Comparative sociogenomics has the potential to provide important insights into how social behaviour evolved. We examined brain gene expression profiles of the primitively eusocial wasp Polistes metricus and compared the results with a growing base of brain gene expression information for the advanced eusocial honeybee, Apis mellifera. We studied four female wasp groups that show variation in foraging/provisioning behaviour and reproductive status, using our newly developed microarray representing approximately 3248 P. metricus genes based on sequences generated from high-throughput pyrosequencing. We found differences in the expression of approximately 389 genes across the four groups. Pathways known from Drosophila melanogaster to be related to lipid metabolism, heat and stress response, and various forms of solitary behaviour were associated with behavioural differences among wasps. Forty-five per cent of differentially expressed transcripts showed significant associations with foraging/provisioning status, and 14 per cent with reproductive status. By comparing these two gene lists with lists of genes previously shown to be differentially expressed in association with honeybee division of labour, we found a significant overlap of genes associated with foraging/provisioning, but not reproduction, across the two species. These results suggest common molecular roots for foraging division of labour in two independently evolved social insect species and the possibility of more lineage-specific roots of reproductive behaviour. We explore the implications of these findings for the idea that there is a conserved ‘genetic toolkit’ for division of labour across multiple lineages.
Studies of the mechanisms underlying social behaviour have the potential to provide important insights into how complex phenotypes can evolve. Recent findings that honeybee social foraging behaviour is regulated by several genes associated with feeding behaviour in Drosophila melanogaster (Ben-Shahar et al. 2002; Ament et al. 2008) suggest that social behaviour can evolve from solitary behaviour via changes in the regulation of deeply conserved genes (Toth & Robinson 2007; Smith et al. 2008).
Based on such findings, it has been further suggested that there may be a ‘genetic toolkit’ for eusocial behaviour in the insects, similar to the genetic toolkit that has been elucidated for development, in which a core set of genes is used repeatedly during social evolution to generate novel forms of behaviour (Toth & Robinson 2007). A key prediction of this idea is that some of the same pathways that regulate division of labour in the well-studied honeybee (Smith et al. 2008) also regulate division of labour in other, independently evolved social insect lineages. However, to date, cross-species comparisons of social insects have been limited for two reasons. First, there has been a paucity of advanced genomic resources for most species (reviewed in Smith et al. 2008). Second, direct comparisons have been challenging owing to differences across studies in the tissues and developmental stages analysed, as well as the techniques used to measure gene expression (reviewed in Sumner 2006; Smith et al. 2008). As a consequence, the extent of overlap across species, and thus the relevance of the ‘toolkit’ idea to social evolution, is unknown.
There have been several studies with the honeybee (Apis mellifera) examining genome-wide expression patterns associated with division of labour, providing an excellent basis to begin cross-species comparisons. This includes work on both the reproductive division of labour between queens and workers (Barchuk et al. 2007; Grozinger et al. 2007) and division of labour among workers (Whitfield et al. 2003, 2006; Cash et al. 2005).
Using honeybee gene expression data as a source of candidate genes, Toth et al. (2007) performed a comparative genomic analysis by using quantitative real-time PCR (qPCR) to measure brain expression for 32 genes in behaviourally differentiated groups of Polistes metricus wasps. Wasps and bees diverged approximately 100–150 Myr ago (Danforth et al. 2006), and social paper wasps have an independent origin of sociality from honeybees. Paper wasp division of labour is regulated by competition and dominance interactions among reproductively competent adults (Pardi 1948). By contrast, worker–worker division of labour in honeybees involves higher levels of cooperation and pheromone regulation (Winston 1987). Despite these differences in social biology, Toth et al. (2007) found a large overlap in patterns of brain gene expression in honeybees and paper wasps, suggesting strong commonalities in the regulation of different systems of division of labour at the molecular level. However, the same study used genes that were carefully selected based on their known associations with honeybee division of labour; thus, the actual extent of overlap between the two species could not be fairly estimated. For a broader set of genes, what is the extent of overlap across species? A more global analysis of P. metricus brain gene expression is necessary to further test the hypothesis of a genetic toolkit for division of labour in the social insects. We did this using a newly developed microarray platform, whose characterization is also presented here.
2. Material and methods
We collected P. metricus from naturally occurring nests at Allerton Park in Monticello, Illinois, as in Toth et al. (2009). We studied four adult female groups: (i) foundresses, which initiate colonies in the spring. Although P. metricus sometimes cooperate to found nests, we only analysed solitary foundresses; (ii) queens, which are successful foundresses that have reared a set of adult worker offspring. Queens specialize mainly in egg-laying and cease foraging and provisioning brood; (iii) workers, which perform several behaviours that relate to colony growth and maintenance including foraging for food and provisioning brood. We focused on foraging workers by collecting only workers with evidence of wing wear, an indicator of foraging experience (Toth et al. 2009); and (iv) gynes, which are future reproductive females that emerge late in the colony cycle. We collected gynes before they dispersed from their natal nests to overwinter, during which time they generally do not work and are also not actively egg-laying. We collected 12 females from each group and attempted to collect queens, workers and gynes from the same nests whenever possible. Wasps were paint-marked every 3–4 days, so the approximate dates of emergence of workers and gynes were known. Complete information on collection dates and ages of wasps is in table S1 of the electronic supplementary material. All wasps were freeze-killed by collecting onto dry ice and stored at −80°C. Brains were dissected from frozen specimens as described in Toth et al. (2007).
As colony usurpation is common in Polistes, microsatellite analyses (Henshaw et al. 2004) were performed to verify that the queens were the mothers of the gynes and workers from their colonies (as in Toth et al. 2009). The results were consistent with mother–daughter relationships for all but one gyne from one colony (table S2, electronic supplementary material).
70 mer oligos were synthesized by Invitrogen (Carlsbad, CA, USA) and custom-printed onto microarrays at the Keck Center for Comparative and Functional Genomics at the University of Illinois. This platform has been previously used successfully with honeybees (Kocher et al. 2008; Alaux et al. 2009a,b). Probes were designed by Invitrogen based on expressed sequence tag (EST) sequences derived from 454/Roche pyrosequencing (described in Toth et al. 2007). The sequences were assembled into consensus contigs using a slightly modified version of the StackPack pipeline (http://www.egenetics.com/stackpack.html). ‘High-quality’ probes were designed based on contigs that: (i) contained over 100 sequences or (ii) had a ‘specificity score’ (supplied by Invitrogen as a metric of specificity) greater than 0.32, and either (iii) had a TBLASTX e-value < 1 E-4 against the honeybee predicted genes (HoneyBee Genome Sequencing Consortium 2006) or (iv) had a BLASTX e-value < 1 E-4 against the GenBank proteins. This resulted in 7207 oligos. Then, we excluded redundant matches to the same bee EST sequence (Whitfield et al. 2002) unless it matched a different GenBank protein, and redundant matches to the same GenBank protein unless it matched a different honeybee EST. This resulted in 5102 oligos. We then excluded all sequences with BLASTN e-values < 1 E-6 to ribosomal RNA and all sequences with ‘probe scores’ (supplied by Invitrogen as a metric of hybridization quality and synthesis efficiency) less than 250, resulting in 4948 oligos.
Forty additional oligos were designed based on genes from consensus assembled sequences of interest (Toth et al. 2007) or their roles in behaviour and physiology in other insects, but were not included based on the aforementioned criteria. The best available oligonucleotide sequence from each targeted contig was used for oligo design. This gave a final number of 4988 P. metricus probes, which we estimated to correspond to at least 3248 different P. metricus genes based on TBLASTX matches to different honeybee predicted genes or BLASTX matches to different GenBank proteins (e-value < 1 E-4). We also added 240 control probes (supplied by Invitrogen) based on unrelated sequences from other organisms. This gave a total of 5228 probes on the array, each spotted twice for a total of 10 456 spots.
We used a standard two dye (Cy3 and Cy5) microarray protocol as previously described (Alaux et al. 2009a). The Maui hybridization system (BioMicro Products, Salt Lake City, UT, USA) significantly improved hybridization efficiency and was used for all experimental arrays. We used a loop design, consisting of 48 individual wasps (12 per group) paired on 72 arrays with dye swaps. Each wasp brain was analysed on either two or four arrays (figure S1, electronic supplementary material).
To validate the performance of the microarray, we prepared a set of two self–self hybridizations derived from pooled samples of wasps. Each pooled sample was split into two, one half labelled with Cy3, the other with Cy5, and these were hybridized together on the same microarray. These tests allowed us to characterize hybridization efficiency of the array probes without the complication of expression differences.
(c) Statistical analyses
Microarray spots were quality filtered by ‘spotfinding’ as in Alaux et al. (2009a). Fluorescence intensities were normalized using a LOWESS transformation (Yang et al. 2002). Duplicate spots for each gene were averaged and adjusted for microarray and dye effects. A mixed-effects ANOVA model using Proc Mixed in SAS was used to describe the expression measurements of each oligo. The model included the fixed effects of dye, foraging/provisioning (P), reproduction (R) and the interaction of the two terms, and the random effects of wasp and array. Each group was coded as follows (as in Toth et al. 2007, 2009): foundresses (P+/R+), gynes (P−/R−), queens (P−/R+) and workers (P+/R−). Significance was based on a false discovery rate of p < 0.05, using an F1 statistic. After the analysis was completed, we eliminated oligos with less than 60 per cent of the spots above background intensity across all arrays. We performed an alternative ANOVA analysis using Proc Mixed that included the fixed effects of dye and group (which had four levels: ‘foundress’, ‘gyne’, ‘queen’ and ‘worker’), and the random effects of wasp, array and colony. These results are not presented in the main text but are included in table S3 of the electronic supplementary material.
The resulting list of oligos showing significant differences in expression (‘wasp differentially expressed, full list’) was used for subsequent analyses. To look at overall patterns in the data and understand the relationships among overall gene expression patterns between the four groups, we performed several multivariate analyses. We performed hierarchical clustering analysis (HCA) by oligo and wasp using Genesis (Sturn et al. 2002). We performed a linear discriminant analysis (LDA) using the ‘lda’ function in R. Because LDA requires there to be more individual samples than variables, we randomly selected 40 oligos from the differentially expressed list for LDA.
In addition, we used p-values for each factor in the ANOVA to cut the Polistes gene list into two lists, one containing genes differentially expressed in association with foraging/provisioning status (‘wasp foraging/provisioning sublist’) and the second containing genes differentially expressed in association with reproductive status (‘wasp reproduction sublist’). We compared these lists with gene lists from prior honeybee studies. We used lists from two studies comparing brain gene expression in nurse and forager honeybees (‘bee foraging lists’; Whitfield et al. 2003; Alaux et al. 2009b) and one comparing brain gene expression in queens and ‘sterile’ workers (‘bee reproduction list’; Grozinger et al. 2007).
We limited the bee–wasp comparisons to wasp oligos for which there was a good TBLASTX hit to a honeybee gene (e-value < 1 E-4) and to genes that were analysed on both bee and wasp array platforms. A gene was counted as overlapping regardless of the direction of expression (whether the expression level was up- or downregulated in one species relative to the other). This approach is justified because several studies suggest that genes (e.g. the foraging gene in bees and ants) can retain conserved roles in behaviour across species, but that the direction of gene regulation may be evolutionarily labile (Robinson & Ben-Shahar 2002; Ingram et al. 2005). We used two independent statistical tests to examine the overlap between genes that were differentially regulated in wasps to the three aforementioned bee gene lists. First, we used one-tailed Fisher exact tests to compare the number of observed overlaps to a null hypothesis of the number of overlaps expected by random chance. Second, simulation tests were performed by drawing a random set of x and y genes from each of the two lists being compared, where x is the number of genes significant in the wasp list and y is the number significant in the bee list. The number of overlapping genes between each set of two samplings represents one randomly drawn overlap. This sampling regime (without replacement) was repeated 4000 times, creating a distribution of overlaps. The probability of the actual overlap was then calculated based on the simulated distribution.
In addition to the aforementioned comparisons, tests of the probability of overlap (again using Fisher exact tests and simulations) were also performed for five additional wasp gene sublists that reflected group-specific patterns, and the results are reported in the electronic supplementary material.
Possible functional significance was explored with gene ontology (GO) analysis. We tested for enriched terms in the three wasp gene lists: ‘wasp differentially expressed, full list’, ‘wasp foraging/provisioning sublist’ and ‘wasp reproduction sublist’ (using DAVID: Dennis et al. 2003; Huang et al. 2009). We focused only on genes for which there were good TBLASTX hits to annotated genes in D. melanogaster (e-value < 1 E-4). We used a raw p-value cut-off of p < 0.05 and a minimum of three gene products in a given category, and only considered GO categories at the level of ‘biological process’.
(a) Validation of microarray performance
In both replicates of the ‘self–self’ hybridizations, approximately 95 per cent of the spots on the two test arrays hybridized at a ‘sum of median intensity in both channels 635/532 nm’ (SMI) of greater than 100 (above background intensity, 40–100). We found a wide range of SMI values (mean ± s.e. = 3569 ± 68). There were control spots on the array consisting of probes with a gradient of incorrect nucleotides (5%, 10%, 15%, 20%, 30%, 40%, 50% mismatch) for 20 randomly selected oligos. As expected, these showed a decrease in average SMI as the per cent mismatch increased (7303 (5%), 1106 (10%), 227 (15%), 310 (20%), 202 (30%), 229 (40%) and 201 (50%)). For each of these 20 oligos, there was a negative control (the same sequence randomly scrambled), which had an average SMI of 149. The average SMI for blank spots on the test arrays was 69 (in the range of background intensity). We also found a significant correlation between the SMI values of duplicate spots (R = 0.96, p < 0.001). These diagnostics gave us confidence that the array was producing reliable hybridizations.
We also used the complete set of experimental microarrays for further validation. We tested whether the expression patterns agreed with two predictions: (i) oligos with a best match to the same putative wasp gene (e.g. had the same best TBLASTX hit to a honeybee gene) should result in a similar level of hybridization and (ii) the expression ratios for the four groups should be correlated with values from our previous qPCR-based study (Toth et al. 2007). For both of these analyses, we used normalized mean expression estimates from the ANOVA described above. To address the first prediction, we compared the average estimate for all pairs of wasp contigs that had a best TBLASTX hit to the same honeybee gene. We found a highly significant correlation across the wasp contig pairs (R = 0.654, p < 0.0001, n = 1206 contigs corresponding to 603 honeybee genes). To address the second prediction, we compared mean fold-difference across six comparisons (foundress–gyne, foundress–queen, foundress–worker, gyne–queen, gyne–worker and queen–worker). For 10 oligos based on the same exact EST contigs from both qPCR and the microarray, we found a significant correlation between the fold-differences (R = 0.30, p = 0.02, n = 10). Although the correlation was not high, we considered it to be confirmatory considering the two sets of analyses were performed on different wasps collected in different years.
(b) Differentially expressed genes
A total of 447 oligos were differentially expressed across the four groups. These 447 oligos corresponded to 389 different genes (based on best TBLASTX hits to honeybee predicted genes or best BLASTX hits to GenBank proteins), which is approximately 12 per cent of the genes represented on the microarray. Of these, expression differences for 45 per cent (201 oligos) were associated with differences in foraging/provisioning status, 14 per cent (63 oligos) with reproduction and 38 per cent (169 oligos) with the interaction between these two terms. A description of the expression patterns of candidate genes from our previous study (Toth et al. 2007) including Pmforaging and PmVitellogenin can be found in the electronic supplementary material. A complete list of all differentially expressed oligos (and BLAST hits for each oligo) is given in table S3 of the electronic supplementary material, and a subset of the ‘wasp foraging/provisioning sublist’ is shown in table 1.
(c) Overall patterns
HCA by oligo showed several different patterns in the expression data (figure 1). For example, there were clusters of genes that reflected group-specific patterns and both reproduction and foraging/provisioning-related patterns. In some clusters, gynes and foundresses showed similar patterns. Although gynes and foundresses have nothing in common with respect to foraging/provisioning or reproductive states, they both have limited social contact in comparison with queens and workers (i.e. gynes at this stage are not part of a dominance hierarchy, and foundresses are alone).
HCA by group (figure 1) showed a different topology from that of the previously published qPCR study on these same four groups (Toth et al. 2007). In the current study, workers and queens had the most similar expression patterns, in contrast to workers and foundresses having the most similar expression patterns in the earlier study. The previous study (Toth et al. 2007) involved only genes known to be associated with honeybee division of labour, whereas this study involved a broader set. In both studies, gynes had the most divergent expression patterns.
LDA resulted in three linear discriminants (LDs), which accounted for 41, 37 and 22 per cent of the variation in the data. A scatterplot of LD1 and LD2 (78% of the variation) lends further support to the result that gynes had the most divergent expression patterns (figure 2). In addition, the LDA further supported the finding from HCA that workers and queens have the most similar expression patterns. However, the LD scatterplot shows minimal overlap between all groups, even queens and workers, indicating each group has a distinct expression profile.
(d) Overlap with honeybee transcriptomic studies
We examined the extent of overlap with previous studies of honeybee brain gene expression related to foraging and reproduction. The number of wasp genes overlapping for each bee gene list is shown in figure 3a–c. Genes on the ‘wasp reproduction sublist’ showed no significant overlap with the ‘bee foraging lists’ (Whitfield et al. 2003; Alaux et al. 2009b; figure 3a,b, respectively). By contrast, based on both Fisher exact tests and simulations, we found significant overlap between the ‘wasp foraging/provisioning sublist’ and the ‘bee foraging list’ from Alaux et al. (2009b), and a marginally significant overlap between the ‘wasp foraging/provisioning sublist’ with the ‘bee foraging list’ from Whitfield et al. (2003). Table 1 gives a list of a subset of the genes that overlapped.
(e) GO functional analysis of gene lists
GO results for enriched ‘biological process’ categories are summarized in table 2 (and detailed results are presented in table S4 of the electronic supplementary material). For the ‘wasp foraging/provisioning sublist’, we found genes in three main biological processes were overrepresented: lipid metabolism, response to stress and heat, and behaviour (including response to other organisms and locomotory behaviour). For the ‘wasp reproduction sublist’, one main biological process was overrepresented—response to heat and chemical stimuli.
This paper provides, to our knowledge, the first large-scale transcriptomic analysis of brain gene expression in a social insect species other than the honeybee. We detected 389 genes (approx. 12% of genes on the array) that differed in brain expression across four female groups in a paper wasp society. Several different patterns of gene expression were found, some that represent group-specific patterns and others that correlate with differences in foraging/provisioning behaviour and reproductive state.
The extent of overlap between paper wasps and honeybees for genes related to foraging behaviour was higher than that predicted by chance, suggesting there may be common elements in the gene networks regulating division of labour in the two lineages, despite their independent origins of sociality. We also found numerous differences in gene expression between the two species, pointing to many lineage-specific mechanisms. This is not surprising, considering that bees and paper wasps are separated by over 100 million years of evolution. Overall, our results suggest a combination of conserved and unique mechanisms regulate social behaviour in these phylogenetically distant social insect species.
Results for foraging, but not reproduction, provide support for the idea of a genetic ‘toolkit’ for division of labour (Toth & Robinson 2007). Several of our analyses show that genes associated with worker foraging in honeybees are likely to be associated with foraging/provisioning behaviour in wasps. First, we found that 12 per cent of the genes on the microarray were differentially regulated across the four female groups, which contrasts sharply with the results of our previous (qPCR) study, in which 63 per cent of genes were differentially regulated (Toth et al. 2007). This large discrepancy suggests that Toth et al. (2007) indeed examined a select gene set; the genes were chosen based on their known associations with division of labour in honeybees. Second, for this select set, overall expression patterns in wasps were most similar in foundresses and workers, the two foraging/provisioning groups (Toth et al. 2007). By contrast, for all 389 differentially regulated genes from the current study, expression patterns were largely distinct for each group (figure 2). Thus, foraging/provisioning-related expression patterns in wasps were more prevalent for genes that relate to foraging in honeybees. Finally, meta-analysis of microarray data showed a higher than random overlap between genes related to wasp foraging/provisioning and honeybee foraging. These results demonstrate common patterns of gene expression associated with foraging behaviour in wasps and bees.
Genes for lipid metabolism, locomotory behaviour and response to heat stress were overrepresented on the list of genes regulated in association with wasp foraging/provisioning. Although the overlapping wasp and bee gene lists were too small for GO analysis, it is notable that several genes in these same categories are also differentially regulated in other species. Heat shock protein and metabolic genes are differentially expressed in several honeybee brain transcriptomic studies related to foraging (Whitfield et al. 2003, 2006; Ament et al. 2008; Alaux et al. 2009b). Moreover, similar genes have been identified in whole body gene expression analyses of other social insect species. Polistes canadensis wasp queens and workers were found to differ in expression of Vitellogenin, a heat shock protein, and numerous metabolic enzymes (Sumner et al. 2006). Graff et al. (2007) discovered queen–worker differences in Vitellogenin expression in the ant Lasius niger. Hoffman & Goodisman (2007) found differences in yellow jackets in Hexamerin storage proteins between queen and worker castes across different life stages. Synthesizing this information with the results of the current study, we suggest that storage proteins such as Vitellogenin and Hexamerin, genes related to stress and heat shock, and metabolic enzymes involved in lipid metabolism may have far-reaching effects on, or may be strongly affected by, caste differences across a wide variety of species and tissue types.
In contrast to the foraging results, we did not find significant overlap in brain gene expression patterns for wasp reproduction and honeybee queen–worker caste differences. These results suggest that different molecular mechanisms regulate reproductive behaviour in the two species. This is consistent with the fact that reproductive physiology differs fundamentally between these two species. In Polistes, ovary development and reproductive behaviours are regulated by the insect gonadotropin, juvenile hormone (JH) (Bohm 1972; Roseler et al. 1985; Sledge et al. 2004). In honeybees, JH levels are uncoupled from ovarian development and egg-laying behaviour in queens, a situation that may also characterize advanced eusocial stingless bees (Hartfelder et al. 2001), but not primitively eusocial bumble bees (Geva et al. 2005). Thus, it is possible that the genetic mechanisms regulating egg-laying behaviour in paper wasps may more closely resemble those of solitary ancestors, whereas a different suite of genes regulate adult reproduction in the derived social system of the honeybee. However, like any finding of this nature, it is also possible that this is a ‘false negative’ result. This can be addressed in the future when more genes are added to the P. metricus microarray after additional sequencing, or with the application of more powerful techniques than comparing lists of differentially expressed genes, e.g. inferring gene regulatory networks based on gene expression and transcription factor binding data (Ament et al. in press). Comparative studies of global brain gene regulation in additional social and solitary Hymenoptera will also help to resolve this issue.
The discrepancies we noted for foraging- versus reproduction-related gene expression are consistent with other recent findings. For example, a study with Cryptocercus termites found that only three out of 10 genes associated with queen–worker differences showed conserved expression patterns across two closely related species (Weil et al. 2009). By contrast, other studies of aggression and reward-seeking behaviour (Alaux et al. 2009b; Barron et al. 2009) suggest that there may be some common elements regulating these forms of behaviour across a wide range of animal taxa, from arthropods to vertebrates. Clearly, additional genomic-level studies on diverse species will be necessary to gain further insight into issues related to how pervasive genetic toolkits are for behaviour.
Our finding that 12 per cent of the genes on the microarray were differentially regulated across the four female Polistes groups contrasts with results from honeybee studies, in which different behaviourally defined groups of individuals differ for 25–40% of the genes on the microarray (Whitfield et al. 2003; Grozinger et al. 2007; Alaux et al. 2009b). This observation suggests that division of labour in primitively eusocial insects might be associated with more subtle differences in brain gene expression when compared with advanced eusocial insects. This is consistent with the fact that primitively eusocial individuals retain more behavioural totipotence than in advanced eusocial insects. Perhaps more changes in gene regulation are required for increased caste specialization. Alternatively, this finding might also change when more genes are added to the Polistes microarray after additional sequencing. Our results provide an important starting point for exploring whether differences in the number of differentially expressed genes are correlated with differences in the degree of behavioural specialization in social insects.
Some cross-species comparisons of gene expression (Jordan et al. 2005) suggest that variation in gene expression for some genes can be tolerated without having an effect on phenotype, in a form of ‘neutral variation’ in expression level. Our data show differential gene regulation between groups of the same population associated with behavioural differences, as inferred from the significant differences in gene expression. If this difference were owing to high levels of phenotype-neutral variance within the population, we would not expect these changes to be statistically significant between these groups. Since the groups were sampled from the same population; population structure, drift or fixation cannot account for these differences. In addition, we see a significantly higher than random overlap of genes related to foraging division of labour in two species, indicating a possible functional/mechanistic component for the observed differences. By contrast, the neutral hypothesis would predict the same extent of overlap for both reproduction- and foraging-related lists.
It remains to be seen which differences in gene expression cause differences in behaviour and which differences in gene expression occur as a result of changes in behaviour. If some are causal, it will be important to consider whether shared mechanisms are the result of convergence on the same mechanism or conservation of function across lineages (Toth & Robinson 2007). One possibility is that convergent phenotypes, such as division of labour for foraging, use convergent mechanisms. This would suggest that certain molecular pathways may be more evolutionarily labile and are recruited multiple times during the evolution of sociality to affect social behaviour. There is evidence for this phenomenon from evolutionary studies of development, e.g. homeotic developmental genes have been recruited to control various forms of morphological novelties in different vertebrate lineages (Carroll et al. 2005).
A second possibility is conservation—division of labour in different lineages evolved from a set of behavioural/physiological states shared by their common ancestor, and these behaviours evolved into foraging/non-foraging and reproductive/non-reproductive castes in different lineages. Several of the genes related to foraging behaviour in both bees and wasps (table 1) are also related to flight and locomotor behaviour in D. melanogaster, and the foraging gene has been implicated in the regulation of foraging in several different social insect species (Ben-Shahar et al. 2002; Ingram et al. 2005; Tobback et al. 2008; Lucas & Sokolowski 2009; present study). West-Eberhard (1996) proposed a similar idea based on ancestral reproductive characters, and additional solitary ‘groundplan’ hypotheses have recently received some experimental support (Amdam et al. 2004; Hunt et al. 2007). Further comparative genomic studies across different social and solitary lineages can help to distinguish between these two evolutionary scenarios and thus provide additional insights into the evolutionary processes by which complex behaviour can arise.
We would like to thank K. S. Pruiett, T. D. Daugherty, S. H. Woodard, J. H. Hunt and K. J. Bilof for assistance with fieldwork, T. Nguyen for performing array hybridizations, F. E. Miguez for statistical advice, M. R. Band for advice on microarrays, L. Qi for spotfinding microarrays and J. H. Hunt and members of the C. M. Grozinger laboratory for comments on the manuscript. This work was supported by NSF grant IOS-0641431 (Behavioural Systems Cluster, G.E.R.) and the Illinois Sociogenomics Initiative (G.E.R.). Microarray data have been deposited into ArrayExpress (http://www.ebi.ac.uk/microarray-as/ae/), ArrayExpress accession: E-MEXP-2558.
↵† Present address: Department of Ecology, Evolution and Organismal Biology, Iowa State University, Ames, IA, USA.
- Received January 14, 2010.
- Accepted February 23, 2010.
- © 2010 The Royal Society