G. E. Robinson, Department of Entomology and Institute for Genomic Biology, Neuroscience Program, University of Illinois at Urbana-Champaign, Urbana, IL, USA. E-mail: firstname.lastname@example.org
Worker honey bees (Apis mellifera) undergo a process of behavioral maturation leading to their transition from in-hive tasks to foraging – a process which is associated with profound transcriptional changes in the brain. Changes in brain gene expression observed during worker behavioral maturation could represent either a derived program underlying division of labor or a general program unrelated to sociality. Male bees (drones) undergo a process of behavioral maturation associated with the onset of mating flights, but do not partake in division of labor. Drones thus provide an excellent reference point for polarizing transcriptional changes associated with behavioral maturation in honey bees. We assayed the brain transcriptomes of adult drones and workers to compare and contrast differences associated with behavioral maturation in the two sexes. Both behavioral maturation and sex were associated with changes in expression of thousands of genes in the brain. Many genes involved in neuronal development, behavior, and the biosynthesis of neurotransmitters regulating the perception of reward showed sex-biased gene expression. Furthermore, most of the transcriptional changes associated with behavioral maturation were common to drones and workers, consistent with common genetic and physiological regulation. Our study suggests that there is a common behavioral maturation program that has been co-opted and modified to yield the different behavioral and cognitive phenotypes of worker and drone bees.
Division of labor underpins the ecological success of social insects, allowing colonies to adaptively shift their demography to exploit changing environments (Wilson 1985). Division of labor is intrinsically regulated through a series of physiological changes occurring during the lifespan of adult workers, and extrinsically regulated by the social environment. Workers of the honey bee, Apis mellifera, transition from performing brood care and in-hive tasks to foraging after 2–3 weeks of adulthood (Winston 1987). This transition is regulated, in part, by increasing titers of juvenile hormone (JH) in the hemolymph (Robinson 2002), and changes in expression of at least several genes (Ben-Shahar et al. 2002, 2004; Nelson et al. 2007). Behavioral maturation is also associated with changes in the expression of thousands of genes in the worker brain (Kucharski & Maleszka 2002; Whitfield et al. 2003, 2006) and is transcriptionally modulated by pheromones present in the social environment (Alaux et al. 2009a; Grozinger et al. 2003).
Drones also undergo a process of behavioral maturation that involves spending a period of time inside the hive prior to engaging in mating flights. Behavioral maturation in drones and workers share several common features. First, JH titers increase prior to the onset of mating flights, and treatment with a JH analog hastens the onset of mating flights in drones (deOliveiraTozetto et al. 1997; Giray & Robinson 1996), similar to the relationship between JH and the onset of foraging in workers (Robinson 2002). Second, there is a genotypic correlation between the age at onset of foraging in workers and the age at onset of mating flights in drones across colonies (Giray & Robinson 1996; Rueppell et al. 2006). Third, the ontogeny of outside activity is also associated with increases in the volume of the neuropil of the mushroom bodies – the insect brain region implicated in learning and memory – in both drones and workers (Fahrbach et al. 1997). These findings suggest common physiological and genetic regulation of behavioral maturation in honey bees.
The biology of drones and workers share many interesting similarities, but also differ in several key aspects. Both drones and workers exhibit behavioral maturation and engage in flight activities that require spatial navigation and homing abilities. However, only workers engage in division of labor and its constituent activities (brood care, colony defense, foraging, dance language, etc.). Furthermore, worker behavioral maturation is socially regulated, while drone behavioral maturation is insensitive to social conditions (Giray & Robinson 1996). Because drones do not participate in division of labor, it is reasonable to assume that they therefore lack the molecular components that regulate behavioral plasticity associated with division of labor in workers.
Drones provide a unique opportunity for dissecting behavioral maturation into common processes as well as worker- and drone-specific processes. We undertook a large-scale microarray experiment to compare and contrast the transcriptional changes associated with behavioral maturation of drones and workers, and to determine unique transcriptional changes associated with worker division of labor.
Materials and methods
Experiments were performed at the University of Illinois Bee Research Facility with Apis mellifera source colonies managed using standard practices. We obtained drones and workers of known age by incubating (33°C) combs containing drone or worker late-stage pupae from two colonies headed by naturally mated queens. One-day-old adults (within 24 h) were either collected in liquid nitrogen for genetic analyses, or were marked with a spot of paint on the thorax and reintroduced into host colonies. The 21-day-old workers were collected in liquid nitrogen as they returned from foraging by partially blocking the colony entrance. The 21-day-old drones were collected in liquid nitrogen by inspecting combs inside the hive. These two time-points were chosen to reflect two distinct behavioral states in both workers and drones. A previous study on workers showed that changes in brain gene expression associated with behavioral maturation were completed approximately midway between our chosen time-points (Whitfield et al. 2006); 21-day-old workers and drones are thus expected to be in a stable behavioral state representing foraging and mate-finding, respectively.
We followed published protocols for microarray experiments for honey bees using a previously characterized microarray (accession A-MEXP-755) containing probes for 13 440 predicted genes (Alaux et al. 2009a). mRNA was extracted from partially lyophilized individual brains. We used MessageAmp II aRNA Amplification kit (Ambon, Austin, TX, USA) for producing Cy3 or Cy5 labeled aRNA, which were hybridized to the oligoarray. Our experiment quantified gene expression in 79 individual bees using 100 microarrays employing dye-swaps in a loop design (Table S1). Our samples belonged to two source colonies and consisted of 1-day-old drones (N = 20), 21-day-old drones (N = 20), 1-day-old workers (N = 19), and 21-day-old workers (N = 20). Arrays were scanned using an Axon 4000B scanner and fluorescence data were quantified using GENEPIX software (Molecular Devices Inc., Sunnyvale, CA, USA). We filtered the raw data by excluding probes with a fluorescence intensity that was less than the median intensity of the negative control spots for a given dye. We also removed probes that target highly expressed genes in the hypopharyngeal glands (which are adjacent to the brain), as well as probes that target honey bee pathogen genes (Alaux et al. 2009a). Duplicate probes were averaged and then normalized using a LOWESS transformation prior to analyses. Microarray data were deposited into ArrayExpress (accession: E-MTAB-523).
Power and limits of microarrays
Our main interest is to examine changes in brain gene expression associated with behavioral maturation and the long-duration behavioral states found in honey bee workers and drones. We have previously shown microarrays to be very powerful (i.e. we can detect subtle ∼1.25-fold changes in expression) for examining long-duration behavioral states in the bee (Cash et al. 2005). We recognize, however, that microarrays may be limited in visualizing transient changes in brain gene expression (Cash et al. 2005), but such changes are not of primary importance for our study. The honey bee oligoarray uses a single probe per gene; we may thus miss some biologically relevant forms of regulation, such as alternative splicing or post-translational modification of resulting proteins. Finally, microarray technology does not immediately lead to knowledge about the relationship between transcription, neuroanatomy, brain circuits and behavior, although it may offer clues and suggest testable hypotheses. Nevertheless, microarrays still provide a wealth of knowledge on the relationship between gene expression in the brain (i.e. the first phenotype) and long-term behavioral states, as shown herein and elsewhere (Chandrasekaran et al. 2011).
We used a linear mixed effect model implemented using restricted maximum likelihood to analyze the normalized log2-transformed data for each gene to account for the effects of the following factors: array, dye, colony, bee, behavioral maturation (1-day vs. 21-day old), sex (worker vs. drone), as well as maturation by sex interactions. We also assessed the effect of behavioral maturation in workers and drones separately using t tests. We only included probes with expression data for at least 74 out of 79 individuals. Excluding probes based on this criterion did not reduce our ability to detect sex- or age-biased gene expression. The number of drones and workers with missing expression values were highly correlated across probes (r2 = 81.6%, P < 2.2e−16), and the ratio of missing vs. present expression values was not significantly associated with caste for any of the excluded probes [χ2 tests; false discovery rate (FDR) >0.05 for all probes]. Similar results were obtained when comparing 1- and 21-day-old bees at excluded probes (r2 = 84.9%, P < 2.2e−16; χ2 tests; FDR >0.05 for all probes). Unless otherwise stated, significant gene lists were based on a FDR of <0.05.
As 21-day-old workers were collected as they returned from the field, while 21-day-old drones were collected from the hive, our sampling scheme potentially confounded sex with activity on gene expression in older bees. As a result, we filtered out 1960 genes known to be transcriptionally affected by flight activity in workers (Naeger et al. 2011) resulting in a final dataset of 6574 genes. Analyses before and after removal of activity genes yielded qualitatively similar statistical and functional (Gene Ontology) results indicating that our interpretations are robust. For example, the number of genes with sex-biased expression remains essentially constant when comparing 1- and 21-day-old bees separately in both filtered and unfiltered datasets. Furthermore, using the filtered dataset did not substantially change the functional analysis and the relative proportion of worker- and drone-specific behavioral maturation genes: 26% and 24% were worker-specific, while 45% and 50% were drone-specific before and after filtering, respectively. Finally, the common behavioral maturation gene set in the unfiltered dataset was significantly enriched for an annotation cluster associated with locomotory behavior and flight behavior, indicating that our filter did indeed target genes associated with flight activity in older bees. All the results presented herein were based on analyses of the filtered dataset.
We generated a matrix of expression values per bee per gene as estimated using a linear model with dye, colony and bee as factors. Individual gene expression estimates were zero-centered and used in principal component analysis (PCA) using the R package PRCOMP (R Development Core Team 2009). We excluded five outlier bees (three drones and two workers) from PCA. These bees had the lowest measures of RNA quality and had aberrant gene expression profiles that distorted higher-order relationships in PCA. We imputed missing expression data for a few individuals using the average gene-specific expression value. We used the R package PAMR 1.42.0 to rank genes that best predicted sex in our study (Tibshirani et al. 2002).
We used real-time quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) to quantify gene expression for 18 genes in 75 out of the 79 bees used in our microarray experiments following published protocols (Alaux et al. 2009a) with one deviation: an exogenous spike-in control (Arabidopsis' RCP) was used to normalize gene expression levels across individuals. We used SYBR green chemistry to measure the relative expression of a specific gene (in triplicate reactions) as well as the spike-in control for each individual. Expression is quantified using a standard curve generated from genomic DNA. Expression levels were measured for 18 genes, half of which showed significant sex-biased gene expression, while the other half showed no significant differences based on microarray data. PCR primers are described in Table S2. We normalized each gene's expression using the spike-in control, and used t tests to compare normalized gene expression between drones and workers. We excluded one gene (GB17452) from the statistical analysis because of inconsistent amplification.
Functional annotation analyses
To explore the functional relevance of differentially expressed genes, we performed functional classification analyses as implemented in DAVID 6.7 (Dennis et al. 2003; Huang et al. 2009) using default parameters. The analysis proceeds by generating clusters or groups of functionally similar annotation terms from different databases (Gene Ontology, KEGG, etc) found in a list of differentially expressed genes. The annotation clusters are then tested for enrichment given a background gene list comprised of all analyzable genes. Functional annotation clusters highlight both high- and fine-level biology, thereby providing better biological insight than traditional Gene Ontology Enrichment analysis (Huang et al. 2009). We converted honey bee genes into Drosophila melanogaster orthologs identified by reciprocal best BLAST match (kindly provided by C. Elsik). We supplemented this list with fly genes identified as the best match to bee genes using BLASTP. Overall, 4690 out of the 6574 probes analyzed herein had annotated fly orthologs.
Comparisons with previous microarray studies
To examine if differences in brain gene expression between drones and workers reflect sex or female caste, we compared our results with a queen vs. worker brain gene expression study, which used a cDNA microarray (Grozinger et al. 2007). If drone–worker expression differences are a function of sex, we would expect genes with drone–worker differences in expression to not show queen–worker differences in expression. However, if drone–worker differences are a function of female caste, we would expect genes with drone–worker differences to also show queen–worker differences in expression. We thus examined if genes that are significantly differentially expressed when comparing drones and workers in our study were also significantly differentially expressed between queens and workers in Grozinger et al. (2007). We also contrasted gene expression differences in 1- and 21-day-old workers and drones with that found between nurses and foragers (Alaux et al. 2009a).
Several studies have documented changes in worker brain gene expression in response to experimental treatments known to affect worker division of labor and behavior, including methoprene (JH-analog), manganese (related to malvolio), cyclic GMP (cGMP, related to foraging) (Whitfield et al. 2006) and queen mandibular pheromone (Grozinger et al. 2003). We examined if the lists of genes significantly regulated by the above treatments in workers were enriched for genes that are common to both worker and drone behavioral maturation.
The effect of behavioral maturation and sex on brain gene expression
Both behavioral maturation and sex were associated with profound changes in brain gene expression in worker and drone honey bees (Fig. 1a). Behavioral maturation was associated with the largest changes in brain gene expression, with 3163 genes showing significant changes in expression between 1- and 21-day-old bees (Behavioral Maturation Gene Set; Table S3). Sex was also a major factor, with 2728 genes showing changes in expression between drones and workers (Sex Gene Set; Table S3). In addition, 1349 genes exhibited a sex by maturation interaction whereby the pattern of differences between 1- and 21-day-old bees differed between drones and workers. The effects of both behavioral maturation and sex on brain gene expression are clearly prevalent as visualized by PCA (Fig. 2): behavioral maturation dominates axis 2, and sex dominates axis 1 and 3. The number of genes with sex-biased gene expression did not substantially change over time; we detected 2393 genes with sex-biased expression in 1-day-olds and 2495 genes in 21-day-olds. We detected a strong effect of colony on brain gene expression for 2264 genes.
Validation of microarray results
We validated the expression of eight sex-biased genes using qRT-PCR; all showed significant differences in brain expression between drones and workers in the same direction as the microarray data (Figure S1). We also measured the expression of nine genes that were not significantly different in our microarray analysis; three of these genes showed significant differences between drones and workers, albeit at much lower significance.
Functional significance of sex- and maturation-associated differences in brain gene expression
Functional classification analyses suggested biological processes and molecular functions that differ between drones and workers. The top 100 genes that best predicted sex in our dataset (Top 100 Drone/Worker Gene Set; Table S3) were significantly enriched for a major annotation cluster associated with receptor-linked signaling, sensory perception and detection and response to light (Table 1). The entire drone vs. worker gene set was enriched for eight annotation clusters (Table 1); these clusters were associated with metabolic processes and oxidative phosphorylation, membrane and transmembrane cellular compartments and humoral immune response. The latter cluster contains many transcription factors and signaling peptides associated with neuronal development (e.g. E2f, Rac1, zfh1, bsk, kay, lola). Of note was a major cluster that contained many terms associated with cytochrome P450, secondary metabolite biosynthesis and catabolism and steroid biosynthesis. This cluster contained many genes involved in hormone and neurotransmitter synthesis and metabolism, including several cytochrome P450 genes, juvenile hormone epoxide hydrolase (Jheh1, JH catabolism, Mackert et al. 2010), Henna (Hn, dopamine and serotonin synthesis; Coleman & Neckameyer 2004) and Tyramine beta hydroxylase (Tbh, octopamine synthesis; Monastirioti et al. 1996). Jheh1 was higher in drones, while Hn and Tbh were higher in workers. This cluster also included the honey bee ortholog of Shade (CYP314A1), which converts ecdysone into active 20-hydroxyecdysone and is known to be expressed in the brains of workers (Yamazaki et al. 2011).
Table 1. Summary of enriched functional annotation clusters for differentially expressed genes in the brain associated with sex and behavioral maturation in the honey bee
Summary of database terms
The enrichment score is the −log10 geometric mean of individual annotation term enrichment P-value in a cluster; values higher than 1.301 represent a geometric mean of P < .05 for the annotation terms in the cluster. Annotation databases: GeneOntology, INTERPRO, SMART, PIR Keyword and Superfamily, UPSEQ, KEGG Pathway and COG Ontology.
Top 100 drones/worker gene set
Membrane targeting protein with C2-domain (Ca2+dependent)
Sensory perception, Cognition Detection of light stimulus & Response to light stimulus
Plasma membrane, Cell junction
Sex gene set
Humoral immune response
Citrate cycle (TCA cycle), Generation of precursor metabolites and energy & Oxidative phosphorylation
Organic acid biosynthetic process, Cellular amino acid biosynthetic process
Organic acid biosynthetic process, Cellular amino acid biosynthetic process
We also detected several annotation clusters associated with behavioral maturation in bees, including two clusters associated with cytoskeleton organization, and a third associated with amine and organic acid biosynthesis (Table 1). The latter cluster contained both Hn and Tbh, which were highly expressed in older bees. Tbh converts tyramine to octopamine, a neuromodulator known to be involved in regulating worker division of labor (Schulz et al. 2002) and dance language (Barron et al. 2007, 2009). A forth annotation cluster contained 34 genes with immunoglobulin-like domains. Such domains are commonly found in transcription factors and signaling molecules involved in neuronal development and behavioral regulation in Drosophila (e.g. Down syndrome cell adhesion molecule Dscam, the netrin receptor unc-5, defective proboscis extension response dpr, frazzled fra and terribly reduced optic lobes trol).
Common and novel patterns of behavioral maturation
We found 2143 and 3147 genes that show differences in expression between 1- and 21-day-old workers and drones, respectively (Fig. 1b; Table S3). A total of 1578 genes were common to both gene lists and showed the same direction in expression between young and old bees in both sexes (Fig. 1b). We detected 565 and 1569 genes that were uniquely differentially expressed during the process of worker or drone behavioral maturation, respectively (Fig. 1b; Table S3).
Genes showing common patterns of maturation in workers and drones were significantly enriched for clusters annotated with oxidative phosphorylation and metabolism (N = 3), immunoglobulin-like domains and gland development and histolysis (Table 2). The latter cluster consisted of 34 genes that are generally involved in morphogenesis, and many of which have documented roles in neuronal development (e.g. netrin receptor unc-5, semaphorin-2A, frazzled and salvador).
Table 2. Summary of enriched functional annotation clusters for genes differentially expressed in the brain associated with common behavioral maturation, as well as drone-specific and worker-specific behavioral maturation
Oxidative phosphorylation & Oxidoreductase activity, acting on NADH or NADPH
Exocrine system development, Gland development & Programmed cell death
Transmembrane transport, Energy coupled proton transport down electrochemical gradient, Nucleotide biosynthetic process, ATP biosynthetic process & ATPase activity
Regulation of Notch signaling
Regulation of transcription, RNA polymerase II transcription factor activity & RNA biosynthetic process
Small GTPase mediated signal transduction, GTP binding & GTPase activity
The gene set unique to worker behavioral maturation was enriched for two functional annotation clusters associated with the regulation of the Notch signaling pathway and signal transduction, and transcription initiation and regulation (Table 2). The gene set unique to drone behavioral maturation was enriched for one functional annotation cluster associated with GTP binding and GTPase activity (Table 2).
Comparisons with previous honey bee microarray studies
To examine if sex or female caste was the primary driver of drone–worker differences in brain gene expression, we compared our data with a study comparing brain gene expression in queens and workers (Grozinger et al. 2007). A total of 472 genes with drone/worker differences in our study were assayed for queen/worker differences by Grozinger et al.; more than 60% of these genes did not show differences in expression between queens and workers (Fisher's Exact test P = 0.006 assuming a ratio of 50:50). These results suggest that sex is the major driver of the observed differences in brain gene expression between drones and workers.
We examined the overlap of gene expression changes associated with behavioral maturation in workers and drones to those associated with the nurse to forager transition in workers (Alaux et al. 2009a). Genes significantly upregulated in 1-day-old workers were enriched for genes that are upregulated in nurses, while genes upregulated in 21-day-old workers were enriched for genes that are upregulated in foragers (χ2 = 250.82, df = 1, p < 0.0001). Similarly, genes significantly upregulated in 1-day-old drones were enriched for ‘Nurse’ genes, while genes upregulated in 21-day-old drones were enriched for ‘Forager’ genes (χ2 = 273.21, df = 1, P < 0.0001). We also cross-referenced the common behavioral maturation gene list with genes known to be regulated by JH, manganese, cGMP and queen mandibular pheromone. We found that genes regulated by methoprene, manganese and queen mandibular pheromone were significantly enriched in the common behavioral maturation gene set (Table 3). We also found a small but significant enrichment of genes regulated by cGMP in the common behavioral maturation gene list (Table 3).
Table 3. The common behavioral maturation gene set is enriched for genes regulated by experimental treatments affecting worker division of labor and behavior
Columns marked + and NS indicate the number of genes with significant or not significant changes in brain expression caused by the experimental treatments, respectively.
Queen mandibular pheromone
Behavioral maturation in workers is associated with profound changes in brain gene expression, as seen previously (Alaux et al. 2009a; Cash et al. 2005; Whitfield et al. 2003) and in the present study. Our study, however, for the first time indicates that behavioral maturation largely involves changes in brain gene expression shared by both drones and workers, despite the fact that drones do not engage in any form of division of labor. We supported this hypothesis by finding 1578 genes with significant and consistent changes associated with behavioral maturation in both workers (74%) and drones (50%). Indeed, gene expression profiles of 1-day-old drones were nurse-like while those of 21-day-old drones were forager-like, suggesting common patterns of behavioral maturation unrelated to division of labor. Most of the common behavioral maturation genes were enriched for annotation terms associated with energy metabolism and mitochondrial function. A recent study discovered differences in mitochondrial function associated with behavioral maturation in workers (Alaux et al. 2009b), but our analysis suggests that such changes are more generally associated with behavioral maturation. These results suggest that the systems that regulate division of labor are built upon molecular pathways that control more fundamental aspects of behavioral maturation.
Both workers and drones spend a period of time inside the hive before embarking on foraging and mating trips, respectively, and previous studies found common physiological and genetic regulation of the ontogeny of behavioral maturation in drones and workers (Giray & Robinson 1996; Rueppell et al. 2006). JH-analog treatments are known to affect the expression of hundreds of genes in the brains of workers (Whitfield et al. 2006), and we found that the common behavioral maturation gene set was enriched for JH-regulated genes. The common behavioral maturation set was also enriched for genes regulated by manganese and cGMP – two treatments that causally affect the age at onset of foraging in workers. Taken together, these findings suggest common transcriptional regulation of behavioral maturation in both drones and workers, which are associated at least in part due to common genetic and physiological regulation.
The common behavioral maturation set was enriched for genes shown in a worker bee experiment (Grozinger et al. 2003) to be regulated by queen mandibular pheromone. This was a surprising result because the behavioral maturation of drones does not appear to be modulated by the social environment (Giray & Robinson 1996). Perhaps this is somehow related to the fact that drones are extremely responsive to queen mandibular pheromone outside the hive, where it acts as a sex pheromone (Sandoz 2006). Another possibility is that social cues as well as physiological and genetic factors act on the same transcriptional networks in the brain to influence behavioral maturation. Indeed, there was considerable overlap between genes regulated by methoprene, manganese and cGMP when compared to genes regulated by QMP treatments (30–35%).
We were able to identify 565 genes that were uniquely associated with worker behavioral maturation. We detected two enriched annotation clusters in worker-specific behavioral maturation genes. The first contained eight genes, mostly all upregulated in older workers, associated with the regulation of Notch signaling and signal transduction, including mind bomb 1 (mib1), deltex (dx) and aristaless (al). Recent studies on Drosophila have shown the importance of Notch signaling in the process of long-term memory (Ge et al. 2004; Matsuno et al. 2009; Song et al. 2009), and Notch signaling is also involved in spatial learning and memory in mammals (Yoon & Gaiano 2005). Our results are particularly intriguing because foragers must develop a map-like spatial memory to learn and communicate the location of profitable flowers relative to the hive (Menzel et al. 2005), and the transition from nursing to foraging involves an expansion in parts of the brain implicated in learning and memory, the mushroom bodies (Withers et al. 1993). Drones, on the other hand, use physical cues to navigate between their colony and drone congregation areas (Ruttner & Ruttner 1972). Perhaps these transcriptomic differences reflect behavioral differences in navigation between workers and drones.
A second major enriched annotation cluster in worker-specific behavioral maturation genes contained 81 genes involved in transcription initiation, regulation and RNA polymerase II activity. This list contains many genes that play a role in mRNA splicing (e.g. Rm62, Dsp1, Rsf1) and gene silencing via RNA interference (e.g. spn-E, Dcr-2, Rm62). If these transcriptomic differences reflect behavioral differences between workers and drones, one possibility is that they relate to the worker bee's greater behavioral plasticity and intricate systems of division of labor, all absent in drones. Perhaps transcriptional regulation is a key molecular function associated with worker behavioral plasticity and division of labor. A recent study implicated alternative splicing as important molecular process associated with female caste differentiation in the honey bee (Lyko et al. 2010) corroborating our findings. Similarly, reconstruction of a brain transcriptional regulatory network for worker bees has shown that simple relationships between transcription factors and their putative target genes are a prominent feature of the networks underlying division of labor (Chandrasekaran et al. 2011). Genes uniquely regulated during worker behavioral maturation provide the best candidates for understanding the molecular processes associated with worker behavioral plasticity and division of labor.
We also identified an enriched cluster of genes uniquely regulated during drone behavioral maturation. This cluster was associated with GTP binding and small GTPase signaling. GTPase signaling is known to play an important role in sensory perception, and many genes in this cluster are involved in vision and sensory perception of light in Drosophila. Drones have massive compound eyes and we speculate that the upregulation of GTP binding and GTPase signaling during behavioral maturation may be functionally relevant given the challenges of locating virgin queens among hundreds of drones during mating flights.
A major finding of our study is the hitherto unprecedented levels of sex-biased gene expression in the bee brain. We found significant differences in the expression of 41% of genes in the brains of drones vs. workers, representing the strongest example of sexual dimorphism in brain gene expression found in animals to date. In comparison, using whole-transcriptome microarrays, only four genes showed sex-biased expression in the brain and ventral nerve cord of Drosophila melanogaster despite very high statistical power to detect such differences (Goldman & Arbeitman 2007). In the mouse, a recent study detected 612 out of 4508 (∼14%) genes with sex-biased brain gene expression (Yang et al. 2006). Similarly, a small number of genes (1–3% out of approximately 22 000 ESTs) showed sex-differences in brain gene expression in two song birds (Naurin et al. 2011). It is unlikely that our results reflect potential contamination of brain tissue with gland and fat cells; the honey bee brain is largely easy to separate from the surrounding tissue when freeze-dried and dissected in ultra-cold ethanol on dry ice. As an extra precaution, we excluded genes that are expressed in the sexually dimorphic hypopharyngeal glands from our analysis, the most likely source of contamination of worker brain samples.
Honey bees deviate from other studied organisms with limited sex-biased brain gene expression in several major ways. Honey bees are haplodiploid (Winston 1987), and they lack chromosomal sex-determination (Beye et al. 2003) and sex hormones (Nijhout 1994). They also show both sex and caste differences in brain morphology (Snodgrass 1956; Winston 1987). Finally, they are highly social (Winston 1987). Further studies are needed to untangle the effects of the above factors on sex-biased brain gene expression. For example, comparative studies of solitary Hymenoptera (e.g. Nasonia) can determine if sociality and caste differences in brain morphology contribute to the high level of sex-biased brain gene expression observed in the honey bee.
Sex-biased brain expression was found for several genes involved in hormone and neurotransmitter synthesis and metabolism, including Tbh and Hn. Workers in general, and older workers in particular, had higher brain gene expression levels of two genes associated with dopamine, serotonin and octopamine synthesis. These neurotransmitters play an important role in the perception of reward in workers (Barron et al. 2009; Barron et al. 2007). Both drones and workers feed inside the hive (Winston 1987), however, workers must develop the cognitive abilities to assess variable nectar and pollen sources while foraging for their colony (Shafir et al. 2005). Although both drones and workers are capable of olfactory learning (Benatar et al. 1995), workers have a higher capacity for reward evaluation and risk aversion when compared to drones (Shafir et al. 2005).
In summary, most of the transcriptional changes associated with behavioral maturation were common to both drones and workers. These genes provide a starting point to explore the mechanistic and evolutionary relationships between the highly derived system of honey bee worker division of labor and the simpler, and presumably more basal, pattern of behavioral maturation exhibited by drone honey bees. We also found several hundred genes that were uniquely regulated during the process of worker or drone behavioral maturation. These genes were enriched for processes associated with learning and memory in workers and visual perception in drones, and may either underlie or respond to caste-specific differences in behavior. Genes that are uniquely regulated during worker behavioral maturation provide the best candidates for understanding the unique aspects of division of labor and behavioral plasticity in worker honey bees.
We thank Karen Pruiett for collecting the bee samples, Tom Newman and Trang Nguyen for assistance with various molecular techniques, and Alyssa Eisenstein for assistance with microarray analysis. This is part of a series of papers arising from a National Science Foundation Frontiers in Biological Research grant (B.R. Schatz, Principal Investigator) that uses large-scale analysis to explore the influences of heredity and the environment on brain gene expression and behavior. Additional funding provided by the Institute for Genomic Biology Fellows Program (AZ).