The genome of the honey bee will inevitably be compared to that of the fruit fly Drosophila melanogaster. Inevitably, because so far we have the genomes of only two other orders - Diptera (Drosophila) and Lepidoptera (the silkworm Bombyx mori) - of the 30 or so orders of insects (the honey bee belongs to the Hymenoptera). Members of three other orders - Coleoptera (beetles), Anoplura (lice) and Heteroptera (bugs) - will soon join this group. At a coarse level, the genomes of fly and bee are quite different: that of the bee is relatively AT-rich, a fact that posed a technical problem to the sequencers, and, even more remarkably, the genes themselves are in regions that average 71% AT; in Drosophila the genes are on average 56% AT. The HBGSC suggests that this difference may be a consequence of cytosine methylation in the honey bee, as unlike Drosophila, the bee genome contains members of all three known families of cytosine-5-methyltransferase genes; indeed, it has two genes from the Dnmt1 family of genes. The presumption is that high levels of cytosine methylation, which tend to repress gene expression, have led to the preferential selection of AT-rich regions as a more favorable context for genes. If so, one might expect the bee genome to be deficient in the dinucleotide CpG; the paradox is that this genome has the highest CpG over-representation (by 1.67-fold) of any known genome. Although there is direct experimental evidence for some CpG methylation in honey bees 2 , neither its extent, nor its significance, is yet known.

Another surprise of the bee genome is its complement of transposable elements, which comprise only 1% of the sequenced genome - in contrast to 5.3% of the euchromatic genome of D. melanogaster 3 . Even more surprising is that this 1% is almost entirely made up of members of the mariner family, which transpose by simple excision and reintegration. Retrotransposable elements, a common feature of most metazoan genomes, are represented by only a small number of very degraded sequences. Whether or not this is a consequence of the haploidy of male bees, as suggested by the HBGSC, is an open question. The other group in which retrotransposable elements are known to be absent are the fully parthenogenetic bdelloid rotifers (see 4 ).

Like most Hymenoptera, honey bees have an extraordinary sex-determining mechanism known as haplo-diploidy: females are normally diploid and a product of sexual congress; males are haploid and develop parthenogenetically from unfertilized eggs 5 . The study of the genetic basis of this mechanism of sex determination in honey bees had to await the development of artificial insemination; otherwise it is impossible to do controlled crosses, a fact that, despite his efforts, defeated Gregor Mendel 6 . It was the great, but much underappreciated, geneticist P.W. Whiting who, working with a more tractable hymenopteran, Bracon hebetor, discovered this mechanism. There is a sex-determining locus with many alleles; heterozygous zygotes develop as females, hemizygous or homozygous zygotes develop as males 7 . This hypothesis was confirmed for honey bees by Woyke 8 and the complementary sex determiner (csd) gene was cloned by Beye and colleagues in 2003 9 . The product of csd is an RNA-binding protein and it may, like the Transformer protein in Drosophila, control sex by determining the splicing pattern of the doublesex gene. Population studies of the sequence of csd show that polymorphism of this gene, essential for sex determination, is maintained by balancing selection 10 .

The development of diploid honey-bee zygotes may follow one of two paths: to sterile workers who devote their lives to collecting nectar and pollen and taking care of the next generation; or to queens who, after a brief mating flight, have a life of leisure laying eggs. The genome sequence of the honey bee will provide a valuable resource for the detailed analysis of differences in gene expression between these castes. Early data from relatively small cDNA libraries already indicate major differences in intermediary metabolism between workers and queens (for example, see 11 ). The role of nutrition in determining caste development in honey bees has been known for over 200 years (see 12 ), and Wheeler et al. 13 have used the official gene list from the HBGSP 1 to implicate the insulin-signaling pathway in this developmental decision.

The rich behavioral repertoire of social bees compared to that of the Diptera has often been invoked to explain the long-established observation that the hymenopteran brain has a dramatic expansion of the mushroom body region. This paired protocerebral structure has 170,000 intrinsic neurons (called Kenyon neurons) per hemisphere in the adult honey bee 14 , compared to a mere 2,500 in Drosophila 15 . In fact, about 15% of bee neurons are dedicated to the mushroom bodies compared to only around 1% in the fly, underscoring the enhanced role of these neural structures in bee behavior. The mushroom bodies have been much studied in Drosophila, and appear particularly important for integrating sensory information, especially in the context of olfaction 16 .

Making and strengthening connections between unconditioned and conditioned stimuli during olfactory learning is a major role of the mushroom bodies in Drosophila 17 , and so it seems reasonable to assume that much of the seemingly more complicated social behavior of Apis may be mediated by this brain center. In support of this view is the observation that odorant receptors are among the gene families most over-represented in Apis compared with the fly 1 . Thus we might guess that the duplication of odorant receptor genes provided a driving force for an exponential enlargement of the brain regions that deal with the extra demands of the huge increase in potential olfactory associations. This enhanced neural plasticity may have led to the retention in Hymenoptera of genes such as Mahya, which is also found in vertebrates but has been lost from Diptera and Lepidoptera. This gene encodes a secreted protein that is expressed in the bee mushroom bodies and antennal lobes, and in vertebrates is present in the olfactory bulb, the structure that shares the same function as antennal lobes in bees, namely the processing and integration of olfactory information. These observations provide an intriguing association between the presence of this gene, its anatomical site of expression, and species with higher cognitive functions 18 .

In contrast, the gene foraging (for), which encodes a cGMP-dependent protein kinase (PKG), is found in both flies and bees and, as its name suggests, is implicated in behavioral strategies for food searching in both organisms 19 20 . In bees, for is expressed in the lamina of the optic lobes and also in a region of the mushroom bodies that receives visual information. Nurse bees age to become foragers when levels of for rise significantly in these brain regions, and these (now) foraging worker bees become positively phototactic. They then leave the darkness of the hive to become bona fide foragers 21 . In flies, however, ablation of the mushroom bodies in the larva does not affect food searching 19 , so an additional level of regulation via these structures has clearly been recruited in the honey bee, further underscoring their critical neurogenic role at the interface between genome evolution and complex social behavior.

The honey bee also misled one of us (C.P.K.) for several years about how one of the canonical circadian clock genes evolved. In 2000, it was revealed that flies and moths have two 'timeless' genes - the one first discovered and called timeless (tim), which has a cardinal role in the 24-hour clock, and tim2 (or timeout), which apparently was the only tim-like sequence found in mammals, nematodes, and other animals 22 23 . Thus it appeared that a relatively recent duplication had occurred in the ancestors of Lepidoptera and Diptera around 300 million years ago, and that tim had evolved rapidly to take on a dedicated circadian role. This view was further strengthened by the fact that mutations in tim2 in mammals or nematodes were lethal 24 25 , whereas mutating tim in Drosophila led to healthy, albeit arrhythmic, flies, revealing tim to be a dedicated 'behavioral' rather than a 'developmental' gene 26 . As the years crept by, peeking at the emerging bee genome did not reveal tim, but did reveal tim2 - the ancestral form of tim. This was consistent with a scenario of a relatively recent duplication of tim2 to generate the clock-relevant tim in the ancestors of Lepidoptera and Diptera. This cosy story has been rudely demolished, however, as the tim sequences have recently been identified in the beetle Tribolium and, even more surprisingly, in sea urchins 27 . This puts back the date for the duplication of tim to pre-Cambrian times.

The genes that we presume encode the circadian clockworks of honey bees show a number of other interesting features, apart from tim evolution, in that their genes seem to be more mouse-like than fly-like. For example, in flies and mice, the Clock (Clk) and cycle (cyc, also called Bmal1) genes encode positive transcription factors that directly regulate the negative autoregulators encoded by period and tim. In flies, the abundance of Clk mRNA cycles with a circadian rhythm but cyc is expressed constitutively, whereas in the mammal, cyc cycles and Clk does not 28 . As if to highlight this species difference, the carboxy-terminal transactivation domain found in fly Clk protein has been transposed to mouse Cyc.

Flies also have a dedicated circadian photoreceptor, encoded by the cryptochrome (cry) gene, whereas mammals have two Cry genes, which act as negative transcriptional regulators, not photoreceptors 28 . Nevertheless, the single copy of Cry in the bee encodes sequences more reminiscent of the mammalian than the fly protein, suggesting that the bee Cry protein also functions as a negative regulator, not a photoreceptor 1 . In fact, Lepidoptera have two copies of Cry; one acts as a negative regulator, the other probably acts as a photoreceptor 29 . Thus basal lineages probably had two types of Cry and two types of tim, and different organisms appear to have mixed, matched and eliminated one or other copy of these two genes according to their needs. Lepidoptera kept both types for each of their tim and cry genes, with both types of functions apparent for each gene 29 . Bees, on the other hand, have the stripped-down version, and have lost one copy of each gene, maintaining obligatory tim developmental, and non-photoreceptor Cry function 27 . Mammals kept developmental tim, but both Cry genes lost photoreceptor function 28 . Drosophila kept both tim genes, but only the photoreceptor cry 22 29 . Evolution surely plays tricks on the unwary biologist.

Most of us have, at one time or another, been stung by a honey bee. Reading the account of the venoms predicted from the genome sequence 1 makes it quite clear why these stings are so painful: bee venom contains perhaps 20 different allergens including "several homologues of scorpion and snake venoms". The domesticated European honey bee (Apis mellifera ligustica) is not, thankfully, very aggressive, but the African A. mellifera scutellata, introduced to Brazil by Warwick Kerr 40 years ago 30 , is (see Bill Hamilton's amusing account of their attack 31 ). One of the consequences of the honey bee genome project is a very dense map of single-nucleotide polymorphisms (SNPs), with nearly 5,500 SNPs already identified and mapped 1 . These have already been used to study the four major groups of subspecies of A. mellifera, with the surprising result that the Eastern (A. mellifera ligustica) and Western (A. mellifera mellifera) European populations result from independent colonizations of Europe by African populations.

Bee researchers, like their colleagues who work with Drosophila, will now distinguish the BG (Before the Genome) and AG (After the Genome) epochs. We can confidently predict that honey-bee research will now be even more vibrant and interesting than BG, with great consequences for both fundamental and applied biology.