Brain mitochondrial bioenergetics change with rapid and prolonged shifts in aggression in the honey bee, Apis mellifera

Energy metabolism is so fundamental to neuronal signaling that metabolic dysfunction underlies most diseases associated with cognitive impairment, including Alzheimer's disease and traumatic brain injury (Sullivan et al., 2005; Lin and Beal, 2006). However, despite the conventional view that the brain invariably demands and consumes high levels of energy substrates such as glucose, a growing body of work suggests that there is variation in brain cellular energy production and use, and moreover, that this variation actually plays an important neuromodulatory role in the healthy brain (Plaçais and Preat, 2013; Chandrasekaran et al., 2015; Hollis et al., 2015; Plaçais et al., 2017; Rittschof and Schirmeier, 2017). Studies have found correlations between brain cellular energy metabolic flux (including differential activities of glycolysis and oxidative phosphorylation pathways) and phenotypes such as aggression (Chandrasekaran et al., 2015; Rittschof and Schirmeier, 2017), anxiety (Hollis et al., 2015; van der Kooij et al., 2017), and learning and memory processes (Plaçais and Preat, 2013; Plaçais et al., 2017). A fundamental role for energy metabolism in neuromodulation could explain why organismal metabolic rate is so often linked to behavioral variation (Biro and Stamps, 2010; Bushman et al., 2014; Rittschof et al., 2015b), and moreover, why the activity of traditional neuromodulators (e.g. biogenic amines) is often insufficient to explain behavioral variation in all contexts (Asahina, 2017). However, relatively little is known about the mechanisms by which neuronal energetic state influences and responds to behavior-relevant signaling events; these properties are fundamental to elucidating an energetic component to neuromodulation.

Mitochondria are particularly critical to fulfilling neuronal energetic demands, as they house oxidative phosphorylation, an oxygen-consuming pathway that efficiently produces ATP for the myriad bioenergetic costs of signaling (Kasischke et al., 2004; Hall et al., 2012); as such, neurons contain large quantities of mitochondria in both vertebrate and invertebrate brains (Rittschof and Schirmeier, 2017). Mitochondria are also important candidates for modulation of energy metabolism as they show exceptional plasticity in terms of their shape, structure, intracellular location and respiration rate (Kasischke et al., 2004; Chan, 2006; Ly and Verstreken, 2006; Detmer and Chan, 2007; Kann and Kovács, 2007; Jendrach et al., 2008; MacAskill and Kittler, 2010; Brenmoehl and Hoeflich, 2013; Hara et al., 2014; Picard and McEwen, 2014; Picard et al., 2015). This variation, which has cognitive impacts (Hara et al., 2014), is linked to a range of factors including stress levels, developmental stage and intracellular energetic demands (Chan, 2006; MacAskill and Kittler, 2010; Picard and McEwen, 2014; Picard et al., 2015, 2018). Plasticity in respiration rate is particularly intriguing as a neuromodulatory mechanism because such changes can occur rapidly in response to neuronal conditions, including signaling events (Kasischke et al., 2004; Nicholls, 2005; Kann and Kovács, 2007). Plasticity in respiration is usually assessed in terms of unidirectional declines – decreased mitochondrial respiration rate is associated with neural pathology (Sullivan et al., 2005; Lin and Beal, 2006). However, recent studies have also found that there is temporal, age and region-specific variation in mitochondrial respiration in the healthy brain (Sauerbeck et al., 2011; Pandya et al., 2016; Plaçais et al., 2017), and moreover, that among-individual differences in brain mitochondrial respiration predict behavioral tendencies (Li-Byarlay et al., 2014; Hollis et al., 2015; Pandya et al., 2015). There is, however, little understanding of how genetic and experiential factors set individual mitochondrial respiration levels in critical areas of the brain associated with cognition and behavioral phenotypes. We evaluate the possibility that social cues that induce a behavioral response also rapidly modulate neuronal mitochondrial respiration. Over time, these experiences could alter the metabolic set-point of the brain, thus having an impact on behavioral tendencies.

In the current study, we evaluated whether shifts in mitochondrial respiration could serve as an energetic mechanism of neuromodulation for a naturally occurring behavioral phenotype, honey bee aggression. Honey bee aggression, expressed by worker bees defending their nest, is rapidly induced by a social alarm cue (Collins and Blum, 1983). Aggressive tendency, i.e. responsiveness to alarm cues, also increases as adult worker bees age over a time frame of weeks (Breed et al., 1990). This shift in behavioral tendency is responsive to social conditions in the colony and broader ecological conditions (Huang and Robinson, 1992; Guzmán-Novoa and Page, 1994; Hunt et al., 2003). Previous studies correlated changes in worker bee aggression with shifts in brain mRNA abundance for genes encoding energy metabolic proteins (Alaux et al., 2009; Chandrasekaran et al., 2015). Whole-brain transcriptomic data show that genes that encode oxidative phosphorylation proteins decline in expression within 1 h of alarm cue exposure, and also decline with age (Alaux et al., 2009; Chandrasekaran et al., 2015). This phenomenon appears to be localized to neurons, and pharmacological manipulations in the honey bee demonstrate a causal and socially responsive relationship between oxidative phosphorylation activity and aggression (Li-Byarlay et al., 2014). Given this established connection to neuronal mitochondrial energy metabolism, and the multiple time scales for variation in aggressive behavior, the honey bee provides an opportunity to evaluate whether changes in mitochondrial respiration play a direct role in experience-dependent modulation of a behavioral phenotype.

Here we take a variety of in vitro and in vivo approaches to demonstrate a link between brain mitochondrial bioenergetics and aggressive behavior. We used laboratory respiration analyses of intact, functioning mitochondria to determine whether brain mitochondria responded to a social alarm pheromone that rapidly induces aggressive behavior. We used similar analyses to determine whether brain mitochondrial respiration was correlated with age-related variation in aggression. In addition, we manipulated diet and energetic substrate availability for intact animals to assess whether enhanced brain mitochondrial respiration affected aggressive behavior in vivo. Carbohydrate or caloric restriction is well established as a method to enhance brain mitochondrial function in mammals (Davis et al., 2008; Lutas and Yellen, 2013); we assessed whether bees fed restricted diets showed changes in aggression. In a second experiment, we manipulated beta-hydroxybutyrate, a ketone body that, when fed to mammals, rescues injury and disease-induced declines in brain mitochondrial respiration (Maalouf et al., 2007). In previous studies, we found that brain ketone body concentration declined for individuals exposed to an aggression-inducing social cue (Chandrasekaran et al., 2015), suggesting that ketone body metabolism may play a role in modulating brain mitochondrial respiration in the context of honey bee aggression. We performed feeding experiments to determine whether beta-hydroxybutyrate had an impact on aggressive behavior. In addition, we assessed whether beta-hydroxybutyrate, provided in vitro to intact mitochondria, affected respiration for bees of different ages, allowing us to evaluate the possibility that age-related differences in responsiveness to aggressive cues reflects differential plasticity in brain mitochondrial respiration. Together, these experiments assess whether modulation of brain mitochondrial bioenergetics may serve as a foundation for an energetic basis of neuromodulation.

We performed analyses from May to October 2016 in Lexington, KY, USA. The honey bees we used represent a variety of different genotypes and strains originating primarily from Apis mellifera ligustica Spinola 1806 and A. m. carnica Pollman 1879. Three unique colonies headed by naturally mated queens are represented in this portion of the study (see details below), and for ketone body feeding experiments we utilized different colonies from similar sources.

We performed experiments in July and August 2015 in State College, PA, USA. The bees originated from colonies headed by naturally mated queens of mixed genetic backgrounds (primarily derived from A. m. ligustica).

For all brain assessments presented here, we analysed whole brains, with all glands removed, following previous studies (Alaux et al., 2009; Rittschof and Robinson, 2013; Rittschof et al., 2014; Chandrasekaran et al., 2015; Rittschof, 2017). We first assessed whether the aggressive behavioral response that results from alarm cue exposure is correlated with a shift in brain mitochondrial respiration. We performed field-based collections of unexposed control bees and compared respiration rates of isolated intact brain mitochondria to bees collected 5 and 60 min following exposure to alarm pheromone (following Alaux et al., 2009; Chandrasekaran et al., 2015; Rittschof et al., 2015a). Upon collection, we anesthetized bees on ice to slow brain biochemical changes and thus achieve a ‘snapshot’ of brain mitochondrial function. Unexposed control bees were foragers we collected as they returned to the colony. To minimize collection time and handling in our study, which can affect bee aggression levels, we considered foragers any bees returning from a trip outside the colony, a category that encompasses primarily pollen, nectar and water foragers, but could also include bees on orientation flights. We immediately transferred these bees to a plastic bag and placed on ice until the dissection step, which began within 5 min of collection. Meanwhile, immediately following the control collection, we applied 5 µl of a 1:10 solution of isopentyl acetate (Sigma-Aldrich, St Louis, MO, USA), the primary component of honey bee alarm pheromone (Collins and Blum, 1983) and mineral oil (Sigma-Aldrich) to a piece of filter paper, and placed this paper in the center of the landing board at the hive entrance. This treatment causes a subset of bees (known as soldiers) to emerge from the hive and display aggressive behaviors, which include lunging at and biting returning nest mates, or the forceps or hands of the experimenter (Breed et al., 1990; Rittschof et al., 2015a). After 1 min, we removed the filter paper, and collected approximately 30 of the most aggressive bees (the ones displaying the most conspicuous aggressive behaviors defined above) with forceps. We placed these bees in two 7.0×8.0×9.0 cm plexiglass boxes to serve as the 60 min collection. We placed the boxes at the colony entrance on the landing board until 60 min total time had elapsed from the initial application of the stimulus. Because bees can starve under these circumstances, these bees were provided with honey during this time frame (Alaux et al., 2009; Chandrasekaran et al., 2015). Following the 60 min collection (which took approximately 3 min), and once 5 min total had elapsed from the end of the alarm stimulus, we used a vacuum to collect approximately 30 additional bees displaying aggressive behaviors as they emerged from the colony. We immediately anesthetized these bees on ice to make up the 5 min treatment group (Chandrasekaran et al., 2015). After 60 min, the boxed bees (above) were quickly transferred to a plastic bag and anesthetized on ice. For both treatment groups, dissection began within 5 min of collection, followed immediately by mitochondrial extraction and respiration analyses. We alternated the dissection of 5 min and control bees so that these were completed within approximately 1 h. We then completed the 60 min dissections, which took an additional 30 min. Mitochondrial extractions and respiration analyses were blocked by treatment for the three groups in order to control for plate-to-plate variation in assay measurements. Data presented represent two identical collections and analyses performed for the same colony across two separate days (spaced approximately 3 weeks apart). The timing constraints of this rapid assay and collection prevented us from measuring the response for additional unrelated colonies, and so the results should be interpreted with caution with respect to generalizability.

To evaluate changes in brain mitochondrial activity as bees age, we compared ‘old’ adult worker bees with ‘young’ adult worker bees collected from the same colony. Old bees were foragers collected as they returned to the hive (see above), and young bees were 2-day-old adults collected as 1-day-olds as they emerged from their brood frame. These groups met our experimental goal, which was to evaluate brain mitochondrial respiration for bees that show strong differences in aggression as a function of age (Rittschof and Robinson, 2013; Rittschof, 2017). However, here we do not differentiate the separate impacts of age and behavior on brain mitochondrial respiration. We collected groups of 20–30 foragers into 7.0×8.0×9.0 cm plexiglass boxes the afternoon prior to dissection and respiration measurements. Bees were provided, overnight in an incubator (34°C, relative humidity uncontrolled), ad libitum 50% sucrose (bees of this age typically eat primarily sugars; Paoli et al., 2014), dispensed in 1.7 ml tubes with two small holes at the base. We collected young bees from the same colony as the old bees to control for colony level variation in respiration measurements. We collected a frame of emerging brood in the afternoon 2 days prior to dissection and respiration analysis. The brood frame was housed in a laboratory incubator (34°C), and the morning following frame collection we collected 1-day-old bees into plexiglass boxes (see above) in groups of 30–50 individuals (Rittschof et al., 2015a). As per their pollen-based natural diet (Ament et al., 2010; DeGrandi-Hoffman et al., 2010), we provided these bees ad libitum with an approximately 2:1 honey:pollen mixture and returned them to the incubator. We performed dissections and respiration analyses the following day when bees were 2-day-old adults. For both age groups, we decapitated live bees and immediately dissected the brains. Dissections, mitochondrial extractions and respiration analyses were performed simultaneously for both age groups. The same protocol was used for in vitro analyses of beta-hydroxybutyrate (BHB, see below) on mitochondrial function, using a different source colony from the collections described above.

All reagents were purchased from Sigma-Aldrich. Brains were quickly dissected in ice-cold isolation buffer (215 mmol l⁻¹ mannitol, 75 mmol l⁻¹ sucrose, 0.1% bovine serum albumin, 1 mmol l⁻¹ EGTA, 20 mmol l⁻¹ Hepes) to preserve mitochondrial function (Pandya et al., 2015). We kept brains in isolation buffer on ice until mitochondrial processing, which began within 1.5 h of dissection. We pooled two brains per sample. Intact mitochondria were isolated using procedures adapted from Pandya et al. (2011). We minced samples in an all-glass Dounce homogenizer containing isolation buffer. Homogenates were centrifuged at 1300 g for 3 min at 4°C and transferred the supernatant to a fresh 2 ml tube, topped off with isolation buffer, and centrifuged at 13,000 g for 10 min. The remaining pellet was re-suspended in 450 µl of isolation buffer. We performed synaptosomal disruption in a nitrogen cell bomb incubated at 8.3 N mm^-2 for 10 min to maximize the yield of total mitochondria from synaptic and non-synaptic populations. The 450 µl solution was then transferred to a new 1.7 ml tube with an additional 1000 µl of isolation buffer, and centrifuged at 13,000 g for 10 min. The supernatant was discarded and the pellet re-suspended in 20 µl isolation buffer. We used 6 µl of solution to assess mitochondrial protein concentration (in duplicate) using a bicinchoninic acid (BCA) assay (Sauerbeck et al., 2011).

All reagents were purchased from Sigma-Aldrich, except for rotenone, oligomycin and FCCP, which were purchased from Biomol (Plymouth Meeting, PA, USA). We analysed bioenergetic function using the Seahorse Biosciences XF24 Flux Analyzer (procedure adapted from Maalouf et al., 2007; Sauerbeck et al., 2011; Kim et al., 2015). Preliminary tests showed that 5 µg of mitochondrial protein yielded a robust signal, and so that standardized amount was loaded for each sample. Measurements were performed in duplicate. We injected pyruvate+malate+ADP, oligomycin, FCCP and rotenone+succinate sequentially through ports A–D, respectively, in the Seahorse Flux Pak cartridges, to final concentrations of 5 mmol l⁻¹ (pyruvate), 2.5 mmol l⁻¹ (malate), 1 mmol l⁻¹ (ADP), 1 µg ml⁻¹ (oligomycin), 1 µmol l⁻¹ (FCCP), 100 nmol l⁻¹ (rotenone) and 10 mmol l⁻¹ (succinate). We measured state 3 respiration (ADP-driven), state 4 respiration (following ADP depletion with complex V inhibition), complex I-driven maximum uncoupled respiration and complex II-driven maximum respiration (Fig. S1). Results are reported (in terms of oxygen consumption rate, pmol min⁻¹) for state 3 respiration, a direct indicator of mitochondrial bioenergetics (Sauerbeck et al., 2011). To assess the effects of ketone bodies on brain mitochondrial bioenergetics, we performed analyses on old and young brains in which we added BHB (1 mmol l⁻¹) directly to the mitochondrial respiration buffer immediately prior to starting measurements (Maalouf et al., 2007). In mammals, this procedure typically results in enhanced state 3 respiration (Maalouf et al., 2007).

Diet restriction is a common method used to manipulate brain mitochondrial respiration in vivo (Davis et al., 2008). Dietary manipulations that enhance brain mitochondrial function are varied, but all involve sugar restriction (Freeman et al., 2007; Lutas and Yellen, 2013; Paoli et al., 2013), and one such treatment is overall caloric restriction (van Praag et al., 2014). To test impacts of diet restriction on behavior in the honey bee, we provisioned caged bees housed in a laboratory incubator. We collected 1-day-old bees (see ‘young’ bee description above) from two colonies and transferred them in groups of 10 to plexiglass boxes. All boxes were provided with sucrose-based diets and ad libitum water in 1.7 ml feeder tubes (as above). We kept bees in groups of 10 to minimize isolation effects on physiology and behavior (Huang and Robinson, 1992). Although this set-up prevented us from precisely monitoring diet consumption on a per-bee basis, in a separate experiment with colored food, we found that all bees consumed at least some food during a given 24 h period (data not shown).

We chose to manipulate sucrose concentration as opposed to other nutrients (pollen, honey) because it contains no other factors that could influence metabolism (lipids, proteins or plant secondary compounds), and the adult honey bee diet is predominantly sugar (Wright et al., 2014). We provided 40 µl of sucrose solution per bee per day, which is well within the daily gut capacity of a worker honey bee. Our restricted diet (DR) was 4 mg sucrose per bee per day, a minimum survival requirement (Brodschneider and Crailsheim, 2010). Our unrestricted diet (UR) control was five times this level, 20 mg sucrose per bee per day, which was essentially ad libitum feeding. For the last 24 h period of the experiment (day 5), we sub-divided the treatments into two additional groups. For half of the bees in each treatment, we fed the same diet, and for the other half, we switched the diets (UR to DR, and DR to UR) in order to compare the effects of short- versus long-term diet restriction on behavior. We assayed aggression when bees were 6 days old (see ‘Behavioral assays’ section below).

For the DR treatment, bees invariably consumed 100% of the diet every day. For the UR treatment, bees typically consumed an average of 71% of available food daily (s.e.m. 3.4%). See Table S1 for details about daily food consumption on a range of sugar-restricted diets. Food was replaced between 08:00 and 10:00 h every morning regardless of the quantity consumed to ensure that sucrose concentrations remained consistent. Tables S2 and S3 provide information on mortality as a result of diet restriction. Mortality over the course of this experiment varied, ranging from 0 to 100% per group of ten bees in the DR group (and typically 0% for UR groups). For the behavioral results presented in the main text, we preferentially selected groups of bees with 100% survivorship to control for the possibility that there were differential effects of diet restriction on mortality as a function of aggression level.

We investigated whether feeding with BHB (Sigma-Aldrich), which is known to enhance brain mitochondrial function in mammals (Freeman et al., 2007; Lutas and Yellen, 2013; Stanley et al., 2014), alters aggression in the honey bee. In pilot trials with DR bees (above), we found that bees readily consumed BHB dissolved in water and sucrose solutions, with no significant increase in mortality (Table S4). This outcome was consistent across a range of doses (58–5800 µg ml⁻¹). For the behavioral assessments described here, we selected doses of 58 and 29 µg ml⁻¹, and fed BHB in high concentration sucrose solution (50% weight/weight). We observed no mortality at these doses. Between 13:00 and 15:00 h, we collected returning foragers (as above) for BHB treatment, as these bees are highly responsive to aggression cues (Robinson, 1987; Breed et al., 1990), and showed in vitro plasticity in respiration following BHB treatment (see Results). We transferred bees using forceps to Petri dish arenas in groups of four. Bees were provided with 800 µl (essentially ad libitum) 50% sucrose, with 0.09 µl of BHB (58 or 29 µg ml⁻¹) or a water control. We kept bees overnight in a 34°C incubator, and aggression assays were performed the following morning.

Aggression assays and analyses are extensively described in previous studies (Li-Byarlay et al., 2014; Rittschof et al., 2015a; Rittschof, 2017). We used the laboratory-based Intruder Assay (described below), which provides multi-modal information to individuals in order to invoke an aggressive response. In previous studies, we correlated the results of this assay with colony-level aggressive responsiveness and variation in individual brain biomarkers of aggression (Rittschof and Robinson, 2013; Li-Byarlay et al., 2014; Rittschof et al., 2014, 2015a; Rittschof, 2017). For the diet restriction experiments, we combined bees into groups of four (one bee per treatment, selected at random) in order to assess relative aggression levels among treatments within each group (Rittschof, 2017). In order to combine bees that had been kept separate throughout the experiment (and thus may show aggression towards foreign bees from other groups), we transferred bees from their boxes into plastic bags and anesthetized them on ice (3–5 min) (Rittschof, 2017). We then marked bees on their dorsal thorax with Testors paint with a treatment-specific color and placed them in 20×100 mm Petri dish arenas for the behavioral assays. We left bees for 1 h in a temperature-controlled room (25–30°C) in full light to revive from the anesthesia prior to behavioral assessment. We did not provide bees with additional food during this time frame, and we observed minimal trophallaxis among group members prior to the assays. For BHB feeding experiments, we assessed aggression displayed by groups of four bees of the same treatment (all BHB treated or all control treated). Comparing aggression this way, across groups as opposed to among bees within a group as above, has lower precision (Rittschof, 2017), but this strategy ensured that treatment and control bees did not exchange BHB via trophallaxis when combined into a group together. Thus the BHB behavioral experiments may underestimate aggression differences as a result of BHB treatment.

To assay aggression, we introduced a foreign intruder bee (collected on the day of the assay from a colony not otherwise involved in the experiment) to a group of four bees in a Petri dish arena, and recorded aggressive actions towards the intruder by group members over a 2 min time period. We tallied interactions and weighted them for severity to produce a total aggression score (Li-Byarlay et al., 2014; Rittschof et al., 2015a). Aggressive behaviors include antennating the intruder with and without mandibles open, biting the intruder, flexing the abdomen, and stinging the intruder (Richard et al., 2008). We tallied aggressive behaviors on a per-treatment basis, and blocked analyses for group identity in the diet restriction experiment. In the diet restriction experiment, we assayed general activity levels in addition to aggression for treatment bees (not used in the aggression analysis). We assessed activity by observing how many times bees of a given treatment crossed two orthogonal lines on a piece of paper under the Petri dish over the course of a 3 min period.

We report statistical tests in the Results section. We analysed respiration data using non-parametric statistical tests or generalized linear models (GLM) as data violated the parametric assumption of equal variances. We analysed behavioral data using GLMs. All statistical tests were performed in JMP Pro 12.1.0. (SAS Institute, Cary, NC, USA).

Exposure to an aggression-inducing social cue (alarm pheromone) was associated with a rapid decline in whole-brain mitochondrial respiration (oxygen consumption rate, OCR). Five minutes after exposure, OCR was significantly lower than the unexposed control, and it remained low up to 60 min later (Wilcoxon test, χ²₂=9.79, P<0.0075; Fig. 1). Although mitochondrial respiration was still significantly lower than control 60 min following exposure (Wilcoxon each-pair test, 5 min versus control: Z=−3.13, P<0.0018, 60 min versus control: Z=−2.18, P<0.03, 60 min versus 5 min: Z=0.29, P<0.78), there was substantial variation in OCR at the 60 min time point, and some individuals re-gained an OCR similar to control. Measured for bees at rest, whole-brain intact mitochondrial respiration increased with age. Forager bees, representing some of the oldest and most aggressive bees in the colony, showed significantly higher mitochondrial respiration (OCR) compared with relatively docile 2-day-old bees (Wilcoxon test, two-tailed, Z=−2.65, P<0.0081; Fig. 2, Fig. S1).

In mammals, diet restriction enhances brain mitochondrial function, with implications for neural excitability and cognition (van Praag et al., 2014). We found that 6 day diet restriction enhanced aggression in honey bees, while 24 h restriction did not significantly alter aggression (GLM, log-link: χ²₃=10.48, P<0.0149; effect tests: long-term restriction, χ²₁=9.96, P<0.0016; short-term restriction, χ²₁=0.51, P<0.47; interaction term, χ²₁=0.20, P<0.66; Fig. 3). This suggests that the behavioral effects of diet are not due to acute starvation, and instead reflect a stable physiological change. The effects of diet restriction are at least somewhat specific to aggressive behavior, as general activity levels showed a different pattern, i.e. predominantly short-term impacts on behavior (Fig. 4). Aggression results from a broader range of restricted diets are shown in Fig. S2.

One mechanism by which diet restriction enhances mitochondrial function is by increasing the generation and use of ketone bodies (e.g. BHB) as metabolic fuel (Lutas and Yellen, 2013). We found that honey bees, like mammals, are responsive to ketone bodies both in vitro and in vivo. In vitro, whole-brain intact mitochondria provided with BHB as a metabolic substrate showed significantly enhanced mitochondrial function for older bees only (GLM, log-link: χ²₃=1665.02, P<0.0001; effect tests: age, χ²₁=1513.12, P<0.0001; BHB treatment, χ²₁=40.31, P<0.0001; interaction term, χ²₁=43.10, P<0.0001; Fig. 5). Notably, this assay replicated the age-related differences in respiration observed in Fig. 2 for a second, unrelated honey bee colony. Additional assessments of age-related differences in brain mitochondrial respiration for other honey bee colonies have shown similar patterns (Fig. S1; C.C.R., H.J.V., J.H.P and P.G.S., unpublished data), suggesting that the age and behavior-related shift in brain mitochondrial respiration is generalizable across genotypes. In vivo, forager bees fed overnight with sugar supplemented with BHB showed significantly enhanced aggression relative to control, sugar-fed individuals (GLM, log-link: χ²₁=5.24, P<0.0221; Fig. 6). A second experiment with blinded behavioral assessment and half the dose of BHB gave similar results (GLM, log-link: χ²₁=5.01, P<0.0251, data not shown).

Here we show that both rapid and stable variation in honey bee aggression is associated with a shift in brain mitochondrial respiration. We also find that two treatments known to increase brain mitochondrial function in mammals, diet restriction and ketone body feeding, cause increased aggression in the honey bee. Although we did not directly examine the impact of diet and ketone body treatment on brain mitochondrial respiration, together, these results suggest that plasticity in mitochondrial bioenergetics is a neural mechanism for energetic modulation of a behavioral phenotype.

Our results suggest, for the first time, that an acute social interaction may induce rapid shifts in brain mitochondrial bioenergetics. The rapid decline in brain mitochondrial respiration is similar in extent to that observed in the context of acute traumatic brain injury (Xiong et al., 1997). However, unlike injury-induced declines in mitochondrial function, which result in neuronal apoptosis (Xiong et al., 1997; Sullivan et al., 2005), honey bee brain mitochondria presumably regain function following the aggressive response, and appear to begin to do so within an hour of the original stimulus. This result builds on a growing body of work linking brain bioenergetics to variation in behavioral phenotypes. For example, in rodents, pioneering work demonstrated that oxidative phosphorylation activity in the nucleus accumbens is causally linked to anxiety behaviors and the outcomes of social dominance interactions (Hollis et al., 2015; van der Kooij et al., 2017). Our findings in this study propose a source of such individual variation in brain energetic state and behavior, i.e. that acute social interactions modulate the energetic set-point of the brain. In a previous study where we demonstrated a causal link between brain oxidative phosphorylation and aggression, we found that long-term prior social history modulates aggressive behavior and the response to pharmacological treatments that manipulate brain energetic state (Li-Byarlay et al., 2014). This provides additional support for the notion that energetic state is set by social experience. The possibility that an acute social interaction may have lasting impacts on brain energetic state is further supported by the finding that such an interaction has an impact on both mitochondrial function and metabolic gene expression (Alaux et al., 2009), a suite of socially induced changes at two distinct levels of biological organization that likely interact to impact neural energetic state, signaling and behavior. These interpretations are based on the hypothesis that neural signaling associated with alarm pheromone processing causes a shift in mitochondrial bioenergetics. However, it is important to note that from our data we cannot state conclusively whether these energetic shifts are caused by pheromone-related neural signaling, or are correlated with the expression of aggressive behavior. Forthcoming studies directly address these alternatives.

Although we did not attempt to explicitly address the role of age versus behavior in modulating brain mitochondrial function in the current study, we found that stable increases in aggression that typically accompany ageing are associated with increased whole-brain mitochondrial respiration, matching the findings of previous studies assessing respiration rates at the organismal scale in the honey bee (Harrison and Fewell, 2002). Similarly, dietary manipulations known to increase brain mitochondrial respiration (diet restriction and ketone body feeding) caused increased aggression, suggesting that these treatments invoke a shift to ‘old bee-like’ brain mitochondrial respiration. These results at the physiological level are surprising because they contradict transcriptomic data which showed, for very similar collection methods, that brain gene expression for oxidative phosphorylation proteins declined in older foragers relative to younger in-hive bees, as well as in bees exposed to alarm pheromone compared with control (Alaux et al., 2009; Li-Byarlay et al., 2014; Chandrasekaran et al., 2015). In apparent agreement with gene expression data, a recent study assessing mitochondrial bioenergetics in the whole honey bee head showed higher respiration in nurses compared with older foragers (Cervoni et al., 2017). However, in addition to processes in the brain, whole-head results from this prior study could reflect the substantial differences in the size and function of the metabolically active hypopharyngeal glands in nurse versus forager bees. Differences in young bee age could also distinguish our current results from those from this previous study. Notably, our prior work showing enhanced aggression with pharmacological inhibition of oxidative phosphorylation did not assess respiration levels at the time of the behavioral assay (Li-Byarlay et al., 2014). Thus, more work is needed to reconcile the relationships between mitochondrial activity, gene expression and behavior, and how these relationships are regulated comparing prolonged and rapid shifts in aggression.

Regardless of differences in regulatory relationships, divergent patterns in mitochondrial activity across rapid and prolonged shifts in aggression suggest potential neuronal functional outcomes of variation in brain bioenergetics. The mitochondrion is an organelle with many important connections to neuronal signaling and excitability. Variation in mitochondrial respiration could reflect several important functional roles beyond ATP generation, including reactive oxygen species regulation and apoptosis initiation (Sullivan et al., 2005; Wang et al., 2013; Yin et al., 2014; Pandya et al., 2015). One possibility is that plasticity in mitochondrial respiration in the honey bee brain reflects calcium-buffering activity. Mitochondria take up, retain and release intracellular calcium ions (Ca²⁺) in neurons during signaling events. Ca²⁺ buffering regulates second messenger signaling, prevents excitotoxicity, and facilitates neurotransmission (Brookes et al., 2004; Chan, 2006; Kann and Kovács, 2007; Chouhan et al., 2012; Hroudova and Fisar, 2013), all of which are essential for the response to aggressive cues. Importantly, mitochondrial Ca²⁺ uptake during a signal event can cause a decline in mitochondrial respiration (as in the alarm pheromone case), while higher baseline respiration enhances buffering capacity (as in the ageing case; Pandya et al., 2013; Wang et al., 2013; Yin et al., 2014). Thus a mechanism like calcium buffering during a signaling event could explain the divergent mitochondrial respiration patterns across aggression time scales in the honey bee. One limitation of our current data is that it investigates mitochondrial function at the whole-brain level. Future studies, focused on assessing brain regional variation in mitochondrial bioenergetics as a function of age and social experience, will be critical to interpret the neuronal functional outcome of changes in mitochondrial respiration in the context of aggression (Picard et al., 2018).

A large number of studies have evaluated the brain genomic correlates of behavioral variation in the honey bee (for a review, see Zayed and Robinson, 2012) and other species. However, few studies have connected these broad-scale transcriptional patterns to higher level physiological processes that could be linked directly to neural function. Such connections are critical for determining how behavioral expression is modulated by social stimuli and understanding the role of brain gene expression dynamics in these processes (Rittschof and Hughes, 2018). We have the opportunity to make such connections in the context of honey bee aggression and neural energetics. The physiological and transcriptomic levels probably interact to set the energetic state of the brain. For example, a rapid shift in mitochondrial oxidative phosphorylation may change the cellular redox state, which can serve as a potent intracellular signal to the neuronal soma, resulting in a shift in gene expression (Dröge, 2002; Brookes et al., 2004; Maalouf et al., 2007). In the honey bee brain, the precise relationship between mitochondrial respiration and gene expression, i.e. whether transcription compensates for, or parallels, physiological changes, is unknown. However, an understanding of this relationship may be critical to determine how social experiences at different time points throughout life exert either lasting or reversible effects on behavioral expression (Rittschof and Hughes, 2018). In honey bees, energy metabolic genes show high degrees of DNA methylation (Foret et al., 2012), which could mean that epigenetic regulation underlies experience-dependent changes in brain energetic state. A recent study demonstrated that epigenetic changes can occur within 120 min of alarm pheromone exposure in the honey bee (Herb et al., 2018). Future work will evaluate gene expression and mitochondrial respiration dynamics, as well as the relationship between these energetic processes and traditional neuromodulatory systems that are associated with variation in the honey bee alarm pheromone response (Nouvian et al., 2018).

Our finding that brain energetic state distinguishes age-related changes in honey bee behavior suggests that the regulation of the energetic state of the brain, along with other known hormone and neuromodulatory changes, may be critical to division of labor and behavioral maturation in worker bees (Sullivan et al., 2000; Le Conte et al., 2001; Schulz et al., 2002; Sullivan, 2003; Amdam et al., 2005). As bees age and become more responsive to aggressive cues, we found that they not only show enhanced mitochondrial function, but mitochondria were more responsive to BHB, an alternative energy substrate to glucose. In previous work, we showed that exposure to aggressive stimuli has significant effects on brain ketone body levels, suggesting that changes in metabolic substrate use may underpin variation in brain mitochondrial function in the honey bee, similar to mammals (Chandrasekaran et al., 2015). Another interpretation of this result is that older honey bee brain mitochondria show greater functional plasticity in addition to higher baseline respiration rates compared with younger bees. Older honey bees are more behaviorally responsive to alarm pheromone compared with younger bees (Robinson, 1987). Perhaps mitochondrial functional plasticity is correlated with capacity for neuronal excitability and thus capacity for behavioral response. A similar mechanism could explain why older bees show greater learning abilities (Ray and Ferneyhough, 1997); a recent study in Drosophila melanogaster found that plasticity in neuronal energy metabolism is required for normal learning and memory functions (Plaçais and Preat, 2013; Plaçais et al., 2017). Future work could assess the relationship between capacity for behavioral plasticity and plasticity in mitochondrial bioenergetics, as well as the localization of these shifts to particular regions of the brain associated with sensory input and/or integration. Though the basis of age-related differences in BHB sensitivity and mitochondrial bioenergetics is unknown, other animals show brain- and tissue-specific mitochondrial respiration rates. In mammals, studies suggest that age-related changes in sensitivity to ketogenic substrates reflect quantity differences in ketone body metabolizing enzymes in the mitochondria (Prins, 2008). More metabolically active mitochondria may also simply show enhanced sensitivity to metabolic substrate availability.

In mammals, diet restriction rapidly and robustly enhances brain mitochondrial function (Davis et al., 2008; Hartman et al., 2013). We found that diet restriction increases aggressive tendency in the honey bee, but only over a prolonged time period. Diet restriction across a range of severity appears to have an impact on aggression (Fig. S2), and the precise tolerance of individuals to various levels of diet restriction and starvation probably varies across genotypes. In the honey bee, diet restriction may stimulate a cascade of physiological processes that are associated with accelerated ageing (Toth et al., 2005; Münch et al., 2013). As bees age, they transition to a high-sugar diet (Paoli et al., 2014), which causes a shift in insulin and TOR (target of rapamycin) signaling (Ament et al., 2008, 2011), and a decline in fat stores (Toth et al., 2005). Diet restriction, which triggers lipid loss (Toth et al., 2005), could induce the same mechanism. One consequence of these physiological shifts is an increase in whole-organism metabolic rate (Schippers et al., 2010; Rittschof et al., 2015b), which, based on our current data, appears to extend to the brain. The precise mechanisms that mediate this change in mitochondrial performance are unknown, but similar insulin-related shifts are observed in mammalian contexts as well. For example, diabetes, which is characterized by disrupted insulin signaling, is a risk factor for brain mitochondrial dysfunction and Alzheimer's disease (Potter et al., 2010; Jauch-Chara et al., 2012; Cholerton et al., 2013; Butterfield et al., 2014; De Felice et al., 2014). Insulin-sensitive shifts in brain mitochondrial function, which occur in a disease context in humans, may occur in reverse under conditions of normal ageing in the honey bee.

We demonstrate that temporal behavioral variation in a natural context is linked to robust shifts in brain mitochondrial function. Thus variation in brain mitochondrial respiration, which has been studied almost exclusively in the context of injury or disease, appears integral to neuronal signaling events that accompany modulation of aggression in healthy individuals. Despite significant differences in neural structure, mammals and insects show a surprising degree of concordance in brain metabolic regulation and energetic substrate use (reviewed in Rittschof and Schirmeier, 2017). Our results therefore contribute to a growing body of evidence for the role of neural energetic state in modulating the expression of cognitive and behavioral phenotypes across a range of animal species including humans (Dong et al., 2009; Lewis et al., 2009; Potter et al., 2010; Shannon et al., 2013; Dong et al., 2015; Hollis et al., 2015; Shannon et al., 2016). Understanding plasticity in mitochondrial respiration in the healthy brain may provide unique insights into the regulatory mechanisms associated with mitochondrial pathologies. Moreover, the regulation of neural energetic state is a critical factor to be considered alongside classical neuromodulator studies towards a more comprehensive understanding of the mechanistic basis of temporal and experience-dependent variation in behavioral expression.

We acknowledge Malinda Spry for assistance in data collection.