We have used Roche 454 deep sequencing to define and compare the transcriptomes expressed by life cycle stages and castes of B. terrestris. Together, these data form a comprehensive gene catalogue for this ecologically and economically important species. The de novo assembly had high contiguity, with a mean contig length of 1,102 bases. The G+C content for the contigs was 36%, which is similar to the worker-sequencing results of Sadd et al. . The majority of contigs (approximately 58%) had significant BLAST matches to previously described proteins. The remaining 42% may derive from untranslated regions of unassembled mRNAs, noncoding RNAs or transcriptional noise (retained introns and similar), or, more interestingly, from B. terrestris genes that are either novel or have diverged significantly from any sequenced relatives. Comparison of the B. terrestris transcriptome against the emerging B. terrestris genome sequence identified matches for 95.9% of the contigs, and we expect that our data will be of utility in annotation efforts for this genome (which is currently being carried out by the Bombus sequencing consortium; K. Worley, P. Schmid-Hempel, G. Robinson, pers.comm.). Separate sequencing of individual life stages permitted the identification of potentially differentially expressed genes across the life cycle stages using R-STAT, and identified potentially important candidate genes underpinning stage- and caste-specific phenotypes. While we drew samples from two natal colonies, comparison of expression between the workers from the two colonies demonstrated significant similarity, justifying an overall comparison of pre-adult stages and adult stages.
We used R-STAT analyses to identify a list of potentially differentially expressed contigs, in order to provide direction for future studies examining gene expression within or across life cycle stages of B. terrestris. However, some data already exist in support of this approach. In a previous study by Pereboom et al. , northern blot analyses examined expression differences between whole bodied larvae and adults (queen and worker stages) and identified three genes (endocuticle structural glycoprotein; hexamerin 70b; and 60S acidic ribosomal protein P2) that were more highly expressed in larvae than in adults, where expression was absent. These genes match contigs generated from our transcriptomic R-STAT analyses that show the same larva-adult expression pattern, providing independent support for the biological validity of the results discussed below.
Differential gene expression linked to developmental processes
We identified a number of key differences in genes implicated in developmental processes across the different life stages. For example, the larval data showed high expression of genes involved in cuticle biogenesis. As the larval stage represents a period of feeding and growth, undergoing four developmental moults over a fourteen day period, this upregulation of cuticular proteins and endocuticle structural glycoproteins is expected. Other contigs identified in the transcriptome with similar expression profiles are thus likely candidates for genes with similar stage-specific roles. For example, a short-chain dehydrogenase/reductase (SDR) had elevated expression in the pupa. SDRs function in hormone and steroid metabolism , and in social insects, may be involved in stage and caste differentiation, for example through binding juvenile hormone. An SDR was demonstrated to be differentially expressed at the mRNA level in the ovaries in developing larvae of honeybee workers and queens, and SDR gene expression was higher in the ovaries of worker larvae in comparison to queen larvae resulting in possible inhibition of ovary development . The likely co-orthologues found in our study of the A. mellifera SDR analysed by Guidugli et al.  were derived solely from pupal reads, while Guidugli et al.  reported expression being highest in final stage honeybee larvae. Thus, we propose that these SDRs are candidate genes for further investigation into their roles in bumblebee caste determination.
Amino-acid storage protein expression across the bumblebee life cycle
Hexamerins are amino acid storage proteins, related to tyrosinases , used during non-feeding periods of adult development to provide amino acids and energy . Hexamerins also have roles in the binding of juvenile hormone during insect development, impacting on growth regulation within the larval stage [47–49]. A number of studies [50–53] have demonstrated differential expression of hexamerin proteins (hexamerin 70a, hexamerin 70b, hexamerin 70c and hexamerin 110) at different stages of development amongst the sexes and castes in the honeybee. In B. terrestris, the pupal sample had elevated expression of hexamerin 110-like genes and a hexamerin 70 homologue, while the larva had elevated expression of distinct hexamerin 70-like genes, and the gyne elevated expression of four additional hexamerin 70-like genes. Thus while only four hexamerin-encoding genes were identified in A. mellifera , we have identified potentially ten in B. terrestris that show differential expression between life cycle stages and castes, suggesting that these proteins may play complex roles in bumblebee development.
Why would gynes express high levels of amino acid storage proteins? Honeybee virgin queens have a higher expression of hexamerins in comparison to workers, with the hexamerins functioning in gonad development , whilst in the wasp, P. metricus, developing gynes have higher quantities of Hex 1 than workers . Studies on hexamerins in other social insects, such as ants and termites, have identified a correlation between depletion of hexamerins within queens and colony formation [55, 56]. Therefore, this potentially high expression of hexamerins would be important for a B. terrestris gyne from a colony formation viewpoint. However, as B. terrestris queens undergo a prolonged hibernation after mating, amino acid storage proteins may be important for maintaining functional operation of crucial biological processes during a period of intense stress. As hibernation is a key stage in the life cycle of bumblebees, many species of which are in drastic worldwide decline , our results provide direction for future work to analyse the mechanisms behind successful hibernation in these insects.
Genes involved in adult behaviour and physiology
Workers had elevated expression of enzymes, such as alpha-glucosidase and a muscle-specific lipase, that would be important for worker task completion. Alpha-glucosidase is involved in carbohydrate metabolism and utilised by foraging honeybees to metabolise nectar into fructose and glucose , while the muscle lipase is important for breaking down lipids during periods of high activity . Interestingly, in our study both enzymes were expressed at an early stage in the adult workers' life (only 72 hours old), in contrast to their temporal pattern of expression in A. mellifera workers. Honeybee workers demonstrate a strict temporal polyethism, where younger workers perform nursing duties while older workers forage , and it is these foraging workers that exhibit higher expression of these enzymes. In comparison, there is no strong age-dependent division of labour within bumblebees . Alloethism within bumblebee workers has been correlated with size, with studies identifying larger workers performing foraging tasks while smaller workers perform in-nest functions, although this division of labour can change depending on the requirements of the colony . Thus future work might focus on size- and age-related differences in gene expression between B. terrestris workers in relation to their subroles within the colony.
Males are underrepresented in genomic studies in social insects as the emphasis has been almost exclusively on females. In the current study, the male had elevated expression of titin, a muscle protein, expressed in the insect flight muscles . As the male bumblebee requires flight for foraging, patrolling and indeed mating, high expression of flight-specific muscles would be required. In relation to behaviour, the male had a high expression of a neuroparsin, queen brain selective protein 1, which has been suggested to function in caste determination during honeybee development through manipulation of the insect insulin-like pathway . However, neuroparsins have been suggested to play roles in a wide variety of functions, including reproduction . Therefore, a neuroparsin-like protein in the bumblebee may have male-specific expression in relation to its behaviour or physiology. The male had elevated expression of several genes that matched hypothetical or otherwise unannotated proteins in the genomes of A. mellifera, C. floridanus and S. invicta. It is particularly interesting that the male received so many fully unannotated protein matches (n = 851 contigs). This may suggest possible novel expression associates with male-specific behaviour and/or physiology. Thus, these data offer valuable insights into the mechanistic basis of male biology in social insects, which has largely been ignored by previous studies (see references above).
The immune response in males
Even in the absence of overt infectious challenge, the background level of immune defence is likely to be regulated through the bumblebee life cycle. We did not explicitly challenge the sampled bees with pathogens, but also did not keep them in germ-free environments (pre-adults in their natal colony, adults in nurseries), and thus we expect a background level of immune activation. B. terrestris queens are monandrous, mating only once , while males can mate up to eight times . In addition, B. terrestris exhibits highly male-biased sex ratios . Together, this suggests high levels of competition among the males to mate with gynes. Consequently, males should invest in reproductive fitness, which has been demonstrated in other insect species to trade-off against immunity (e.g. Anopheles gambiae ). We identified genes involved in pathogen recognition, the transduction of recognition signals, and immune effectors, and analysed their patterns of expression for data to support this hypothesis. Surprisingly, in contrast to our expectations based on life-history theory, the male had elevated expression (compared to other stages) of AMPs involved in the removal of infectious agents as part of the immune system [69–72], including hymenoptaecin, defensin, abaecin and apidaecin. This is the first account of an apidaecin-type protein being expressed in B. terrestris. Wilfert et al.  found no trade-offs between either branch of the immune system (prophenoloxidase (PPO) and AMP) and reproductive investment, but rather a positive correlation between AMPs and reproduction. Wilfert et al.  suggest the basis for the positive correlation may be because males pass on AMPs with their sperm to mates during copulation. However, the male in our study was very young and we sampled the whole body rather than just reproductive tissue, making the mating-gift hypothesis less convincing. Unlike the majority of social insect males, bumblebee males do not remain within the colony post-emergence from the pupal case. Bumblebee males forage for themselves and spend the majority of their time patrolling scent-marked trails . Bumblebee males can survive outside the colony for up to 60 days (Brown, M.J.F., unpublished data). Thus males cannot take advantage of proposed colony-level social immunity  and a primed immune system might be an adaptation to life outside the colony. Thus, our results suggest that the life-history differences between males (effectively solitary) and females (colonial) may impose divergent selection on expression of immune system genes in social insects.
Olfaction in the bumblebee
Olfaction and the ability to discriminate a number of volatiles is of immense importance to insects in general, and social insects in particular. They must recognise nest-mates, discriminate and control subordinates, select mates, and discriminate between a wide range of plants for food collection. Here we discuss two classes of olfaction genes, the odorant binding proteins (OBPs) and the chemosensory proteins (CSPs).
In the honeybee genome, 21 OBP genes have been identified and examined for patterns of expression . Within our transcriptome dataset, we found significant matches for eight A. mellifera OBPs (OBP1, OBP2, OBP3, OBP4, OBP6, OBP9, OBP11 and OBP13). This enables us to compare their expression in bumblebees to that of their homologues in honeybees. In the honeybee, OBP1, OBP4, OBP6 and OBP11 are expressed exclusively in the antennae of adults  and OBP11 was identified as having gender-specific expression (absent from honeybee drones), suggesting a role in female recognition of mates . We found low expression of a B. terrestris orthologue of A. mellifera OBP1 (formerly known as antennal specific protein 1), which is involved in binding of queen pheromone in honeybees . In honeybees, OBP1 is expressed in workers' and drones' antennae after approximately 14 days post-emergence . As we sampled our bees before this late timepoint, level of expression of the OBP1 homologue may simply be due to developmental staging. OBP6 was expressed in all the adult bumblebee stages, but not the larva and pupa, matching expression patterns in the honeybee. In contrast, OBP11 was expressed in only the male and gyne. Consequently, OBP11 in the bumblebee may have a role in mate recognition within both sexes, as opposed to the putative female-specific role in honeybees.
OBP2 had elevated expression in all adult stages of B. terrestris. In honeybees, OBP2 is expressed specifically in the antenna with weak expression in the legs and head, possibly from chemosensory sensilla in these body parts . In contrast, OBP3 is ubiquitously expressed in all adult body parts of the honeybee with the exception of the antennae . In B. terrestris, OBP3 was highly expressed in all the bumblebee adult stages, with higher expression in the gyne than in the worker, but was absent from the larva and pupa, which again matches honeybee expression patterns. OBP9 is only expressed in the ovaries and eggs of the queen in honeybees  but in B. terrestris the male was the only sample to show expression of OBP9. Lastly, OBP13 was highly expressed in the B. terrestris pupa and male, but Forêt and Maleszka  identified expression of A. mellifera OBP13 in old larvae, with expression continued throughout the pupal stage. It appears that expression of these odorant binding proteins has been conserved in some cases, whilst diverging in others, presumably in response to taxon-specific selection processes.
We identified four putative CSPs (homologues of CSP2, CSP3, CSP4 and CSP6 in A. mellifera) in the B. terrestris transcriptome. B. terrestris CSP3 and CSP4 had elevated expression in the larva, matching the results of Forêt et al.  who found A. mellifera CSP3 to exhibit highest expression in the larva before pupal and imaginal moults. While Briand et al.  proposed that CSP3 had a role in brood pheromone recognition, Forêt et al.  proposed that CSP3 may play a role in cuticle maturation. In A. mellifera, CSP4 expression was restricted to olfactory tissues in adult, but not pre-adult stages . In contrast, while we detected CSP4 expression in adult B. terrestris, the highest expression was in the larva. In honeybees, CSP2 was expressed at low levels throughout the life cycle stages while CSP6 was expressed throughout the larva, pupa and adult stages, with elevated tissue-specific expression in queen ovaries and eggs . The expression of CSP2 and CSP6 across the bumblebee life stages is consistent with that of the honeybee. We were surprised not to have detected a homologue of A. mellifera CSP1, which is expressed ubiquitously across honeybee life cycle stages. Finally, A. mellifera CSP5 is expressed only by mature queen honeybees in eggs and ovaries, stages we did not sample in B. terrestris.
Overall, the examination of olfaction-related genes in B. terrestris reveals both similarities and also striking differences from their expression in A. mellifera. We expect that these divergences reflect the different social structuring of bumblebee compared to honeybee colonies, and in particular the differing roles and strategies adopted by different castes in these two species. Again, our results provide indications of fruitful lines of future research into how patterns of gene expression relate to the evolution of primitive (bumblebees) and advanced (honeybees) sociality.