Report

Wasp Gene Expression Supports an Evolutionary Link Between Maternal Behavior and Eusociality

See allHide authors and affiliations

Science  19 Oct 2007:
Vol. 318, Issue 5849, pp. 441-444
DOI: 10.1126/science.1146647

Abstract

The presence of workers that forgo reproduction and care for their siblings is a defining feature of eusociality and a major challenge for evolutionary theory. It has been proposed that worker behavior evolved from maternal care behavior. We explored this idea by studying gene expression in the primitively eusocial wasp Polistes metricus. Because little genomic information existed for this species, we used 454 sequencing to generate 391,157 brain complementary DNA reads, resulting in robust hits to 3017 genes from the honey bee genome, from which we identified and assayed orthologs of 32 honey bee behaviorally related genes. Wasp brain gene expression in workers was more similar to that in foundresses, which show maternal care, than to that in queens and gynes, which do not. Insulin-related genes were among the differentially regulated genes, suggesting that the evolution of eusociality involved major nutritional and reproductive pathways.

A major challenge in biology is to understand the evolution of animal society in molecular terms. Eusociality is the most extreme form of cooperation, typified by individuals that care for siblings rather than reproduce themselves, i.e., “workers.” The evolution of eusociality has been ascribed to kin or colony-level selection (1, 2), but these explanations do not specify mechanistic routes.

It has long been suggested (35) that sibling care by hymenopteran (ant, bee, wasp) workers evolved from maternal care, which involves provisioning brood by foraging for food and then feeding them. According to this idea, two principal behaviors exhibited by solitary Hymenoptera, reproduction (egg-laying) and maternal care (brood provisioning), became uncoupled during the early stages of social evolution (6), and these behaviors eventually occurred in separate castes, queens and workers, respectively (7). Linksvayer and Wade (8) added a molecular dimension to this idea by predicting that sibling care and maternal care behaviors should be regulated by similar patterns of gene expression.

We used Polistes paper wasps to test Linksvayer and Wade's idea. Polistes are primitively eusocial, which means that although individuals specialize as either workers or reproductive individuals, these two castes are less distinct than in advanced eusocial species. In Polistes, both workers and reproductives display provisioning behavior, but at different points in the life of a colony. Advanced eusocial insects, by contrast, have morphologically distinct queen and worker castes, and in some species, such as the honey bee, queens no longer exhibit any maternal care, which precludes comparing the molecular basis of sibling and maternal care. Primitively eusocial insects like Polistes afford the opportunity to explore the molecular basis of maternal and worker behavior within a single species.

We measured brain gene expression in 87 individuals from four distinct behavioral groups of females from naturally occurring colonies of the temperate species Polistes metricus (Fig. 1A). Foundresses are females that establish new colonies in the spring, often as solitary individuals. Foundresses exhibit both reproductive (egg-laying) and maternal (foraging and brood-feeding) behavior. After rearing a first generation of female brood that develop into workers, successful foundresses become queens and cease caring for brood. Workers take over provisioning the brood—their siblings—by foraging for food and then feeding them; workers show little, if any, reproductive behavior. By contrast, queens focus exclusively on reproductive behavior. Gynes are reared late in the season; they engage in no reproductive or maternal care behavior (9). After successfully mating, gynes overwinter and then become foundresses (10). We hypothesized that brain gene expression patterns in P. metricus workers and foundresses should be most similar to each other from among these four groups, because they both show brood provisioning behavior despite their different reproductive status. Alternatively, if brain gene expression more closely reflects reproductive behavior, expression in foundresses and queens should be most similar to each other.

Fig. 1.

P. metricus wasp brain gene expression analysis tests the prediction that maternal and worker (eusocial) behavior share a common molecular basis. (A) Similarities and differences in reproductive and brood provisioning status for the four behavioral groups analyzed in this study: foundresses (n = 22), gynes (n = 20), queens (n = 23), and workers (n = 22). Each individual wasp (total of 87) was assigned to a behavioral group on the basis of physiological measurements (14). (B to D) Results for 28 genes selected for their known involvement in worker (honey bee) behavior. (B) Heatmap of mean expression values by group and a summary of analysis of variance (ANOVA) results for each gene. Genes were clustered by K-means clustering (37); those in red showed significant differences (ANOVA, P < 0.05; table S1) between the behavioral groups. P. metricus gene names were assigned on the basis of orthology to honey bee genes (reference in parentheses); putative functions were assigned on the basis of similarity to Drosophila melanogaster genes. (C) Results of linear discriminant analysis show that foundress and worker brain profiles are more similar to each other than to the other groups. (D) Results of hierarchical clustering show the same result (based on group mean expression value for each gene). Four genes (PmVg, Pmg5sd, PmGlyP, and PmRfaBp) were excluded from these analyses because they showed high levels of expression in tissue adjacent to the brain (fig. S2); results for all three analyses were similar with and without these four genes (fig. S3).

Social behavior is a complex and polygenic trait, so an appropriate test of the idea that maternal and worker behavior share a common molecular basis requires analysis of multiple genes in different pathways. But Polistes wasps, though venerable models for studies of social evolution (11, 12), have until recently lacked genomic sequence information (13). To provide a ready source of test genes for quantitative reverse transcription–polymerase chain reaction analysis, we used 454 sequencing to obtain 45 megabases (Mb) in 391,157 cDNA sequence fragments from the P. metricus brain transcriptome (14). We were interested to see whether this low-cost, high-throughput sequencing method would be successful for this purpose, despite short sequence read lengths (average of 120 bp) and an estimated 100- to 150-million-year divergence time between P. metricus and the honey bee, Apis mellifera (15), the most closely related species with a sequenced genome to use as reference (16).

We generated a map of the honey bee genome combined with known transcripts and their relative abundance in the combined bee expressed sequence tag (EST) data sets (1618). P. metricus transcript fragments predicted to encode proteins orthologous to those encoded by A. mellifera genes were plotted on the map according to the number of fragments identified for a particular locus; matches were found for 39% of all honey bee mRNAs. The relative abundance of P. metricus sequence fragments corresponded well with the abundance of A. mellifera ESTs for the respective loci (Fig. 2). The combined P. metricusA. mellifera transcriptome data set was then used to select the genes for this study.

Fig. 2.

A representation of P. metricus brain transcripts overlaid on a honey bee genome template (16) shows wide coverage and similar transcript abundance for P. metricus relative to known honey bee transcripts. P. metricus brain cDNA sequence fragments were matched as predicted proteins to A. mellifera transcripts with experimental support (known cDNA or EST sequences). A. mellifera transcripts (red points, right of axis) and their closest P. metricus orthologs from our survey (blue points, left of axis) were then mapped to the corresponding genomic locus in the A. mellifera genome. The vertical lines represent A. mellifera chromosomes 1 to 16. The distance of each point from the midline is proportional to the logarithm of the abundance of the mRNA (the number of sequences for each P. metricus or A. mellifera transcript corresponding to the A. mellifera gene at that locus) (1618). P. metricus orthologs were obtained for a total of 3017 A. mellifera transcripts. The P. metricus transcriptome data contained putative orthologs for 39% of known A. mellifera mRNAs. An additional 252,556 transcript sequence fragments obtained from P. metricus did not have a clearly orthologous transcript in A. mellifera.

Prior information allowed us to focus on genes implicated in honey bee foraging and provisioning behavior, rather than a set of randomly chosen genes that might be less informative. We selected 32 genes (Fig. 1B and table S1) from the P. metricus EST set that are orthologs of A. mellifera genes known to be associated in some way with worker bee behavior, based on results from studies with microarrays (22 genes) (19, 20) and candidate genes (10 genes) (2129). Twenty-two of the genes have been shown by microarray analysis to be both differentially expressed in the brains of honey bees engaged in foraging or feeding brood [on the list of the “top 100” genes most consistently associated with bee foraging behavior (19)] and regulated by juvenile hormone (20), which also causes worker bee foraging behavior (30). Five candidate genes are differentially expressed in honey bees engaged in foraging or feeding brood (2124, 29), three of which also have been shown to play causal roles in the regulation of worker bee foraging behavior (21, 22, 31). Five additional candidate genes involved in insulin signaling were selected because this pathway is implicated in honey bee queen-worker caste determination (25, 26, 32) and worker foraging behavior (27, 28). Patterns of gene expression in P. metricus were not used as criteria for gene selection.

There was a robust association between individual wasp brain gene expression and naturally occurring behavioral differences among the wasp groups. Leave-one-out cross-validation analysis (19) resulted in 68, 69, 70, and 47% correct assignments to the foundress, gyne, queen, and worker groups, respectively. For the less conservative resubstitution method (33), the results were 89, 100, 100, and 95%. The predictions from both classification methods were significantly better than random (Chi-square tests, P < 0.0001, 25% expected). This honey bee–derived gene set thus demonstrates extensive brain regulation across the four wasp groups, making it an informative set to explore the molecular relationship between maternal and worker behavior in P. metricus.

Sixty-two percent of the genes in the gene set were differentially regulated in P. metricus as a function of reproductive or provisioning behavior (Fig. 1B and table S1). Multivariate analysis of variance showed that brain gene expression varied significantly with reproduction (F = 3.28, P = 0.0002) and provisioning (F = 4.76, P < 0.0001), with a significant provisioning × reproduction interaction (F = 2.48, P = 0.002). Three out of the five insulin-related genes showed significant associations with provisioning and/or reproductive behavior, consistent with known nutritional effects on behavior and physiology in honey bees and other social insects (34).

Three statistical analyses demonstrated that brain gene expression for worker wasps was more similar to that of maternal females (foundresses) than to that of females not showing maternal care (queens and gynes). First, K-means clustering (Fig. 1B and fig. S1) revealed five clusters of coexpressed genes. The first cluster contained genes (n = 8) that showed coexpression in foundresses and workers compared to queens and gynes. The second cluster of genes (n = 7) was mainly characterized by up-regulation in gynes, and the third (n = 6) by down-regulation in queens, but in both of these clusters, foundresses and workers also showed patterns of expression that were similar to each other (Fig. 1B and fig. S1).

The second statistical analysis, linear discriminant (LD) analysis, also showed similarities between foundress and worker brain gene expression (Fig. 1C). A plot of LD1 versus LD2 (which accounted for 90% of the variation in brain gene expression across all four groups) revealed group-specific expression patterns, but foundresses and workers showed the greatest overlap. This is consistent with the poorer performance of classification methods (described above) for those two groups; overlap in gene expression patterns made them difficult to distinguish from each other. Gynes, which engage in neither reproductive nor provisioning behavior, were the most distinct group. The third statistical analysis, hierarchical clustering by group, supported the patterns found in the other two analyses—brain gene expression of workers and foundresses was most similar, and that of gynes was most divergent (Fig. 1D).

There are marked temporal changes in brain gene expression as females shift from foundress to queen status, i.e., from maternal to reproductive behavior. These findings demonstrate heterochronic expression of genes associated with maternal behavior, a form of plasticity that is considered to be necessary for the evolution of worker behavior (8). They also reflect the apparent modularity of egg-laying and brood provisioning behavior and their underlying regulatory networks; this type of modularity also is thought to be important in the evolution of novel traits (35).

We used the honey bee genome, together with “next-generation” sequencing technology, to rapidly bring genomics to the relatively closely related wasp P. metricus; this is an early example of the utility of 454 sequencing for transcriptomics (36). Our results demonstrate that it is possible to use species that have had their genomes sequenced as “hubs” to efficiently generate genomic resources for clusters of related species that might each be especially well suited to address particular evolutionary problems. This “hub and spokes” approach should enable genomics to be deployed for a broader range of species than is currently being done, until whole-genome sequencing of eukaryote genomes becomes routine.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1146647/DC1

Materials and Methods

Figs. S1 to S3

Tables S1 and S2

References

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article