Ann. Zool. Fennici 43: 423–442 ISSN 0003-455X Helsinki 29 December 2006 © Finnish Zoological and Botanical Publishing Board 2006

Determining the molecular basis of sociality in insects: progress, prospects and potential in sociogenomics

Seirian Sumner

Institute of Zoology, Zoological Society of London, Regent’s Park, London NW1 4RY, UK

Received 1 Nov. 2005, revised version received 27 Oct. 2006, accepted 29 Aug. 2006

Sumner, S. 2006: Determining the molecular basis of sociality in insects: progress, prospects and potential in sociogenomics. — Ann. Zool. Fennici 43: 423–442.

How complex biological diversity can arise from seemingly simple strands of DNA is one of the most pertinent of questions confronting 21st century biologists. With the increasing availability and cost effectiveness of genomic techniques, there has been a rapid expansion in the taxonomic breadth of species that can now be studied at the molecular level of the genes. Consequently, behavioural ecologists are now able to examine the behaviours of their study organisms at an entirely new level. Here I review the current progress made in sociogenomics — the study of the molecular basis of sociality — with particular emphasis on what genome-level studies can reveal about social evolution. First I discuss the evolutionary interactions that occur between the genome and sociality. Next I review the current literature on how genes underlie queen and worker caste evolution: I identify 19 genes that are likely to be of importance in the evolution of caste systems across eusocial taxa; I make predictions on how gene expression patterns might orchestrate the evolution of social complexity, and make preliminary tests using the available data. Finally, I outline major questions in social evolution that can be addressed for the first time using a sociogenomic approach, high- light practical considerations in sociogenomics and discuss suitable model systems for future research on the molecular basis of sociality.

Introduction other genes in complex networks. Determining how variation in gene regulation brings about It has long been recognised that biological diver- phenotypic novelties (e.g. in terms of an individ- sity evolves at the level of the genes (Darwin ual’s morphology, physiology and/or behaviour) 1859, Dawkins 1976, 1982). Recent advances in is crucial for understanding how the diversity of molecular biology have revealed that even the life evolves (Carroll et al. 2005). most complex organisms are made up of a sur- Genes are regulated in response to the envi- prisingly small number of genes. For example, ronment. The environment of a social animal is the human genome consists of a mere 20 000– especially variable, as any individual is influ- 25 000 genes (Li et al. 2001). How does such a enced by its social interactions with other group seemingly limited genetic toolkit produce com- members in addition to the usual biotic and plex biological diversity? Clearly, a single gene abiotic influences experienced by non-social ani- must have many functions and interact with mals. This may explain why the phenotypic 424 Sumner • ANN. ZOOL. FENNICI Vol. 43 diversity displayed by the social insects (bees, to date. I review the literature in order to iden- wasps, ants and termites) is amongst the most tify common genes and expression patterns that impressive in the animal kingdom. The morpho- underpin different caste systems, and provide a logical, physiological and behavioural variation synthesis of our understanding of caste evolu- displayed between individuals within a single tion at the level of the genome. Finally, I discuss colony or amongst species within a taxonomic future directions in sociogenomics, outlining key family provides an unrivalled opportunity to questions in social evolution that can now be explore the relationship between the genome addressed, and discuss the potential of social and biological diversity at a range of differ- insects as genetic-model organisms of the future. ent unit levels (Keller 1999). Studying social insects at the molecular level across different levels of sociality provides critical insights into Social evolution and genome how complex, highly derived, social systems evolution: cause or consequence? evolved from simpler, ancestral ones. Determin- ing the genetic basis of sociality has become There may be something special about the a rapidly growing field of research, as tech- genome of social animals that predisposes them niques for studying the genome have become to become social. For example, they may have more accessible and affordable (Robinson 1999, genome ‘signatures for eusociality’, or the 2002b, Evans & Wheeler 2001). The integration ancestors of social lineages may have had a of molecular genetics with the study of sociality rich complement of genes coding for recogni- — ‘sociogenomics’ sensu Robinson (Robinson tion or communication receptors which facili- 1999) — therefore leads the classical study of tated social evolution. Conversely, being social social behaviour to new levels, making it one might affect the way the genome evolves. For of the most promising fields for studying how example, there may have been major genome novel gene functions evolve, as well as revealing rearrangements through chromosome breakage how variation in transcript abundance of certain and exchange of genes during the transition from genes reflects behaviour and influences social non-social to social. Disentangling the causes evolution (Fitzpatrick et al. 2004). Questions of eusocial evolution from those that arise as raised include whether the same genes underlie a consequence of it is important if we are to similar behaviours in different taxa, and identify- develop realistic models on the molecular basis ing genes that have been crucial in influencing of social evolution, as well as understand how social evolution, or which have been completely the (social) environment influences genome evo- lost or gained. These gene-level insights into lution. A major obstacle in disentangling cause sociality enable us to build a more realistic pic- from consequence is the feedback between the ture of how the genome orchestrates diversity, in genome and sociality. A likely scenario is out- the form of sociality. lined in Fig. 1: Haplodiploidy, a property of most There are several excellent reviews of social insect (all Hymenoptera) genomes, can sociogenomics, which in particular describe the predispose individuals to altruistic behaviour by ground-breaking progress that is being made in altering the social structure of a group such that understanding how the genomic properties of some individuals help others reproduce at the the honeybee (Apis mellifera) underlie its social- expense of their own direct reproduction. In this ity (Evans & Wheeler 2001, Robinson 2002a, way, the genome has the potential to influence 2002b, Robinson et al. 2005). In this paper, I social evolution. The resulting change in social explore the realms of sociogenomics beyond the structure (i.e. reduction in the ratio of breeders honeybee. I draw together current data on social to non-breeders) means that the effective popula- insect genomes in order to discuss the complex tion size is reduced (Crozier 1979), which might interaction between sociality and genome evolu- alter the rate of molecular evolution (Bromham tion. In particular, I focus on the genetic basis & Leys 2005). Any resulting gene losses, gains of queen and worker castes, the topic in sociog- and duplications, nucleotide substitutions and enomics that has received the most attention deletions could contribute to the evolution of ANN. ZOOL. FENNICI Vol. 43 • Sociogenomics and the evolution of sociality 425

6. Novel gene function

2. Genes for 1. Haplodiploidy altruistic behaviour

SOCIALITY GENOME

Fig. 1. A hypothetical exam- ple of how the genome may influence the evolution of sociality and sociality influ- ence genome evolution. 3. Reduced effective 5. Loss/gain of Examples of feedback are population size genes numbered 1–6 in the order 4. High rate of in which they are likely to molecular evolution take effect. See text for fur- ther details. novel genes or gene functions, which in turn may 1971). However, haplodiploidy is not a prereq- alter the behaviour of individuals and their social uisite for eusocial evolution. Neither is it likely structure. In sum, sociality has the potential to to have evolved as a consequence of eusociality. act as a vehicle for genome evolution through Evidence for this is found amongst the eusocial a constant feedback: changes at the gene level taxa that are diplodiploid (e.g. naked mole rats, alter behaviour which alters how natural selec- termites and aphids, shrimps and an ambrosia tion acts on the genes, and so on. Determin- beetle), the many large taxonomic groups that ing how sociality evolves ultimately demands a are haplodiploid but lack any eusociality (e.g. better understanding of this feedback. some mites, beetles, white fly and scale insects The genomes of many social insects exhibit (Normark 2003)). unique properties that appear to be associ- Cause and consequence cannot be distin- ated with sociality. For example the haplodip- guished so easily for other genome proper- loid genetic system, found in all Hymenoptera ties found in social insects. For example, there (bees, wasps and ants), can provide the basis is some evidence that social insects tend to for eusocial evolution by providing relatedness have higher chromosome numbers than their incentives for sib-rearing. Females are diploid, non-social counterparts (Sherman 1979, Cro- developing from fertilised eggs, but males are zier 1987). Also, worker specialisation in ants haploid, developing from unfertilised eggs. The roughly correlates positively with chromo- asymmetry in a female’s relatedness to her sis- some number, and primitively eusocial wasps ters (r = 0.75), brothers (r = 0.25) and own appear to have lower chromosome numbers than offspring (r = 0.5) has the potential to predis- advanced eusocial wasps, suggesting that chro- pose haplodiploid females to raise sisters instead mosome number most likely became elevated of their own offspring (Hamilton 1964, Trivers in response to sociality, rather than the other & Hare 1976). Under certain conditions (e.g. way round (Sherman 1979). High chromosome single-mating and female-biased sex ratios) this number is likely to reduce the variance in sib- can form the basis for a reproductive division of sib relatedness, but it is not clear how this labour between queens and workers as well as might benefit sociality. Sherman suggested that overlapping mother–daughter generations, both because workers would be less able to determine of which are conditions for eusociality (Wilson the relative fractions of genes carried by differ- 426 Sumner • ANN. ZOOL. FENNICI Vol. 43 ent siblings, they would raise siblings unselec- ally significant relationship. However, advanced tively, which would be the queen’s preference. eusocial species did generally have a higher However, this requires that workers can identify nucleotide substitution rate than non-social spe- with which siblings they share the most alleles, cies, suggesting that over long evolutionary for which there is little evidence (Queller et al. periods complex sociality may increase overall 1990, Keller 1997). Moreover, any change in rates of molecular evolution. Their analysis was selection pressure that might arise as a result of based on 1–3 genes per social/non-social com- a mutant queen with high chromosome number parison, and so it is possible that more sequence would take many generations to exert an effect data might reveal further relationships between (Dawkins 1982). High chromosome number, that sociality and molecular evolution. for whatever reason evolved, may instead be a Whether a high rate of molecular evolution pre-adaptation to eusociality (Dawkins 1982). is a cause or consequence of sociality, it will This conclusion would also account for the many certainly give rise to the evolution of new genes non-social species which have high chromo- or new gene functions through gene duplica- some numbers, and the eusocial species which tion, single nucleotide substitutions, insertions have a very low chromosome number (e.g. the or deletions. There is already evidence for this primitive ant genus Mymecia has only 1 pair as some conserved genes have novel functions in of chromosomes (Crosland & Crozier 1986)). social animals (Robinson & Ben-Shahar 2002). The interplay in the evolution of chromosome For example, major royal jelly protein (MRJP) number and sociality evidently requires more is used by honeybee workers in the production rigorous testing of the data with phylogenetic of royal jelly (Whitfield et al. 2003), whilst in correction and a more considered explanation of other social insects it is associated with queen- the mechanism. specific functions (Tian et al. 2004, Sumner et There has been some discussion as to whether al. 2006). Moreover, Krieger and Ross (2002) high rates of molecular evolution may have recently identified for the first time a gene that driven social evolution. The honeybee (Apis determines a complex social behaviour, illus- mellifera) has a very high recombination rate trating that simple nucleotide mutations can and exhibits great colony-level genetic diver- strongly influence sociality. They showed that sity (Hunt & Page 1995), whilst more primi- allelic differences in the gene Gp-9 in the fire ant tively-eusocial species (e.g. Bombus terrestris) Solenopsis invicta determine whether a colony and solitary species (e.g. Nasonia vitripennis) will be single- or multi-queened. Workers that have much lower recombination rates. Although are heterozygous at the locus (Bb) will tolerate limited, these data suggest that social complex- multiple queens (polygyny) if queens carry the ity and recombination rates are positively cor- b allele, but will kill BB queens. Thus, the social related (Gadau et al. 2000). High recombina- structure of a colony depends on the presence tion rates may produce greater genotypic, and or absence of the b allele. The Gp-9 sequence is hence probably phenotypic, diversity amongst similar to pheromone-binding proteins in moths, workers through multi-gene traits where vari- and is used in chemical recognition by S. invicta, able loci are linked on chromosomes. A related prompting the authors to suggest that it is used issue is whether the small effective population as a phenotypic signature, or “green-beard” by sizes that arise as a result of some individuals workers to distinguish between queens. Interest- becoming workers affects the rate of molecular ingly, the two alleles differ only at 9 nucleotides, evolution in social animals. Bromham and Leys which code for 8 amino acids (1 substitution is (2005) conducted a phylogenetic analysis to see synonymous, causing no change in amino acid if there is a link between sociality and the rate production). Moreover, the authors were able to of molecular evolution across the Hymenoptera, identify 3 amino acids shared uniquely by the b- termites, shrimps and naked mole rats, by com- like allele, of which one or more is likely to play paring related non-social and social species. In an essential role in producing the gene product contrast to an earlier study that used fewer taxa that induces polygyny. It is unclear whether the (Schmitz & Moritz 1998), they found no gener- role of Gp-9 in controlling polygyny is con- ANN. ZOOL. FENNICI Vol. 43 • Sociogenomics and the evolution of sociality 427 served across taxa beyond the genus Solenopsis, The genetic basis of queen and as the authors were unable to amplify the gene worker castes in a related myrmicine genus, Monomorium. This suggests that the role of Gp-9 as a social One of the major drivers in the evolution of regulator evolved as a consequence of sociality social complexity is task partitioning between in Solenopsis invicta rather than being a cause, group members. In social insects it is taken to although more extensive comparative studies are extremes in the form of reproductive (queen) required to confirm this. The most probable and non-reproductive (worker, soldier) castes. scenario is that the rate of molecular evolution Caste specialisation in social insects correlates arises as a consequence of sociality as well as strongly with ecological success (Bourke 1999). being a cause. Thus, studying the genetic basis of castes in That gene regulation and function can be extant social insects provides insights into the influenced by sociality, strongly suggests that molecular machinery central to eusociality. More unique genes (or gene networks) that are absent broadly, such studies can also reveal how a single in non-social animals are likely to evolve. A genome can give rise to phenotypic diversity, candidate for this might be a gene(s) for altru- what and how novel genes and gene functions ism (Wade 1980, Michod 1982, Queller 1992). evolve, and what the relationship is between In order for an altruistic gene to spread it must the transcriptome and behavioural, physiological be present in the beneficiary (i.e. the queen) and morphological diversity. as well as the altruist (i.e. the sterile worker). Either it must be facultatively expressed (e.g. some adults help, and others reproduce) or it Genes underpinning queen and worker must be obligately expressed at different life castes stages (e.g. young adults help and old ones reproduce) (Charlesworth 1978, Crozier 1979, Queens and workers are different phenotypes Bourke & Franks 1995). If the altruism gene is arising from the same genome. With few excep- newly evolved as a product of sociality it will tions (Julian et al. 2002, Volny & Gordon 2002, be absent in the solitary relatives of the social Helms-Cahan & Keller 2003), whether an indi- lineage. Linksvayer and Wade (2005) elaborate vidual becomes a queen or worker depends on on this by explaining eusocial evolution in terms how it responds to environmental stimuli at criti- of maternal carers (queens) and sib-social carers cal periods in caste determination, rather than (workers) whose behaviours are derived via dif- genotypic differences (Wilson 1971, Wheeler ferential expression of ancestral maternal care 1986). Phenotypic variation exhibited by castes genes. The authors make predictions which can therefore usually arises through differential be — and to some extent have already been (see expression of shared genes. Insights into the below) — tested using sociogenomic techniques, evolutionary origins of caste are best obtained namely that queen (maternal care) and worker from studying simple societies, which exhibit (sib-social care) behaviours arise through het- the most primitive caste systems (where queens erochrony (temporal expression differences) of and workers differ only in behaviour) and are ancestral maternal-care genes. There is already likely to represent the ancestral state of advanced good evidence that queen and worker castes from castes (where queens and workers may differ a range of social insect societies differentially in morphology) found in more complex insect express shared genes, both at the developmental societies. The over-arching question is whether and adult stages. Studies on the molecular basis the genes underlying behavioural differences in of castes in social insects have figured promi- castes of simple societies (i.e. those associated nently in sociogenomics to date, providing the with the origins of caste, from an ancestral, best picture yet of social evolution at the level of solitary state) are the same as those underlying the genome. I review this blossoming research morphological differences in castes of complex topic in the next section. societies (i.e. those associated with the mainte- nance of castes, in the derived state) (Evans & 428 Sumner • ANN. ZOOL. FENNICI Vol. 43

Wheeler 1999). Thus, comparisons of behav- expression profiles of 50 predictive genes to ioural castes in simple societies with morpholog- accurately determine whether a bee was a nurse ical castes in more complex societies can reveal or forager. These studies illustrate very elegantly the identity of some of the molecular drivers of how complex behaviours that are coordinated caste evolution. It is a fortunate coincidence that by multiple genes in undefined networks can be the studies in which gene expression differences studied in a genetic non-model organism. between castes have been studied cover a siz- Aside from the honeybee, queen and worker able complement of social complexity, ranging castes have been studied at the gene level in from simple behavioural castes in paper wasps several other species that exhibit morphological (Polistes canadensis) to morphological castes in caste differences. Genes that are differentially honeybees (Apis mellifera), fire ants (Solenopsis expressed amongst castes in adults and brood invicta) and termites. Moreover, the studied spe- of the bumblebee Bombus terrestris have been cies represent amongst them at least 6 independ- isolated (Pereboom et al. 2005), revealing some ent origins of eusociality, in each of which queen intriguing patterns of expression. Pereboom et and worker castes have evolved independently al. (2005) found that larval caste development in (Wilson 1971). Here I summarize the findings the bumblebee is associated with temporal vari- of these studies and identify genes that appear ation in gene expression, such that larvae which to be ubiquitously differentially expressed with up-regulate genes early in development become respect to caste across taxa. queens, but those that up-regulate these same genes late in development become workers. Tem- poral regulation of gene expression is increas- Differential gene expression in castes of ingly recognised as a way in which a seemingly social insects small number of genes can produce such bio- logical complexity (Arnosti 2003). Pereboom et Genome-level research in social insects is cur- al. (2005) made the first comparisons between rently best established in the honeybee Apis expression patterns of brood and adults within a mellifera, and with a fully sequenced genome species. They found that different genes appear near completion in this species it will be a pow- to underlie caste differences in brood and adults. erful resource for comparative genome research This study also facilitated the first comparisons in social insects (Robinson et al. 2005). Evans of the genetic basis of castes across species. Four and Wheeler (1999, 2000) paved the way for genes were identified as differentially expressed honeybee genomics by identifying differentially in both B. terrestris and A. mellifera, but only one expressed genes in queen and worker larvae. of them (hexamerin II) shared similar expression These authors were the first to empirically patterns between the two species. Five differen- illustrate that differential expression of many tially expressed genes have also been identified genes underlie queen and worker polyphenisms amongst newly-emerged adult queens and work- in social insects. Subsequent development of a ers of the stingless bee Melipona quadrifasciata, microarray of the honeybee brain has facilitated and one of these (cytochrome P450) was also dif- large-scale transcriptome screening of adult hon- ferentially expressed in A. mellifera (Judice et al. eybees, and some inspiring studies on the genes 2004). The pioneering of transcriptome studies underlying adult worker behaviours in honey- on species other than A. mellifera opens avenues bees have recently been published. For example, for conducting comparative genomic analyses on honeybee workers change their behaviour from many other social insects. nursing to foraging with increasing age. Kuchar- Ants exhibit more derived caste systems than ski and Maleszka (2002a) illustrated the spatial do bees and wasps, with extreme caste polymor- and temporal changes in gene expression that phism amongst workers and in some species the take place during this behavioural development evolution of a soldier caste (Hölldobler & Wilson by comparing expression in the brains and abdo- 1990). To date, there has only been one study on mens of naïve worker bees, nurses and foragers. gene-level caste differences in ants. Tian et al. More recently, Whitfield et al. (2003) used the (2004) isolated differentially expressed genes in ANN. ZOOL. FENNICI Vol. 43 • Sociogenomics and the evolution of sociality 429 dealate queens (mated 24 hrs prior to collection, Termites are phylogenetically distant from who have shed their wings) and alate queens the hymenopterans, being more closely related to (virgin queens, yet to mate and shed their wings) cockroaches and mantids. All termites are euso- of the imported fire ant Solenopsis invicta. They cial and represent an origin of caste evolution identified 7 genes that are up-regulated in dealate that is independent of the Hymenoptera (Thorne queens, suggesting these genes are important in 1997). Comparisons of termite caste systems queen maturation. Tian et al. (2004) also pro- with hymenopteran caste systems are compli- duced the most comprehensive data so far on cated for several reasons. Firstly, termites are how expression levels of single genes change hemimetabolous and so immature stages have in the different brood and adult developmental a similar body form to adults such that a single stages, which includes eggs, larvae, pupae, adult individual can pass through worker (pseudogate) workers, alate and dealate queens, although only and reproductive or soldier castes within its life- 9 genes were examined. Because there are no time. Since all termites are eusocial there is no solitary or even primitively eusocial species of intermediate, primitively eusocial termite and so ants, studies on ants can only reveal how castes a direct comparison of caste evolution with the are maintained rather than how they evolved Hymenoptera cannot be made. However, because (Crozier 1982). Nonetheless, these studies are an individual termite can change caste over important since the highly complex societies (and time, there are elements of caste development hence extreme caste differentiation) observed in in termites that may be considered analogous ants have not been achieved by wasps or bees. to the temporal, behavioural castes observed in To date, there has only been one study on the extant primitively eusocial Hymenoptera (e.g. genes underlying queen and worker castes in a Polistes as discussed above). Another problem truly primitively eusocial insect, where queen and is that sterile castes (workers and soldiers) in ter- worker behavioural castes are determined during mites are composed of both males and females, adulthood. Genes associated with adult devel- whilst in the Hymenoptera the sterile castes opment in the paper wasp Polistes canadensis are always female. However, sex-specific caste show gradual changes in expression with social specialisations are restricted to the most derived status, such that queens (high-ranked females) termite groups, for which there are currently no and newly emerged (low-ranked) females exhibit gene expression data. Despite these complica- distinct patterns of gene expression, with work- tions, gene-level comparisons made with queen ers (mid-ranked females) exhibiting intermedi- and worker pathways in the social Hymenoptera ate patterns (Sumner et al. 2006). Sumner et may reveal important insights into convergent al. (2006) looked for similarities amongst the evolution of caste systems amongst the social genes and patterns of expression displayed by insects (Miura 2004). Much of the gene expres- adult castes in the 4 species mentioned above sion research to date has focussed on the lower (P. canadensis, B. terrestris, A. mellifera and termite Reticulitermes flavipes, where workers, S. invicta) and identified 9 genes that appear to soldiers, alates (winged reproductives) and sup- be differentially expressed with respect to caste plementary reproductives (wingless) have been across 2 or more of these species. This com- examined (Scharf et al. 2003, 2004). Genes parison suggested that although the same genes that are exclusively expressed in the mandibular may be differentially expressed with respect to glands of soldiers of another species, Hodoter- caste across taxa, the patterns of expression mopsis japonica, have also been isolated (Miura appear to be widely divergent. Research is to be et al. 1999, 2003, Miura 2001, Hojo et al. 2005). strongly encouraged on species with behavioural Comparisons of these data with soldier castes caste systems, especially through comparisons in hymenopteran species (e.g. Atta leafcutting with extant solitary relatives, which are likely ants) may prove insightful when such data is to represent the ancestral non-social state. Such available, since soldiers in the Hymenoptera are comparisons may be crucial in revealing the modified workers whilst soldiers in the termites genome changes that accompany the first steps are thought to have evolved independently of in eusocial evolution. workers. 430 Sumner • ANN. ZOOL. FENNICI Vol. 43

Identifying genes associated with caste et al. 2004) and the 3 termites n = 17 (Miura et across eusocial taxa al. 1999, Scharf et al. 2003, 2004, Hojo et al. 2005). These sequences were aligned into over- Here, I compare the genes and expression pat- lapping regions (contigs) using the sequence terns that have been identified as differentially viewing program Sequencher, using either large expressed with respect to caste across 8 species gap alignment or ‘dirty’ realignment with 60%– of social insect. These are 3 bees (Bombus ter- 90% minimum match and 70%–90% minimum restris, Apis mellifera and Melipona quadri- overlap. Contig sequences that had not scored a fasciata), 1 wasp (Polistes canadensis), 1 ant similarity match on Genbank at the time of pub- (Solenopsis invicta) and 3 termites (Reticuli- lication were subjected to BLASTX in order to termes flavipes, Nasutitermes takasagoensis and verify homology and infer gene function. Hodotermopsis japonica). To my knowledge, I identified 19 ESTs that are differentially these are the only species in which castes have expressed with respect to caste in 2 or more been examined at the level of the transcriptome. species (listed in Appendix 1) and hence may Unfortunately, the current diversity of species have an evolutionary important role in caste studied is too limited and the number of genes systems in social insects. Although differentially studied too small to determine whether the genes expressed ESTs from M. quadrifasciata had underlying behavioural and morphological castes similarity matches with A. mellifera sequences are the same (see ‘Genes underpinning queen (Judice et al. 2004), none of these are currently and worker castes’). Thus, my purpose here is known to be differentially expressed with respect to simply identify any genes that have been con- to caste in A. mellifera, which explains the served across developmental stages and across absence of M. quadrifasciata in Appendix 1. No evolutionary time with respect to caste. Accord- single EST was consistently expressed by the ingly, I (1) identify genes that underlie caste dif- same caste in all species for which there are data, ferences in larvae and adults of any one species; suggesting that the specific roles of these genes (2) identify any genes that underlie caste differ- have changed over evolutionary time, in terms of ences in any one life-stage across different spe- either their specific function or how they interact cies. Finally, I (3) compare expression patterns with other genes or gene networks. of specific genes within these groups and discuss I used these data to determine whether there any functional divergence of caste-associated are any genes that underlie caste differences in genes that has occurred over evolutionary time. both larvae and adults within species. Although In order to identify genes that are ubiqui- some genes are unlikely to be expressed by tously expressed across taxa with respect to brood and adults (e.g. those involved in egg caste, I looked for similarities in the published production), genes that are expressed by both sequences (available on Genbank) rather than life-stages may have important roles in caste used the gene category given to each expressed regulation. Pereboom et al. (2005) found in the sequence tags (EST — sub-section of a gene) at bumblebee that 8 out of 12 genes were differ- the time of the publication. This was necessary entially expressed with respect to castes in both because many of the similarity matches recorded brood and adults, and concluded that the same at the time of publication are now out of date genes do not always underlie caste differences as new sequence data are constantly submitted between adults and larvae. In addition to this to Genbank. I looked for similarities amongst study, I found comparable data for the honeybee cDNA ESTs available for A. mellifera n = 156 and fire ant, and was able to identify 7 genes that ESTs (Corona et al. 1999, Evans & Wheeler were differentially expressed in larvae and adults 2000, Toma et al. 2000, Ben-Shahar et al. 2002, of one or more of these three species (see Table Kucharski & Maleszka 2002a, 2002b, Piulachs 1). Expression patterns in adults and either early et al. 2003, Whitfield et al. 2003), B. terrestris or late instar larvae (but not in both instars) were n = 12 (Pereboom et al. 2005), M. quadrifasciata the same for 4 genes (Table 1, column 1), e.g. n = 16 (Judice et al. 2004), P. canadensis n = Cytochrome oxidase I was upregulated in both 43 (Sumner et al. 2006), S. invicta n = 10 (Tian adult workers and late-instar worker-destined ANN. ZOOL. FENNICI Vol. 43 • Sociogenomics and the evolution of sociality 431 larvae, but not early-instar larvae of B. terrestris. to caste (Table 2). Three genes have similar These 4 genes may be important in regulating the patterns of expression across species: peroxire- switch from one caste to another. Five genes had doxin, ribosomal protein and hexamerin II are different expression patterns in larvae and adults all up-regulated in late-instar worker-destined of any one species, and 2 of these were expressed larvae relative to queen-destined larvae (Table only in larvae (cuticle protein and hexamerin II). 2, column 1). The other 3 genes exhibit different The remaining 3 genes (cytochrome oxidase I expression patterns across species: cytochrome & II and ATP synthase) were all up-regulated oxidase I, ATP synthase and cuticle protein are by one caste in larvae and another caste in all up-regulated in late-instar worker-destined adults (Table 1, column 2). In conclusion, some larvae of the bumblebee and late-instar queen- genes are differentially expressed with respect destined larvae of the honeybee. Amongst adults to caste in both brood and adults, but many (which includes all species listed in Appendix 1), genes appear to have different roles in the differ- five genes appear to be consistently up-regulated ent life-stages. Direct comparisons of expression in workers (Table 3, column 1). This suggests patterns between queens and workers at brood that worker behaviours in even distantly related and adult stages for more genes, and studies on species (such as the paper wasp and honeybee) the specific functions of genes at different life- are underpinned not only by some of the same stages are evidently required. (or closely related) genes, but also by similar My second aim was to identify any genes expression patterns. These genes may be particu- underlying caste differences across species. larly fruitful to pursue in future studies on adult These genes may shed light on the molecular castes in social insects. A further 6 genes clearly processes involved in caste evolution as well as do not share expression patterns across species reveal mechanisms of gene evolution. I exam- (Table 3, column 2). These genes may still be ined larvae and adults separately, since the rel- of evolutionary importance in the caste systems evance of examining the same genes at different life-stages is unclear (see previous paragraph). Table 2. Genes that are differentially expressed with In the larvae of the honeybee and bumblebee respect to caste in larvae of both B. terrestris and A. (fire ant is excluded as there are too few data) 6 mellifera. Genes listed in column 1 have similar pat- genes were differentially expressed with respect terns of expression, and those in column 2 differ in their patterns of expression.

Table 1. Genes that are differentially expressed with Same expression Different expression respect to caste in both adults and larvae of B. terres- patterns in larvae patterns in larvae tris (Bombus), A. mellifera (Apis) or S. invicta. Genes across species across species are listed as those that have similar patterns of expres- sion in adults and larvae (column 1: i.e. up-regulated in Hexamerin II Cytochrome oxidase I the same caste in both life-stages) and those that differ Peroxiredoxin ATP synthase in their patterns of expression (column 2: i.e. up-regu- Ribosomal protein Cuticle protein lated in different castes at different life-stages).

Same expression Different expression Table 3. Genes that have similar expression profiles patterns in adults patterns in adults (column 1) and mixed expression profiles (column 2) and larvae and larvae amongst adults of different species.

Cytochrome oxidase I Cytochrome oxidase I Genes up-regulated in Genes varying in (Bombus) (Apis) adult workers of expression patterns across ATP synthase ATP synthase different species adults of different species (Bombus) (S. invicta, Apis) Peroxiredoxin Cytochrome oxidase II Cytochrome oxidase I Peroxiredoxin (Bombus) (S. invicta) ATP synthase Ribosomal protein Ribosomal protein Cuticle protein Heat shock protein IDGF (Bombus) (Bombus) Lectin like Transferrin Hexamerin II SPARC Vitellogenin (Bombus) MRJP 432 Sumner • ANN. ZOOL. FENNICI Vol. 43 of social insects, but their functions are likely to the production of royal jelly and this suggests have diverged. that vitellogenin may have helped direct the evo- It is clear now that some of the same genes lution of age polyethism in the honeybee, since are associated with caste determination across young nurse bees up-regulate it relative to older, several social insect species. There is currently forager bees (Amdam et al. 2003a, Piulachs et little overlap between the data available for both al. 2003). termites (Isoptera) and Hymenoptera (only 6 Major royal jelly proteins (MRJP) are a genes) and so it is difficult to determine to what family of proteins with multiple functions and extent the same genes underlie castes in these are of particular interest because of the novel two distantly related lineages. However, there is functions they seem to have evolved in some evidence of convergent evolution in the genetic social insects. In some species they are evidently basis of caste systems in both the Hymenoptera associated with active queen behaviour: they are and Isoptera, since 4 genes (18S ribosomal pro- expressed by queens but not workers in the paper tein, cytochrome oxidase 1, Hexamerin I and 28S wasp (Sumner et al. 2006), and a related protein ribosomal protein) are differentially expressed (yellow protein, from the same family as MRJP) with respect to caste in both termites and at least is expressed by dealate but not alate queens of one other hymenopteran species (Appendix 1). the fire ant, S. invicta (Tian et al. 2004). How- ever, in the honeybee MRJPs are up-regulated by foragers (Kucharski & Maleszka 2002a) where Functional divergence in genes associated they play a fundamental role in worker pro- with queen and worker castes duction of royal jelly (Kucharski & Maleszka 1998). MRJPs also appear to play an important In the previous section, I identified genes that are role in the honeybee brain where 8 different likely to have evolutionary and/or mechanisti- proteins have already been identified, indicating cally important roles in some of the caste sys- high rates of gene duplication and diversification tems displayed in social insects. More research is (Albert & Klaudiny 2004). These functions of required in order to understand the functions of MRJPs are evidently divergent and complex in these genes in the different castes, life-stages and some socially advanced species of social insect. species, as patterns of expression with respect to It remains to be seen how widespread divergent caste are clearly poorly understood at present. evolution in this gene family has been in other In this section I review what we know about eusocial lineages. the putative functions of a few of these genes Another gene that has a clearly complex in order to illustrate the point that many of functional role in caste regulation is transfer- these genes have multiple functions within spe- rin, an iron-binding protein that is thought to be cies and divergent functions amongst species or involved in innate immunity and vitellogenesis life stages. Our understanding of gene regula- in insects. In the honeybee it is upregulated in the tion and multi-functionality is limited and this brains of mature adult foragers and in the abdo- may explain why the genetic patterns underlying mens of virgin queens (Kucharski & Maleszka caste evolution (reviewed in ‘Identifying genes 2003). Such temporal and context-dependent associated with caste across eusocial taxa’) are fluctuations in expression of transferrin indicate currently difficult to interpret. that it has a dual function in the honeybee. In Vitellogenin is a gene that is primarily asso- contrast, its role in the paper wasp appears to be ciated with egg production and as expected is more simple, where it is up-regulated strongly in up-regulated in queens relative to workers in queens, with very weak expression in workers honeybees (Piulachs et al. 2003), paper wasps (Sumner et al. 2006). The multi-functionality (Giray et al. 2005, Sumner et al. 2006) and of transferrin is well documented in vertebrates termites (Scharf et al. 2004). Recent studies, (Espinosa-Jeffrey et al. 2002), and so its com- however, have suggested that its role in the more plexity in insects is of little surprise. advanced eusocial insects may have diverged. The genes discussed here represent a small For example, in honeybees it is also involved in fraction of the genes actually involved in the ANN. ZOOL. FENNICI Vol. 43 • Sociogenomics and the evolution of sociality 433 evolution and maintenance of castes in social phological castes are determined during larval insects. The emerging pattern to date is that development) are likely to have evolved from many genes have been co-opted into regulating simple societies (in which behavioural castes caste systems — both across different life-stages are determined during adulthood). Understand- within a species, and also amongst species sepa- ing how queen and worker castes evolve from a rated by varying phylogenetic distances — but solitary state and how subsequent caste polymor- that gene function and regulation are clearly not phisms evolve are fundamental questions in the ubiquitously conserved. It is likely that a key study of social evolution. It has been suggested mechanism for social plasticity in many social that as castes evolve, the reproductive and non- animals, invertebrates and vertebrates, is spatial reproductive parts of the solitary phenotype are and temporal variation in gene expression. The decoupled, such that the two alternatives (queens design of experiments in future studies on the and workers) are mutually dependent, comple- genes involved in sociality should bear this in mentary morphs (West-Eberhard 1996, Gadagkar mind. Moreover, it is important to note that most 1997, Giray et al. 2005, Hunt & Amdam 2005). studies are biased towards finding convergence, Thus, in addition to changes at the level of indi- since researchers tend to choose to study ESTs vidual genes (discussed in ‘Genes underpinning that have similarity matches with known protein queen and worker castes’), the evolution of queen sequences on Genbank, or candidate genes that and worker castes from a solitary state may have have been studied in other animals. This means a involved redirecting transcriptome regulation into great many ESTs that we know nothing about are caste-specific roles. This could take place through omitted because they fall below the similarity suppression or induction of particular genetic threshold and remain unclassified. One such gene pathways in one or both castes such that they is listed in Appendix 1 (Unknown 1). Unknown express different, complementary sets of genes, genes are of particular interest as they may have the null hypothesis being that queens and workers evolved specifically in response to sociality and are equally divergent. It has also been suggested are likely to perform unique functions. that there are likely to be more genes with caste- specific expression in the complex (advanced) eusocial species than in the simple (primitively) A model of caste evolution at the eusocial species (West-Eberhard 1996, Gadagkar transcriptome level 1997, Linksvayer & Wade 2005). Thus, transcrip- tome patterns in queens and workers are pre- A single-gene level approach to studying the dicted to diverge such that more genes become molecular basis of sociality, however, may caste-specialised with increasing social complex- obscure any large-scale patterns of differential ity (see Fig. 2: Predictions). expression. Here I make predictions about the large-scale genomic changes that are likely to accompany the processes involved in caste evo- Observations lution and discuss whether the gene expression data currently available fit these predictions. These predictions can be tested by comparing the large-scale expression patterns of genes asso- ciated with queen and worker castes in simple Predictions societies with those in more complex societies. One way of doing this is to compare the relative Extant eusocial species likely evolved from soli- frequency of genes that are equally expressed tary species in which all females would have in queens and workers of species with differ- carried out both queen tasks (i.e. egg-laying) ent caste systems. This is a crude assessment and worker tasks (i.e. foraging, nest building and and assumes that genes have equally important brood rearing) (West-Eberhard 1987, Amdam et effects on the developing phenotype, but it is al. 2004, Hunt & Amdam 2005). Subsequent the only feasible way to address these predic- evolution of complex societies (in which mor- tions with the currently available data. I exam- 434 Sumner • ANN. ZOOL. FENNICI Vol. 43

PREDICTIONS OBSERVATIONS

Solitary spp. Solitary ancestor 100% Predicted

Simple society: P. canadensis Behavioural castes 19% 39% 42% n = 36 genes

Complex society A. mellifera Morphological castes 43% 19% 47% n = 156 genes

Fig. 2. Predictions of how the transcriptome changes with increasing social complexity. Observations of transcrip- tome differences between simple and complex societies. Patterns are based on the limited available gene expres- sion data for simple societies of P. canadensis (adults) and more complex societies of A. mellifera (brood). Circles with dotted lines represent worker transcriptome, and those with solid lines represent the queen transcriptome. Overlap of the two circles indicates the proportion of genes that are equally expressed (less than 2-fold differences in observed expression); non-overlapping indicates the proportion of genes up-regulated in one caste. ined published data on expression patterns in When queen and worker-destined larvae of adult queens and workers of P. canadensis (from A. mellifera are examined in the same way it is Sumner et al. 2006) and queen- and worker-des- clear that they differ strikingly from P. canaden- tined larvae of A. mellifera (from supplementary sis (Fig. 2, Observations). First, there is no sig- data in Evans and Wheeler (2000)), which repre- nificant difference between the number of genes sent two extremes of sociality. I found some pre- up-regulated in queens and workers of A. mellif- liminary support for these predictions. I regarded era (74 and 68 genes respectively, Fisher’s exact: genes as being differentially expressed between p = 0.55). Second, the difference in the degree of queens and workers if there was at least a 2-fold decoupling in A. mellifera and P. canadensis difference in expression level. In the behavioural is highly significant: queens and workers in castes of P. canadensis, a large proportion of the A. mellifera expressed only 10% (16/158) of genes studied were equally expressed (less than the sampled genes equally, whilst P. canadensis 2-fold difference in expression level) by workers equally expressed 39% (14/36) of sampled genes and queens (39%, 14/36). Of the genes that were (Fisher’s exact: p = 9.73 ¥ 10–5). Third, there is a differentially expressed, fewer were upregulated significant difference in the proportion of genes in workers with respect to queens (19%, 6/36), up-regulated by workers relative to queens in than in queens with respect to workers (42%, A. mellifera and P. canadensis (43% and 19% 15/36; Fisher’s exact: p = 0.037). This suggests respectively Fisher’s exact: p = 0.0083). This that the transcriptomes that give rise to queens comparison suggests that the degree of decou- and workers in P. canadensis have been une- pling between A. mellifera queens and workers qually decoupled. Workers express fewer unique has been equal and more extensive relative to P. genes or expression patterns than queens. One canadensis. interpretation of this is that workers are prima- Although this analysis is necessarily pre- rily queen-like but with part of the queen path- liminary and crude, the patterns that appear to be way suppressed (see Fig. 2, Observations): this emerging are that evolving from a solitary life- would suggest that workers in Polistes evolved style to a society with simple, behavioural castes primarily by delaying the development of the involves suppression of queen pathways in order reproductive state. to evolve workers, rather than workers becoming ANN. ZOOL. FENNICI Vol. 43 • Sociogenomics and the evolution of sociality 435 specialised. Increasing social complexity and able across research disciplines and taxonomic the evolution of morphological castes, by what- groups. Secondly, they illustrate how sociog- ever mechanism, is apparently accompanied by enomics allows us to address questions beyond true decoupling of the transcriptome such that the scope of current genetic-model organisms queens and workers exhibit separate and equally like Drosophila, which are not social. For exam- complementary phenotypes. Evidently, more rig- ple, we are beginning to understand the com- orous tests of this model are required, which plexity of the interactions that occur between the demand large-scale screening of random cDNA genome evolution and sociality, and the genetic microarrays for a range of species representing basis of reproductive division of labour. Thirdly, different levels of social complexity, as well as studying key issues in social evolution at the a better understanding of the relative effects of gene level reveals novel insights into the patterns different genes on phenotypic traits. Although and processes involved in sociality, invoking the data available were obtained from macroar- new ideas and clarifying theoretical models of rays of selective libraries for both P. canadensis both social and genome evolution. In this sec- and A. mellifera, the patterns outlined in Fig. tion I identify key areas for future research in 2 are unlikely to be artefacts. This is because sociogenomics, with a particular emphasis on the the genes examined were not all obtained from evolution of sociality. I summarise some practi- the same queen/worker subtractive libraries. For cal considerations for the future of sociogenom- example, almost half the genes used in the P. ics and discuss the potential for social insects as canadensis study were obtained from a library of genetic-model organisms of the future. newly emerged females which are neither queen nor worker (Sumner et al. 2006). I repeated the analysis using just these genes and similar pat- Major questions for sociogenomic terns were obtained, indicating that the pattern studies observed for P. canadensis in Fig. 2 is unlikely to be an artefact (Queens upregulate 4 genes A very general question in biology is how the whilst workers upregulate 1; another 4 genes are genome, with its limited genetic toolkit, produces equally expressed). Likewise, the A. mellifera the diversity of life observed around us (Carroll study included genes that were isolated in young, et al. 2005). The morphological, behavioural bipotent larvae that had not fully embarked on and social diversity displayed by social insects queen or worker pathways. Future analyses of makes them ideal for addressing this question, larger, random datasets using quantitative real- although it is important to distinguish between time PCR to quantify expression across more studies on the origins of sociality from those on species, including solitary ones, are however its maintenance. Here I outline key questions on needed in order to confirm the patterns suggested the genetic basis of the origin and maintenance here. of sociality in social insects that can now be addressed using sociogenomic techniques. Future prospects in sociogenomics How does the genome restrict the evolution The current achievements of sociogenom- of eusociality? ics (reviewed in this article) have already secured a promising future for social insects Eusociality evolves through the interplay of the as model systems in genomic research. Firstly, environment and genes. But it appears to be they demonstrate that viable and informative difficult to become eusocial, as it evident by gene-level studies can be performed on species the rare and clustered nature of eusocial clades whose genomes have been little studied (non (Wilson 1971). Unlike the role of the environ- genetic-model organisms). This is facilitated by ment in eusocial evolution, there has been little the increasingly affordable techniques available discussion of what genomic properties might for analysing genomes that are now transfer- restrict social evolution, partly because it is 436 Sumner • ANN. ZOOL. FENNICI Vol. 43 impossible to locate failed attempts of social Do the same genes/regulatory networks evolution. However, we can gain an insight into underlie both the origin and maintenance of the changes that occur in the transition between sociality? the two life styles by comparing the genomes of primitively eusocial species with solitary rela- The genome properties responsible for the onset tives and identifying the genomic changes that of eusocial evolution are not necessarily identi- occur. For example, is the regulation of certain cal to those that maintain it. There is growing genes uniquely altered in eusocial animals as evidence that a relatively small number of genes compared to solitary ones? Have eusocial ani- produces vast biological diversity through tem- mals evolved novel genes (e.g. for altruism) or poral and spatial variation in expression (Arnosti gene functions? Have they lost gene families or 2003 and ‘The genetic basis of queen and worker gene functions (Robinson & Ben-Shahar 2002)? castes’). The morphological, behavioural and Genes that are lost in eusocial species may be social diversity found amongst social insects some of the key inhibitors of social evolution. provides an excellent opportunity to examine We have begun to address these issues (reviewed how phenotypic diversity evolves at the level of in ‘Social evolution and genome evolution: the genes (West-Eberhard 1987, 1989). Primi- cause or consequence?’), but a more directed, tively eusocial species are the best representation comparative approach (Wei et al. 2002) is to be of a species at the onset of eusociality as few of encouraged. Examining how genomes change their traits are likely to be evolutionarily derived. after apparent reversions from eusociality to Within-genus/family comparisons of species (i.e. solitary living (Wcislo & Danforth 1997) would which share a recent common ancestor) occupy- be particularly informative. ing different levels of sociality can reveal how genes and/or their expression patterns change in order to produce different levels of social- Are polyphenisms that have evolved in ity. A good candidate group for such a study is several different social lineages maintained the bumblebee genus Bombus which displays a by the same genes and expression range of caste mechanisms that may reflect dif- patterns? ferent levels of sociality (Röseler 1991).

The major polyphenisms that characterise social insects have evolved several times in different Practical limitations and opportunities in lineages. The evolution of queen and worker sociogenomics castes is a prime example. Whether castes in dif- ferent lineages of the Hymenoptera are underlain It is important that sociogenomics progresses by the same genes, despite their independent away from simply correlating gene expression evolutionary pathways, was discussed in ‘The profiles with social attributes, towards manipu- genetic basis of queen and worker castes’. Other lating genes and altering phenotypic traits. A phenomena that have evolved multiple times in strong reason for doing this is that much of the the eusocial insects and would be good candi- gene expression detected in adult social insects dates for genomic studies include social parasit- may be a consequence of the behaviour rather ism, colony founding strategies, multiple mating than a cause (i.e. down-stream rather than up- and nest-mate communication and recognition stream genes). An indication of this is that so far systems. A phylogenetic approach of studying we have not found any cis-regulatory elements, the genes that underlie the same polymorphisms which regulate gene expression and are the real in different social lineages will reveal whether clues to the interaction between genes and social these genes/regulatory patterns are products of evolution. Moreover, the functions of isolated convergent evolution or whether they are con- genes are almost entirely inferred rather than served genes, inherited from a shared ancestor. proven. Studying gene expression has its own problems. Firstly, it is essentially the study of the set of reactions that control the abundance ANN. ZOOL. FENNICI Vol. 43 • Sociogenomics and the evolution of sociality 437 of gene products. Thus, although gene tran- and the action of evolution. Such comparisons script abundance reflects behavioural variation, can provide insights into the evolution of a spe- it does not necessarily predict gene product cies or particular trait (e.g. sociality), for exam- (e.g. protein) abundance. Protein targeting may ple in terms of gene birth and death, phylogenies, be a more accurate approach to studying the species/trait origins and adaptations. Compara- relationship between the genome and diversity tive genomics usually involves looking for simi- (Gerlai 2001), although proteomic techniques larities in sequences that infer homology (i.e. are currently less transferable between species that they have a common evolutionary ancestor). and less affordable than genomic techniques. Genome comparisons allow the reconstruction Secondly, gene expression studies are cluttered of genetic changes (or genetic ‘footprints’) that with noise. Distinguishing meaningful patterns occurred at crucial stages in the evolution of from noise is a challenge as we currently do not sociality. fully understand the sources of noise (Raser & Recent sequence data suggests that the hon- O’Shea 2005). Moreover, standardising expres- eybee shares over 95% of its orthologs with D. sion levels between samples is problematic as melanogaster (Whitfield et al. 2003). The long many housekeeping genes are not (as their name history of developmental genetics in Drosophila suggests) equally expressed. Some of these prob- therefore provides a very lucrative spring-board lems can be solved by manipulating transcript for sociogenomics. Researchers are now exploit- abundance and observing the altered phenotype. ing this resource in ingenious ways. For exam- Honeybee genomics is already testing the waters ple, it has been shown that Apis mellifera work- in this exciting new research field (Robinson ers up- or down-regulate an ortholog of the 2002a): gene expression has been localised to Drosophila foraging gene ( for) depending on certain body regions and even specific brain whether they are foragers or nurses (Ben-Shahar regions (Whitfield et al. 2003, Cash et al. 2005), et al. 2002). This indicates that gene regulation and RNAinterference (RNAi) techniques have of specific behaviours can be highly conserved, been employed to inactivate/activate target genes even over the 300 millions of years of evolution- (Amdam et al. 2003b). However, it is a steep ary history that separate Drospophila from Apis learning curve and likely riddled with unex- — see also Toma et al. (2000), Abouheif and pected problems. Behavioural traits are likely to Wray (2002), Ben-Shahar et al. (2002) for more be influenced by large suites of genes as well as examples of how Drosophila orthologs have environmental factors (e.g. Whitfield et al. 2003) been used to study social evolution. and our understanding of how genes interact with each other in general is still very basic, even for genetic-model organisms (Brazhnik et Social insects as genetic-model al. 2002). This leads to problems when trying organsims to use gene targeting to determine the function of a specific gene: for example, inactivating one The Hymenoptera lend themselves to compara- gene may instigate compensatory responses by tive genomics because they offer an unrivalled other genes, such that no phenotypic effect will range of behaviours, social adaptations and be observed. social complexity, with extant representatives Comparative genomics is a large-scale, for each of the steps likely to have been taken holistic approach to studying genome evolution, during social evolution (Evans & Wheeler 2000, which looks for similarities and differences in 2 2001). The future of comparative sociogenom- or more genomes (or parts of genomes), at any ics rests on careful selection of species from level, e.g. different species, subspecies or strains across the spectrum of sociality. There are practi- of the same species (Wei et al. 2002). It rests on cal considerations as well as scientific ones to the premise that the two genomes under exami- be made when selecting future model species nation share common ancestry such that every for sociogenomics (Evans & Gundersen-Rindal base pair in each organism can be explained as 2003). Behavioural phenotypes are especially the combination of the original ancestral genome difficult to study because they are sensitive to 438 Sumner • ANN. ZOOL. FENNICI Vol. 43 social and environmental fluctuations and they such as dominance hierarchies, queen superce- must be studied in a semi-natural context (Rob- dure, conflicts over egg-laying and choice of inson & Ben-Shahar 2002). Ideal model species reproductive strategy. Polistes is the basal genus for sociogenomics would be easy to rear under (most deep branching in a phylogeny) of a large controlled laboratory conditions, as well as easy clade (the Polistinae), making the study of these to study in their natural environments. They wasps especially informative because they can would be common and cosmopolitan in their divulge the state of genes before the evolu- distributions in order to maximise accessibility tion and radiation of more recent and divergent for an international body of researchers, and groups (Arevalo et al. 2004, Carroll et al. 2005). their natural history should be well understood. The Polistinae is the largest family of wasps The honeybee satisfies all these requirements and displays almost the entire range of sociality, and consequently is well on its way to being the from the simplest forms of true eusociality (e.g. first eusocial genetic-model organism. There is in independent nest founders like Polistes and already a wealth of quantitative genetic data Mischocyttarus), through to complex eusocial available for the honey bee, e.g. from QTL species (e.g. swarm founders like Agelaia and analyses (Hunt & Page 1995, Page et al. 2002) Polybia). The potential for exploring the origins, moreover, the annotation of the honey bee’s evolution and maintenance of sociality using genome sequence is now complete (The Honey- comparative genomics within this group is there- bee Genome Sequencing Consortium 2006). The fore immense. honeybee has emerged as an ideal candidate for sociogenomics, facilitating the study of social evolution at one of its most complex states. This Acknowledgements is excellent news for students of sociality, as social insects now have a credible reputation for I would like to thanks the participants in the symposium “Polistes paper wasps: emergence of a model system” for genomic studies. But insights into the origins of helpful discussions, and Andrew Bourke, Bill Jordan and sociality are best obtained through complemen- Zjef Pereboom for insightful comments on earlier drafts of tary studies of simpler social organisations than the manuscript. those exhibited by the honeybee. Polistes paper wasps are probably the best- studied genus of primitively eusocial insects References (Turillazzi & West-Eberhard 1996). They are cosmopolitan in their distribution, can be reared Abouheif, E. & Wray, G. A. 2002: Evolution of the gene net- work underlying wing polyphenism in ants. — Science relatively easily in natural, semi-natural (e.g. 297: 249–252. nest boxes) and lab environments, facilitating Albert, S. & Klaudiny, J. 2004: The MRJP/YELLOW protein manipulation experiments. Their genome is rela- family of Apis mellifera: identification of new members tively small (300Mb, Johnson et al. (2004)), in the EST library. — J. Insect Physiol. 50: 51–59. and transcriptome studies (i.e. on mRNA) have Amdam, G. V., Norberg, K., Fondrk, M. K. & Page, R. E. 2004: Reproductive ground plan may mediate colony- already commenced on one species (Sumner et level selection effects on individual foraging behav- al. 2006). Colony-level social life in Polistes is iour in honey bees. — P. Natl. Acad. Sci. USA 101: highly dynamic as queen and worker castes are 11350–11355. behavioural and determined during adulthood Amdam, G. V., Norberg, K., Hagen, A. & Omholt, S. W. such that nest-mates compete to realise their 2003a: Social exploitation of vitellogenin. — P. Natl. Acad. Sci. USA 100: 1799–1802. reproductive potential (Reeve 1991). Moreover, Amdam, G. V., Simoes, Z. L. P., Guidugli, K. R., Norberg, K. reproductive options for females are varied: a & Omholt, S. W. 2003b: Disruption of vitellogenin gene female may remain as a worker all her life, or function in adult honeybees by intra-abdominal injection she may succeed the existing queen or found of double-stranded RNA. — BMC Biotechnology 3, art. her own colony (West-Eberhard 1969, Hughes 1. doi:10.1186/1472-6750-3-1. Arevalo, E., Zhu, Y., Carpenter, J. M. & Strassmann, J. E. et al. 1987, Peters et al. 1995). Thus, Polistes 2004: The phylogeny of the social wasp subfamily Polis- provides opportunities to study the genetic basis tinae: evidence from microsatellite flanking sequencing, of behavioural traits not found in honeybees, mitochondrial CO1 sequence and morphological char- ANN. ZOOL. FENNICI Vol. 43 • Sociogenomics and the evolution of sociality 439

acters. — BMC Evol. Biol. 4, art. 8. doi:10.1186/1471- — Genome Biol. 4: 107. 2148-4-3. Evans, J. D. & Wheeler, D. E. 1999: Differential gene Arnosti, D. N. 2003: Analysis and function of transcriptional expression between developing queens and workers in regulatory elements: Insights from Drosophila. — Annu. the honey bee, Apis mellifera. — P. Natl. Acad. Sci. USA Rev. Entomol. 48: 579–602. 96: 5575–5580. Ben-Shahar, Y., Robichon, A., Sokolowski, M. B. & Robin- Evans, J. D. & Wheeler, D. E. 2000: Expression profiles son, G. E. 2002: Influence of gene action across different during honeybee caste determination. — Genome Biol. time scales on behaviour. — Science 296: 741–744. 2: 1–6. Bourke, A. F. G. & Franks, N. R. 1995: Social evolution in Evans, J. D. & Wheeler, D. E. 2001: Gene expression and ants. — Princeton University Press, Princeton. the evolution of insect polyphenisms. — BioEssays 23: Bourke, A. G. F. 1999: Colony size, social complexity and 62–68. reproductive conflict in social insects. — J. Evol. Biol. Fitzpatrick, M. J., Ben-Shahar, Y., Smid, H. M., Vet, L. E. 12: 245–257. M., Robinson, G. E. & Sokolowski, M. B. 2004: Can- Brazhnik, P., de la Fuente, A. & Mendes, P. 2002: Gene net- didate genes for behavioural ecology. — Trends Ecol. works: how to put the function in genomics. — Trends Evol. 20: 96–104. Biotechnol. 20: 467–472. Gadagkar, R. 1997: The evolution of caste polymorphism in Bromham, L. & Leys, R. 2005: Sociality and the rate of social insects: genetic release followed by diversifying molecular evolution. — Mol. Biol. Evol. 22: 1393–1402. evolution. — J. Genet. 76: 167–179. Carroll, S. B., Grenier, J. K. & Weatherbee, S. D. 2005: From Gadau, J., Page, R. E., Werren, J. H. & Schmid-Hempel, DNA to diversity, 2nd ed. — Blackwell Publishing, P. 2000: Genome organisation and social evolution in Oxford. Hymenoptera. — Naturwissenschaften 87: 87–89. Cash, A. C., Whitfield, C. W., Ismail, N. & Robinson, G. E. Gerlai, R. 2001: Gene targeting: technical confounds and 2005: Behaviour and the limits of genomic plasticity: potential solutions in behavioural brain research. — power and replicability in microarray analysis of honey- Behav. Brain Res. 125: 13–21. bee brains. — Genes, Brain Behav. 4: 267–271. Giray, T., Giovanetti, M. & West-Eberhard, M. J. 2005: Juve- Charlesworth, B. 1978: Some models of the evolution of nile hormone, reproduction & worker behaviour in the altruistic behaviour between siblings. — J. Theor. Biol. neotropical social wasp Polistes canadensis. — P. Natl. 72: 297–319. Acad. Sci. USA 102: 3330–3335. Corona, M., Estrada, E. & Zurita, M. 1999: Differential Hamilton, W. D. 1964: The genetical evolution of social expression of mitochondrial genes between queens and behaviour. I. — J. Theor. Biol. 7: 1–16. workers during caste determination in the honeybee, Helms-Cahan, S. & Keller, L. 2003: Complex hybrid origin Apis mellifera. — J. Exp. Biol. 202: 929–938. of genetic caste determination in harvester ants. — Crosland, M. W. J. & Crozier, R. H. 1986: Myrmecia pilo- Nature 424: 306–309. sula, an ant with only one pair of chromosomes. — Sci- Hojo, M., Morioka, M., Matsumoto, T. & Miura, T. 2005: ence 231: 1278. Identification of soldier caste-specific protein in the Crozier, R. H. 1979: Genetics of sociality. — In: Hermann, frontal gland of nasute termite Nasutitermes takasagoen- H. R. (ed.), Social insects, vol. 1: 223–286. Academic sis (Isoptera: Termitidae). — Insect Biochem. Molec. Press, New York. Biol. 35: 347–354. Crozier, R. H. 1982: On insects and insects: twists and turns Hölldobler, B. & Wilson, E. O. 1990: The ants. — Springer in our understanding of the evolution of eusociality. Verlag, Berlin. — In: Breed, M., Michener, C. & Evans, H. (eds.), The Hughes, C. R., Beck, M. O. & Strassmann, J. E. 1987: biology of social insects: 4–9. Westview Press, Boulder, Queen succession in the social wasp, Polistes annularis. Colorado. — Ethology 76: 124–132. Crozier, R. H. 1987: Towards a sociogenetics of social Hunt, G. J. & Page, R. E. 1995: Linkage map of the honey- insects. — In: Eder, J. & Rembold, H. (eds.), Chem- bee, Apis mellifera, based on RAPD markers. — Genet- istry and biology of social insects: 325–328. Verlag J. ics 139: 1371–1382. Peperny, Munich. Hunt, J. H. & Amdam, G. V. 2005: Bivoltinism as an ante- Darwin, C. 1859: On the origin of species. — Penguin cedent to eusociality in the paper wasp genus Polistes. Books, London. — Science 308: 264–267. Dawkins, R. 1976: The selfish gene. — Oxford University Johnston, J. S., Ross, L. D., Beani, L., Hughes, D. P. & Press, Oxford. Kathirithamby, J. 2004: Tiny genomes and endoreplica- Dawkins, R. 1982: The extended phenotype. — W.H. Free- tion in Strepsiptera. — Insect Mol. Biol. 13: 581–585. man and Co., Oxford. Judice, C., Hartfelder, K. & Pereira, G. A. G. 2004: Caste- Espinosa-Jeffrey, A., Kumar, S., Zhao, P. M., Awosika, O., specific gene expression in the stingless bee Melipona Agbo, C., Huang, A., Chang, R. & De Vellis, J. 2002: quadrifasciata — are there common patterns in highly Transferrin regulates transcription of the MBP gene and eusocial bees? — Insectes Soc. 51: 352–358. its action synergizes with IDF-1 to enhance myelinogen- Julian, G. E., Fewell, J. H., Gadau, J., Johnson, R. A. & Lar- esis in the md rat. — Dev. Neurosci. 24: 227–241. rabee, D. 2002: Genetic determination of the queen caste Evans, H. & Gundersen-Rindal, D. 2003: Beenomes to in an ant hybrid zone. — P. Natl. Acad. Sci. USA 99: Bombyx: future directions in applied insect genomics. 8157–8160. 440 Sumner • ANN. ZOOL. FENNICI Vol. 43

Keller, L. 1997: Indiscriminate altruism: unduly nice parents Piulachs, M. D., Guidugli, K. R., Barchuk, A. R., Cruz, J., and siblings. — Trends Ecol. Evol. 12: 99–103. Simoes, Z. L. P. & Belles, X. 2003: The vitellogenin Keller, L. (ed.) 1999: Levels of selection. — Princeton Uni- of the honey bee, Apis mellifera: structural analysis of versity Press, Princeton. the cDNA and expression studies. — Insect Biochem. Krieger, M. J. B. & Ross, K. G. 2002: Identification of a Molec. Biol. 33: 459–465. major gene regulation complex social behaviour. — Sci- Queller, D. C. 1992: A general model for kin selection. ence 295: 328–332. — Evolution 46: 376–380. Kucharski, R. & Maleszka, R. 1998: A royal jelly protein in Queller, D. C., Hughes, C. R. & Strassmann, J. E. 1990: expresses in a subset of Kenyon cells in the mushroom Wasps fail to make distinctions. — Nature 344: 388. bodies of the honey bee brain. — Naturwissenschaften Raser, J. M. & O’Shea, E. K. 2005: Noise in gene expres- 85: 343–346. sion: origins, consequences and control. — Science 309: Kucharski, R. & Maleszka, R. 2002a: Evaluation of differ- 2010–2013. ential gene expression during behavioural development Reeve, H. 1991: Polistes. — In: Ross, K. G. & Matthews, R. in the honeybee using microarrays and northern blots. W. (eds.), The social biology of wasps: 99–148. Cornell — Genome Biol. 3 art. 2. doi:10.1186/gb-2002-3-2- University Press, Ithaca. research0007. Robinson, G. E. 1999: Integrative animal behaviour and Kucharski, R. & Maleszka, R. 2002b: Molecular profiling sociogenomics. — Trends Ecol. Evol. 14: 202–205. of behavioural development: differential expression of Robinson, G. E. 2002a: Genomics and integrative analysis mRNAs for inositol 1,4,5-triphosphate 3-kinase iso- of division of labour in honeybee colonies. — Am. Nat. forms in naive and experienced honeybees (Apis mel- 160: S160–S172. lifera). — Mol. Brain Res. 99: 92–101. Robinson, G. E. 2002b: Sociogenomics takes flight. — Sci- Kucharski, R. & Maleszka, R. 2003: Transcriptional profiling ence 297: 204–205. reveals mutilfunctional roles for transferrin in the honey- Robinson, G. E. & Ben-Shahar, Y. 2002: Social behaviour bee, Apis mellifera. — J. Insect Sci. 3: 1–8. and comparative genomics: new genes or new gene Li, W.-H., Gu, Z., Wnag, H. & Nekrutenko, A. 2001: Evolu- regulation? — Genes, Brain Behav. 1: 197–203. tionary analyses of the human genome. — Nature 409: Robinson, G. E., Grozinger, C. M. & Whitfield, C. W. 2005: 847–849. Sociogenomics: social life in molecular terms. — Nat. Linksvayer, T. A. & Wade, M. J. 2005: The evolution- Rev. Genet. 6: 257–270. ary origin and elaboration of sociality in the aculeate Röseler, P.-F. 1991: Roles of morphogenetic hormones in Hymenoptera: maternal effects, sib-social effects and caste polymorphism in bumblebees. — In: Gupta, A. P. heterochrony. — Q. Rev. Biol. 80: 317–336. (ed.), Morphogenetic hormones of arthropods. Roles in Michod, R. E. 1982: The theory of kin selection. — Annu. histogenesis, organogenesis and morphogenesis: 384– Rev. Ecol. Syst. 13: 23–55. 399. Rutgers University Press, New Brunswick. Miura, T. 2001: Morphogenesis and gene expression in the Scharf, M. E., Wu-Scharf, D., Pittendrigh, B. R. & Bennett, soldier-caste differentiation of termites. — Insectes Soc. G. W. 2003: Caste- and development-associated gene 48: 216–223. expression in a lower termite. — Genome Biol. 4, R62. Miura, T. 2004: Proximate mechanisms and evolution of doi:10.1186/gb-2003-4-10-r62. caste polyphenism in social insects: from sociality to Scharf, M. E., Wu-Scharf, D., Zhou, X., Pittendrigh, B. R. & genes. — Ecol. Res. 19: 141–148. Bennett, G. W. 2004: Gene expression profiles among Miura, T., Koshikawa, S. & Matsumoto, T. 2003: Winged immature and adult reproductive castes of the termite presoldiers induced by a juvenile hormone analog in Reticulitermes flavipes. — Insect Mol. Biol. 14: 31–44. Zootermopsis nevadensis: implications for plasticity and Schmitz, J. & Moritz, R. F. A. 1998: Sociality and the rate of evolution of caste differentiation in termites. — J. Mor- rDNA sequence evolution in wasps (Vespidae) and hon- phol. 257: 22–32. eybees (Apis). — J. Mol. Evol. 47: 606–612. Miura, T., Kamikouchi, A., Takeuchi, H., Natori, S., Kubo, Sherman, P. W. 1979: Insect chromosome numbers and euso- T. & Matsumoto, T. 1999: Soldier caste-specific gene ciality. — Am. Nat. 113: 925–935. expression in the mandibular glands of hodotermopsis Sumner, S., Pereboom, J. J. M. & Jordan, W. C. 2006: Dif- japonica (Isoptera: Termopsidae). — P. Natl. Acad. Sci. ferential gene expression and phenotypic plasticity in USA 96: 13874–13879. behavioural castes of the primitively eusocial wasp, Normark, B. B. 2003: The evolution of alternative genetic Polistes canadensis. — Proc. R. Soc. Lond B 273: systems in insects. — Annu. Rev. Entomol. 48: 397–423. 19–26. Page, R. E., Gadau, J. & Beye, M. 2002: The emergence of The Honeybee Genome Sequencing Consortium 2006: Hymenoptera genetics. — Genetics 160: 375–379. Insights into social insects from the genome of the hon- Pereboom, J. J. M., Jordan, W. C., Sumner, S., Hammond, R. eybee Apis mellifera. — Nature 443: 931–949. L. & Bourke, A. F. G. 2005: Differential gene expression Thorne, B. L. 1997: Evolution of eusociality in termites. in queen-worker caste determination in bumblebees. — Annu. Rev. Ecol. Syst. 28: 27–54. — Proc. R. Soc. Lond. B 272: 1145–1152. Tian, H., Vinson, B. & Coates, C. J. 2004: Differential gene Peters, J. M., Queller, D. C., Strassmann, J. E. & Solis, C. R. expression between alate and dealate queens in the red 1995: Maternity assignment and queen replacement in a imported fire ant, Solenopsis invicta Buren (Hymenop- social wasp. — Proc. R. Soc. Lond. B 260: 7–12. tera: Formicidae). — Insect Bioche. Molec. Biol. 34: ANN. ZOOL. FENNICI Vol. 43 • Sociogenomics and the evolution of sociality 441

937–949. West-Eberhard, M. J. 1969: The social biology of polistine Toma, D. P., Bloch, G., Moore, D. & Robinson, G. E. 2000: wasps. — Misc. Publ. Mus. Zool. Univ. Michigan 140: Changes in period mRNA levels in the brain and divi- 1–101. sion of labor in honeybee colonies. — P. Natl. Acad. Sci. West-Eberhard, M. J. 1987: Flexible strategy and social USA 97: 6914–6919. evolution. — In: Ito, Y., Brown, J. L. & Kikkawa, J. Trivers, R. L. & Hare, H. 1976: Haplodiploidy and the evolu- (eds.), Animal societies: theories and facts: 35–51. Japan tion of the social insects. — Science 191: 249–263. Scientific Societies Press, Tokyo. Turillazzi, S. & West-Eberhard, M. J. 1996: Natural history West-Eberhard, M. J. 1989: Phenotypic plasticity and the ori- and evolution of paper wasps. — Oxford University gins of diversity. — Annu. Rev. Ecol. Syst. 20: 249–278. Press, Oxford. West-Eberhard, M. J. 1996: Wasp societies as microcosms Volny, V. P. & Gordon, D. M. 2002: Genetic basis for queen- for the study of development and evolution. — In: worker dimorphism in a social insect. — P. Natl. Acad. Turillazzi, S. & West-Eberhard, M. J. (eds.), Natural Sci. USA 99: 6108–6111. history and evolution of paper-wasps: 290–317. Oxford Wade, M. J. 1980: Kin selection: its components. — Science University Press. 210: 665–667. Wheeler, D. E. 1986: Developmental and physiological Wcislo, W. T. & Danforth, B. N. 1997: Secondarily solitary: determinants of caste in social Hymenoptera: evolution- the evolutionary loss of social behaviour. — Trends ary implications. — Am. Nat. 128: 13–34. Ecol. Evol. 12: 468–474. Whitfield, C. W., Cziko, A. M. & Robinson, G. E. 2003: Wei, L., Liu, Y., Dubchak, I., Shon, J. & Park, J. 2002: Gene expression profiles in the brain predict behaviour Comparative genomics approaches to study organism in individual honey bees. — Science 302: 296–299. similarities and differences. — J. Biomed. Inform. 35: Wilson, E. O. 1971: The insect societies. — Harvard Univer- 142–150. sity Press, Cambridge, MA.

This article is also available in pdf format at http://www.annzool.net/ 442 Sumner • ANN. ZOOL. FENNICI Vol. 43 - . W vs vs . W reproductive R. flavipes Q vs . W Alate Sub-

vs . S S. invicta

vs . W Delate Adults

vs . F alate Q Q (workers) . F vs

et al . 2003. COX I & COX II = cytochrome oxidase I & II respectively. MRJP . non- 0 F Naïve N N N N N 0 N N N Eq W W W 0 N N N N N N N N N N N N N N N vs

B. terrestris A. mellifera vs . W Reproductive Nurse Naïve Q

vs . Q F vs. canadensis

vs . W Q invicta

Larvae N Eq Q N N Nurse N Dealate Dealate Eq N N N N N N N 0 N N N Q N N Q N N N N N Nurse N Naïve N N Nurse N Naïve F F Virgin Q N N Q N N Dealate N N N N N N N N N N N N N N N N N N F/Q N 0 N N N N N W W Sub N N N N N Young N 0 N N Naïve N N W Eq Sub Q N Q N N N N N N N N N Q W F/Q N N N N Eq Dealate N N N Q N N N N N N N N Eq Alate Sub Eq N F N N F Naïve N N Eq Eq Eq vs . W Q mellifera

vs . Q 0 0 N N N N N N N N Q W(late) W N Q 0 W (early & Young W B. terrestris A. S. P. late instars) female reproductive W Q(early)/ W(late) Q Eq N W Non-reprod N Naïve Eq Eq S N N Vitellogenin Lectin like MRJP Peroxiredoxin SPARC Transferrin Unknown 1 Q(early)/W(late) W N F/Young Q Non-reprod N N N N N N N Heat shock protein 70 Hexamerin I Hexamerin II N Imaginal disc Cuticle protein COX I COX II Q(early)/W(late) Q N N 0

A summary of the genes expressed in 2 or more species of social insect with respect to caste. The comparison made is indicated for each species. Direction species. each for indicated is made comparison The caste. to respect with insect social of species more or 2 in expressed genes the of summary A 1. Appendix of expression is indicated by named caste; Equal (Eq) = no difference in expression detected between castes; 0 = no expression detected; N = no data available. Refer ence sources given in text. Q = queen, W worker, F forager, S soldier, Y Young. Putative gene 18S ribosomal protein N I included only the 50 most predictive ESTs for nurse and foraging honeybees from Whitfield = major royal jelly protein. 28S ribosomal protein Ribosomal protein ATP synthase Q(early)/W(late) N CG31605 W Q(early)/W(late) Q N Q/Young W W F Reprod W N N N N N Naïve N Eq N Eq N N N N N