony. The experiment was replicated twice in two different host colonies on two different sampling days.

To determine the effect of feeding status on the lipogenic capacity of bees in two different worker classes, nurses (8 day old bees observed inserting their heads into larval cells for at least 3 s) and foragers (28 day old bees collected when returning to the colony) were placed into (5×12×16 cm) Plexiglas and mesh cages (30 bees per cage) and fed 30% sucrose solution in 12 ml syringes ad libitum for 3 h. Bees were then either fed 30% sucrose solution for another 21 h (feeding treatment) or instead fed only water for the same period (starvation), for a total of 24 h of caging prior to sampling. For measurements of abdominal FAS activity and total protein, 6 bees were collected from each cage and pooled in groups of 2 as described below. The experiment was replicated 4 times.

To test whether a difference in macronutrient consumption between nurses and foragers could cause the differences in FAS activity we measured in the previous experiment, we designed an experiment to measure the effect of dietary protein and lipid on lipogenic capacity in workers. To address this question, we fed artificial diets with variable protein and lipid content to newly emerged bees until they were 8 days old, considered to be near peak nursing age. Such an approach of manipulating nurse-age bees to look for factors that precociously create forager-like physiology is commonly used to study the regulation of the nurse–forager transition (Toth et al., 2005; Antonio et al., 2008; Traniello et al., 2020). Bees were maintained in a humidified incubator at 31°C in cup cages (Evans et al., 2009). Similar to previous work (Arien et al., 2018), we created diets using honey as a carbohydrate source, soy as a protein source and corn oil as a lipid source (Table 1). The composition of the diets was 70.26–78.26% honey, which contains negligible protein and lipids; 0 or 21.74% soy protein isolate (composed of 92% protein, amounting to 20% net protein in the diet); 0 or 21.74% powdered sucrose, as a replacement for the soy protein isolate in protein-deficient diets; and 0 or 8% corn oil. These percentages of dietary protein and lipid were chosen based on past work showing that these values produced bees with the highest brood-rearing capacity (Arien et al., 2020). A two-factor design was used for diets to test the effect of presence/absence of these concentrations of protein and lipid. The dietary mixes were fed ad libitum to the bees in 1.5 ml Eppendorf tubes and replaced every 24 h. In addition, bees were fed 30% sucrose solution ad libitum in 10 ml syringes. Mortality was monitored daily in each cage and dead bees were removed. Consumption of the paste diets was calculated by mass changes every 24 h. Sucrose solution consumption was measured by mass difference between the start and end of the experiment along with a correction for evaporation of the solution. All consumption metrics were calculated by dividing cage-level mass changes by the number of bees still alive at each time point. Total macronutrient consumption was estimated by multiplying mass of sucrose solution and artificial diets by macronutrient fraction and summing all macronutrients together. Total calories consumed by each bee per day was estimated by converting macronutrient consumption in mg per bee per day into calories, using 4, 4 and 9 calories mg⁻¹ as estimations for protein, carbohydrate and lipid, respectively. For measurements of abdominal FAS activity and total protein, 6 bees were collected from each cage on day 8 and pooled in groups of 2 as described below. For measurements of HG size, a separate set of 6 bees per cage was collected. The experiment was replicated 2 times. For measuring the effect of dietary protein on total abdominal lipids, 5 bees were collected from each cage and the experiment was replicated 3 times.

Because recent findings indicate that FAS is regulated post-transcriptionally by autoacetylation based on nutrient availability (Miao et al., 2022), we chose to measure the activity level of fat body FAS activity rather than gene expression. Fat body FAS activity was assayed using a previously described assay (Lu et al., 2017) with minor modifications. In brief, workers were dissected and fat bodies (complete abdominal cuticle with adhering fat body tissue minus the stinger, gut, crop and ovaries) from two bees were pooled and homogenized in 250 µl of phosphate-buffered saline (7 mmol l⁻¹ NaCl, 2.7 mmol l⁻¹ KCl and 10 mmol l⁻¹ PO₄, pH 7.4) containing cOmplete^TM protease inhibitor cocktail (Roche) and kept on ice until further processing. The fresh samples were sonicated for 30 s (Qsonica Q800R2) and centrifuged at 5000 g for 10 min at 4°C. The supernatant was then collected and aliquoted: 133.2 μl was used immediately for the FAS activity assay, while the remainder was frozen and maintained at −20°C until it was used for the lipase activity and protein quantity assays.

To measure FAS activity, 33.3 μl of the supernatant was added to a buffer solution containing 163.3 μl of 2.0 mol l⁻¹ potassium phosphate buffer, pH 7.1, 16.7 μl of 20 mmol l⁻¹ dithiothreitol, 20 μl of 0.25 mmol l⁻¹ acetyl-CoA, 16.7 μl of 60 mmol l⁻¹ EDTA and 16.7 μl of 6 mmol l⁻¹ nicotinamide adenine dinucleotide phosphate (NADPH) in a 96-well microplate. To initiate the reaction, 33.3 μl of 0.39 mmol l⁻¹ malonyl-CoA was added to each well, while the same quantity of molecular-grade water was added to background wells. FAS activity was then measured as the oxidation of NADPH at 340 nm using a UV/VIS spectrophotometer (Biotek HT1) for 10 min at 37°C. A background correction was made for the oxidation of NADPH in the absence of malonyl-CoA.

Lipase activity was measured in thawed aliquots of the fat body supernatant using a fluorometric Lipase Activity Assay Kit (Cayman Chemical) as has been previously done with honey bee fat body tissue (Corby-Harris et al., 2019). Finally, the total amount of soluble protein in each sample was measured with a BCA Assay Kit (Thermo Scientific) according to the manufacturer's instructions, primarily as a normalization factor. Both FAS and lipase activity were calculated as whole abdominal values (units of nmol min⁻¹ abdomen⁻¹) and normalized to total protein in each sample (units of nmol min⁻¹ mg⁻¹). Enzyme activity data were collected using a BioTek Synergy H1 Plate Reader, housed in the Regenerative Medicine Imaging Facility at Arizona State University.

The HGs of bees that were fed or not fed protein and lipid for 8 days in cages were measured to determine whether HG size differed depending on diet. Bees were collected from each cage, their heads removed, flash frozen in liquid nitrogen, and maintained at −80°C until their glands were measured. For each bee, the HGs were dissected into phosphate-buffered saline, stained using Giemsa stain for 7 min, and then photographed using a Leica M205C stereoscope with a Leica DFC450 camera using Leica Applications Suite v4.5. The area of 10 acini per bee was measured using ImageJ and the mean value per individual was used for statistical analysis. The selection criteria used for acini in each image were that they had clear margins indicating that they were in focus, had an attachment point visible to the collecting duct and appeared to be of average size relative to other acini in the photograph.

Because the measurements were carried out by a researcher who was not blind to treatment identity, assessment of possible selection bias was carried out by randomly selecting 20 photos from the pool of 106 using the R package ‘random’ that generates random numbers from atmospheric noise, giving those 20 images to an observer blind to treatment identity, and asking them to select acini in the photos based on the described selection criteria. Selected acini were then measured, and the measurements compared with those made by the researcher. Measured acini size did not significantly differ based on whether the acini were selected by the researcher or an observer blind to study conditions (Mann–Whitney U=2854, n₁=n₂=75, P=0.878).

The abdominal fat body lipid content of 3, 8, 13 and 21 day old bees and of bees fed or not fed protein was measured. For the age-based colony samples, bees were collected, flash frozen in liquid nitrogen, and maintained at −80°C until their fat bodies were dissected as described above, following a previous approach (Corby-Harris et al., 2019). The tissue samples were Folch extracted overnight at 4°C in 2 ml of 2:1 chloroform:methanol with 25 mg ml⁻¹ of butylated hydroxytoluene (BHT) to inhibit lipase-based lipid degradation and lipid oxidation. The samples were then briefly vortexed and centrifuged for 15 min at 4°C. The bottom chloroform phase was retained and dried using a CentriVap at room temperature. The dried chloroform-soluble fraction was subjected to a sulfuric acid–vanillin–phosphoric acid assay as described previously (Van Handel, 1985). Sample absorbances were evaluated against a standard curve of corn oil. For the diet experiment, methods were identical except that fat bodies were dissected prior to flash freezing in liquid nitrogen.

To address the effect of age and replicate on abdominal FAS activity, normalized FAS activity, normalized lipase activity and total abdominal lipids, we used ANOVA with age, replicate and their interactions as factors. Following a significant effect of age, replicates were pooled and a post hoc Tukey's HSD test was used to determine which age groups differed from each other. As abdominal protein and abdominal lipase activity did not meet the assumption of normality even when log transformed, a Kruskal–Wallis test was used to assess the effect of age, while a Wilcoxon signed rank test was used for replicate effects.

For the nurse/forager starvation experiment, ANOVA was used for the effect of worker class, feeding status, replicate and their interaction on abdominal FAS activity. As abdominal protein quantity did not meet the assumptions of normality, a Kruskal–Wallis test was used to test the effect of treatment group (fed nurse, starved nurse, fed forager, starved forager) and replicate. A post hoc Dunn's test with a Bonferroni correction for multiple comparisons for the family-wise error rate (FWER) was used to determine which groups differed significantly from each other.

Similarly, ANOVA was used for the effect of dietary protein/lipid on abdominal FAS activity and protein and for consumption metrics. Normalized FAS activity did not meet the assumptions of normality, so ANOVA was used with log-transformed data. For the HG acini data, log transformation did not rescue the heteroscedasticity and Wilcoxon signed rank test was used for replicate effects while a Kruskal–Wallis test was used for the effect of dietary treatment with a post hoc Dunn's test as described above.

All analyses were carried out using RStudio version 2023.06.1.

To look for evidence that decreased synthesis or increased breakdown of lipids in the fat body is responsible for lipid loss during the nurse–forager transition, we first collected age-marked bees from hives and isolated their fat body tissue to measure protein and lipid levels and enzyme activity. Total abdominal protein was not affected by age (Fig. 1A; Kruskal–Wallis, χ²₂=2.99, P=0.22) or replicate (Wilcoxon signed rank test, P=0.81). Similarly, total abdominal lipids were not significantly affected by age (Fig. 1B; F_3,60=1.35, P=0.27) or replicate (F_1,60=1.00, P=0.32).

Abdominal FAS activity was significantly affected by both age (Fig. 1E; F_2,42=10.98, P<0.001) and replicate (F_1,42=10.64, P=0.002). Because the pattern was the same across replicates (Fig. S1), data from the two replicates were pooled for post hoc testing: bees at 13 days of age had significantly lower FAS activity than those at 3 days (P=0.02) and 8 days (P<0.001), while 3 and 8 day olds did not differ significantly (P=0.45). Because there was no significant trend in the protein data, we normalized FAS activity to the amount of protein in each sample, producing a similar result: there was a significant effect of both age (Fig. 1F; F_2,42=8.07, P=0.001) and replicate (F_1,42=4.10, P=0.049) on normalized FAS activity. Because the pattern was again the same across replicates (Fig. S2), data were pooled as previously: bees at 13 days of age had significantly lower FAS activity than those at 3 days (P=0.005) and 8 days (P=0.003), while 3 and 8 day olds did not differ significantly from each other (P=0.98).

Abdominal lipase activity was not affected by age (Fig. 1C; Kruskal–Wallis, χ²₂=2.68, P=0.26) or replicate (Wilcoxon signed rank test, W=289, P=0.99). Similarly, normalized lipase activity was not significantly affected by age (Fig. 1D; F_2,44=1.26, P=0.29) or replicate (F_1,44=1.45, P=0.23).

To test whether the lower FAS activity in 13 day old bees was due to a transition to foraging behavior or a difference in feeding status, we designed a two-factor experiment where 8 day old nurses and 28 day old foragers were either fed 30% sucrose solution ad libitum or starved (given only water) for 21 h in cages. Total abdominal protein was significantly lower in starved foragers than in the other three groups (Fig. 2A; Kruskal–Wallis, χ²₃=21.12, P<0.001; post hoc Dunn's test with a Bonferroni correction for FWER) and not significantly affected by replicate (Kruskal–Wallis, χ²₃=3.89, P=0.27). Abdominal FAS activity was significantly higher in nurses than in foragers (Fig. 1B; F_1,32=74.57, P<0.001) and in fed compared with starved bees (F_1,32=6.37, P=0.02). Replicate (F_3,32=1.73, P=0.18), worker class×treatment interaction (F_1,32=0.88, P=0.35), worker class×replicate interaction (F_3,32=0.35, P=0.79), treatment×replicate interaction (F_3,32=0.93, P=0.44) and worker class×treatment×replicate interaction (F_3,32=1.53, P=0.23) were all not significant. Because abdominal protein quantity was significantly affected by the starvation treatment (in foragers) and not by worker class, it was thus not used as a normalization factor and normalized FAS activity was not analyzed for this experiment.

To test whether dietary composition could produce the differences in FAS activity between nurses and foragers, we designed artificial diets that allowed us to manipulate dietary protein and lipid content and fed these diets to bees in cages for 8 days. Bees fed protein diets consumed significantly more sucrose solution (Fig. S3A; F_1,8=25.40, P=0.001) and less of the artificial diets (Fig. S3B; F_1,8=17.48, P=0.003); lipid presence in the diets did not affect consumption of sucrose solution (Fig. S3A; F_1,8=1.45, P=0.26) or artificial diets (Fig. S3B; F_1,8=0.61, P=0.46). Consumption of total macronutrients was not affected by either dietary protein (Fig. S3C; F_1,8=1.82, P=0.21) or lipid (F_1,8=1.79, P=0.22). Total estimated calories consumed per bee per day was not affected by dietary protein (Fig. S3D; F_1,8=2.42, P=0.16) but was significantly higher in bees fed lipids (F_1,8=5.44, P=0.048).

Abdominal protein was not significantly affected by dietary protein (Fig. 3A; F_1,20=0.25, P=0.62) or dietary lipid (F_1,20=2.12, P=0.16). Abdominal FAS activity was significantly higher in bees fed protein than in those deprived of protein (Fig. 3B; F_1,21=63.78, P<0.001) but was not significantly affected by dietary lipid (F_1,21=0.80, P=0.38). ANOVA of the log-transformed data showed that there was a significant positive effect of dietary protein (Fig. 1C; F_1,20=53.75, P<0.001) and negative effect of lipid (F_1,20=6.53, P=0.02) on normalized FAS activity.

To further address whether the diets produce bees with nurse-like physiology, we also measured the size of HGs and the total quantity of abdominal lipids. A Kruskal–Wallis test showed a significant difference in mean HG gland acinus area between treatment groups (Fig. 3D; χ²₃=35.59, P<0.001). A post hoc Dunn's test with a Bonferroni correction for FWER showed that dietary protein but not fat increased mean HG gland acinus area. Abdominal lipid was significantly affected by replicate (Kruskal–Wallis, χ²₂=8.16, P=0.02). However, because the pattern across replicates was similar (Fig. S4), data were pooled to determine treatment effects: protein-fed bees had significantly higher abdominal lipids than protein-deprived bees (Fig. 4; Wilcoxon signed rank test, W=60, P=0.03).

In this study, we found that honey bee worker fat body lipogenic capacity was significantly affected by age, role within the colony and dietary macronutrient availability. Lipogenic capacity declined as workers aged past peak nursing behavior (8 days old; Fig. 1C,D) and, whether they were starved or fed, nurses had consistently higher lipogenic capacity than foragers (Fig. 2B), though it is important to note that we did not fully decouple worker age from role, as nurses were also much younger than foragers. We also tested how availability of dietary macronutrients affects lipogenic capacity, and found it was significantly decreased by carbohydrate starvation in both nurses and foragers collected from colonies (Fig. 2B) and by protein (but not lipid) deprivation for bees kept in a controlled environment in cages for 8 days (Fig. 3B,C). Protein deprivation similarly decreased the size of HGs and total abdominal lipid content (Figs 3D and 4), matching past studies showing that sufficient protein consumption during development is necessary for workers to develop nurse-like physiology, while protein deprivation causes premature development of forager-like physiology (DeGrandi-Hoffman et al., 2010; Corby-Harris et al., 2016; Omar et al., 2017; Martelli et al., 2022).

Our finding here that FAS activity and not lipase activity or total protein is downregulated in older workers and foragers supports the hypothesis that reduced capacity to synthesize lipids is specifically targeted for downregulation during the nurse–forager transition. Reduced lipogenic capacity provides one mechanistic explanation for why foragers become and remain lean, despite consistently maintaining higher hemolymph sugar levels than nurses (Mayack et al., 2019). A limitation of our study is that, because we used a spectrophotometric assay for measuring the activity of the FAS enzyme complex rather than directly measuring the rate of lipogenesis itself, we cannot conclude definitively that stored fat body TAG correlates with the measured FAS activity. Our measurement of total abdominal lipid content provides some support for this correlation, as treatment groups with higher FAS activity generally had higher total abdominal lipids (Figs 1B and 4). However, it should be noted that we did not conclusively show that FAS-derived fatty acids are stored as TAG in fat body lipid droplets: synthesized fatty acids may instead be directly catabolized via β-oxidation, used as a source for biosynthesis of lipid molecules used for cuticular hydrocarbons, tracheal waterproofing, cell membranes or hormones (Huang et al., 2022), or transported by lipoproteins to the head glands to be used for the fatty acid fraction of jelly fed to developing bees.

The differences we found in lipogenic capacity between nurses and foragers persist even when carbohydrate feeding status is controlled by feeding bees ad libitum for 24 h in a controlled laboratory environment: while acute starvation did decrease abdominal FAS activity in both nurses and foragers, worker class had a much larger effect. Feeding on carbohydrates thus clearly upregulates lipogenic capacity, but it does so within a dynamic range determined by the social role of the bee. Interestingly, while in nurses the downregulation of FAS activity during starvation seems to be specific, with no reduction in total abdominal protein, starved foragers had significantly reduced total protein in the abdominal fat body relative to fed foragers and to nurses (Fig. 2). This suggests that the downregulation of FAS activity in starved foragers is less specific and part of a broad metabolic response to catabolize fat body proteins to provide energy during starvation. The lack of this response in nurses could be because they lack the necessary proteolytic machinery necessary for rapid catabolism of protein reserves, or because they are better buffered against nutritional stress, perhaps because of their higher lipid stores (Toth and Robinson, 2005), higher levels of hemolymph storage proteins such vitellogenin that can be catabolized instead of fat body protein (Nakaoka et al., 2008), or their lower maximal metabolic rate (Schippers et al., 2010) or lower activity level. Further work will be needed to address age class-specific differences in protein catabolism.

Our finding that protein deprivation causes reduced lipogenic capacity (Fig. 3B,C), lower abdominal lipids (Fig. 4), and smaller HGs (Fig. 3D) in nurse-age bees fits into a broad range of research linking dietary protein with development and maintenance of nurse physiology. A nurse's function in the colony is primarily to produce and distribute royal jelly to the queen, larvae and other workers, acting as a net nutritional source while other bees are nutritional sinks (Crailsheim, 1991). To support the synthesis of royal jelly, nurses are the primary consumers of pollen in a hive (Crailsheim et al., 1992) and they have much higher levels of digestive proteases than other bees so that they can process difficult to digest pollen into useful nutrition (Crailsheim and Stolberg, 1989). Royal jelly is a complete food source composed of protein, carbohydrates and lipids (Wang et al., 2016). It is secreted from the HGs and mandibular glands (MGs) of the head. Adequate dietary protein during the first week of life is critical for the development of the physiological features enabling effective nursing behavior: in bees deprived of pollen, the HGs degrade between 3 and 8 days (Corby-Harris et al., 2022), and our replication here of this pattern using artificial diets confirms the protein component of pollen as the primary factor necessary for the development of large, nurse-like HGs. Pollen deprivation similarly decreases total HG and MG protein (Peters et al., 2010) and vitellogenin hemolymph titer (Bitondi and Simões, 1996), and there is a positive correlation between pollen in the worker intestines and fat body vitellogenin (Wegener et al., 2018). Vitellogenin is used by nurse bees as an amino acid donor for producing royal jelly, and high vitellogenin titers are linked with nursing (Amdam et al., 2003). Because successful nursing behavior requires protein, it is beneficial for colonies if nurses that do not have sufficient dietary protein transition to foraging to increase pollen and hence protein flow into the hive (Amdam and Page, 2005).

While here we found no effect of dietary lipids on HG size (Fig. 3D), another study found that feeding bees lipids did increase HG size (Stabler et al., 2020). The difference may be due to the older age of the bees in that study (10 days), differences in consumption based on the state of the diets (liquid instead of semi-solid), or the different lipid source (soy lecithin instead of corn oil). Supporting this last idea, past work has shown that bees fed corn oil specifically, which is rich in linoleic acid, had smaller HGs than those fed oil with a more balanced omega 6:omega 3 ratio (Arien et al., 2015). Thus, a limitation of this study is that we did not address the extent to which lipid composition may be more important for worker development than the absolute concentration of lipid in the diet. Additionally, it is important to note that we did not assess the effect of dietary lipids on abdominal lipid stores, but past work shows that both quantity and composition of dietary lipids strongly affect worker lipid stores (Arien et al., 2020; Stabler et al., 2020; Wang et al., 2021). Lipids are an important but variable part of royal jelly, ranging from approximately 2% to 7% in one study (Ferioli et al., 2014). This variability is probably because lipids are the most variable composition in pollen, ranging from 1% to 13% (Campos et al., 2008). Lipids are likely a valuable resource to colonies, as increasing the dietary lipid concentration causes nurse-age bees to consume more total food (Stabler et al., 2020) and have increased success in brood rearing (Arien et al., 2020).

We found mixed evidence regarding whether lipogenic capacity is affected by dietary fat. While groups fed lipids generally had lower FAS activity, this difference was only significant when normalizing to total protein (Fig. 3C). Caution should be taken in interpreting this result, as, while the difference was not significant, bees deprived of lipids had lower total abdominal protein, potentially biasing the result (Fig. 3A). It is worth nothing that bees fed lipids consumed more total calories (Fig. S3D), a finding that is consistent with past work (Stabler et al., 2020), suggesting that any potential upregulation in FAS activity in bees deprived of lipids is not due to increased consumption. It is possible that workers increase de novo lipogenesis to compensate for a deficiency in dietary lipids or decrease lipogenesis when dietary lipids are adequate, but if so, the effect is very small compared with the effect of dietary protein.

Vitellogenesis has been evolutionarily co-opted in honey bees to provide larval nutrition from a sterile worker class (Amdam et al., 2003). Much of nurse physiology seems to be optimized for high levels of fat body vitellogenesis, and this may extend to the high nurse fat body FAS activity we measured here as well: in planthoppers, knockdown of FAS lowered expression of both Vg and VgR (Cheng et al., 2023), suggesting that de novo synthesized fatty acids are used for vitellogenesis, possibly because vitellogenin requires lipidation by specific fatty acids before secretion (Silversand and Haux, 1995).

Our finding that dietary protein deprivation prematurely induces a forager-like reduction in lipogenic capacity and fat body lipids suggests that the dietary switch during the nurse–forager transition could be an important upstream signal lowering synthesis and storage of lipids in the fat body. This study does not directly address why foragers consume less protein than nurses, but a range of studies support that that they do: experiments using paired diets in a geometric framework showed that the intake target of young, queenless bees shifted from an EAA:C ratio of 1:50 towards 1:75 over a 2 week period, accompanied by a reduced lifespan on diets high in EAAs. Foragers required a diet high in carbohydrates (1:250) and also had low survival on diets high in EAA (Stabler et al., 2020). Foragers consume little or no pollen (Crailsheim et al., 1992), and thus jelly received from nurses must be their primary source of protein (Crailsheim, 1992). The amount of jelly given to adult workers does significantly decrease with age, suggesting that foragers receive less total protein than younger workers (Crailsheim, 1992).

While we focused on synthesized lipids in this study, dietary lipids also likely play a role, as foragers consume less pollen than nurses, and pollen is the primary source of dietary lipids. We did not address the relative contributions of dietary and de novo lipids to total lipid stores in workers, but past work has shown that treating workers with a fatty acid synthesis inhibitor significantly decreases abdominal lipids only when workers are deprived of pollen (Toth et al., 2005), suggesting that de novo lipogenesis becomes more important for lipid storage when dietary lipids are not available. As foragers consume minimal amounts of pollen (Crailsheim et al., 1992), reduced lipogenic capacity is likely to be especially important in maintaining lipid stores at the low levels found in foragers.

Taken together, our findings implicate a developmental reduction in protein intake targets as a central factor in mediating metabolic plasticity in honey bee workers. This supports the hypothesis that honey bee worker division of labor evolved through co-option of, and novel, social inputs into the regulation of appetite, but much more experimental work will be needed before the synthesis of a complete model linking hive inputs through regulation of protein appetite into mediation of caste-specific metabolic physiology.

We thank Cahit Ozturk for his assistance with bee husbandry, Jenna Dobson for help with HG acini quantification, Vanessa Corby-Harris for advice on lipase activity and lipid quantification assays, and Geraldine Wright and Daniel Stabler for help with the design of artificial diets. We acknowledge resources and support from the Regenerative Medicine Imaging Facility, part of the Biosciences Core Facilities at Arizona State University, and thank Zerrin Uzum for advice on imaging hypopharyngeal glands. We thank Tim Linksvayer as well as two anonymous reviewers whose comments helped to improve the manuscript.