Determinants of adult worker honey bee behavioral maturation and their correlated effects on brain gene expression. (A) Bees switch from hive work (primarily nursing brood) to foraging outside for nectar and pollen. We studied the regulation of new and previously reported brain transcriptomic differences between nurses and foragers and responses to 10 determinants of the timing of the age at onset of foraging (+ stimulating or − inhibiting foraging onset; e.g., rich diet delays onset of foraging). Previous studies also have characterized stimulatory (+) and inhibitory (−) relationships among some of the maturation determinants themselves. For genetic comparisons (AvE, LvM, and PH), the first genotype shows faster maturation than the second one. AvE, Africanized vs. European subspecies; BP, brood pheromone; cGMP: cyclic-guanosine monophosphate treatment; diet, rich vs. poor diet; JHA, juvenile hormone analog treatment; LvM, Northern (A. mellifera mellifera) vs. Southern European (A. mellifera ligustica) subspecies; Mat, maturation (nurses vs. foragers); Mn, manganese treatment; PH, high vs. low pollen-hoarding genetic strains; QMP, Queen Mandibular Pheromone; Vg, vitellogenin RNAi (details in Table S1). (B) Pairs of maturation determinants with effects on gene expression that were correlated in a direction consistent with their shared effects on behavior (green) or opposite to their effects on behavior (red) across most (solid colors) or relatively few (stripes) bootstrap gene sets.

Meta-associations between GO terms, cis-regulatory motifs, and genotypic, environmental, and physiological determinants of honey bee behavioral maturation. (A) Meta-associations reported by Metalysis between maturation determinants and GO biological processes, with EVD P value ≤ 0.05. The EVD P value for each meta-association is based on comparing its meta-P value to the smallest meta-P values returned from each of many randomization tests (negative controls). Orange and blue cells indicate the individual maturation determinants contributing to a meta-association; orange and blue denote associations with genes up-regulated in the fast maturation and slow maturation conditions, respectively. (B) Meta-associations (at EVD P value ≤ 0.01) revealed by single motif cis-Metalysis (Upper) and subjection of SAMBA biclusters to single motif enrichment tests (Lower). No significant meta-associations were discovered using BiMax biclusters. Asterisks denote motifs reported by cis-Metalysis (at EVD P value ≤ 0.01) when configured to search only unidirectional meta-associations. (C, Top) The best meta-associations (red ticks) between cis-regulatory motifs and determinants were much stronger than the best found in many randomized datasets (blue ticks; best-fit γ-distribution shown as the blue curve). Each blue tick represents the strongest meta-association in a distinct (randomized) dataset. Same plots for the results from BiMax (C, Middle) and SAMBA (C, Bottom) biclusters subjected to motif enrichment tests.

Metalysis and cis-Metalysis reveal shared transcriptional regulatory mechanisms for determinants of honey bee behavioral maturation. (A) The cis-Metalysis procedure. Step 1: Scan promoters of genes (G_i) for conserved cis-regulatory motifs (M_i). Step 2: Compute statistical associations between motif presence and differential gene expression across multiple conditions (C_i). Step 3: Combine P values to characterize meta-associations across some or all conditions. (B) Combinations of cis-regulatory motifs shared by 11 determinants of fast vs. slow honey bee behavioral maturation (from hive work to foraging). A wiring diagram shows the most significant combinatorial rules involving motifs recognized by the transcription factors mtTF-A and USF. The roles of individual motifs differed between conditions. “And” and “not” logic operations are depicted by rounded rectangles and triangles, respectively.

Informatics techniques for the discovery of meta-associations across transcriptomic experiments. (A) Flowchart for cis-Metalysis. Steps 1 and 2: Motif modules are defined based on the presence (green cells) of single motifs (e.g., M₁) or their Boolean combinations (e.g., M₁ and M₂) in the 5-kb promoter sequences of each gene. Steps 3 and 4: Sets of up- (orange) and down- (blue) regulated genes are identified from experimentally profiling gene expression for each determinant. Step 5: Statistical enrichments are tested between motif modules and expression sets producing all of the association P values (shaded cells) that are combined by cis-Metalysis. In its most flexible mode (shown), cis-Metalysis combines the best P value (bordered in yellow) from each determinant when considering all combinations of motifs. Step 6: Meta-P value calculation. Top shows the best association P value from each condition (D₁ … D₄). For each K = 1 … 4, the statistic ϕK combining the best K P values is computed and translated to a P value P(ϕK) in Middle. The minimum P(ϕK) over all K is the meta-P value (highlighted in yellow within the red border) and considers the number and strength of the combined P values. The meta-association is represented in Bottom with selected significant conditions colored and the remaining conditions in gray. (B) Multiple modes of cis-Metalysis enable discovery of simple or combinatorial identical or plastic forms of regulatory logic shared across transcriptomic states (conditions) by selecting specific subsets of association P values to combine. For each mode, an example subset of associations (shaded cells) and the selected best P values (bordered in yellow) are depicted. In single motif cis-Metalysis, P values are selected from the association tests with the gene modules defined by a single motif (M₁). In the identical logic configuration, a rigidly defined combination of two motifs (!M₁ and M₂ in this example) defines the gene module that is tested for association with a fixed direction of regulation (down). In role-consistent and flexible cis-Metalysis, a meta-association of a motif pair (M₁ and M₂) is allowed to use different Boolean combinations of the two motifs and different directions of regulation in different determinants. In role-consistent mode, each motif–expression association must use a particular motif in the same role (activator or repressor); here, M₁ is an effective activator, and M₂ is a repressor. This constraint is relaxed in flexible cis-Metalysis (example in A, Step 5).

Combinatorial regulation of maturation-related gene expression. (A) Top three most significant meta-associations between maturation determinants and pairs of motifs that interact with role-consistent logic. Motif module–gene expression associations with fast maturation and slow maturation genes in each experiment are denoted by orange and blue, respectively. Role-consistent meta-associations of motifs with fast and slow maturation are indicated by green and red, respectively. For example, the third row indicates that the presence of the Dl motif and absence of the ETS motif are together associated with fast maturation genes in the QMP, Mat, BP, AvE, LvM, and JHA experiments, whereas the presence of the ETS motif and absence of the Dl motif are associated with slow maturation genes in the Mn, diet, Vg, and cGMP experiments. (B) Combinations of motifs recognized by the TFs Dl, Xbp1, and ETS are associated with fast and slow maturation, respectively, and together, they predict gene expression patterns across all 11 maturation determinants. A wiring diagram shows the most significant combination of motifs for each determinant. “And” and “not” logic gates are depicted by rounded rectangles and triangles, respectively. (C) Combinations of motifs recognized by mtTFA and USF are associated with the responses to all 11 maturation determinants, with each motif bidirectionally associated with up- and down-regulated genes in different experiments. Same notation as in B.