Transcriptomic basis and evolution of the ant nurse-larval social interactome

Abstract

Development is often strongly regulated by interactions among close relatives, but the underlying molecular mechanisms are largely unknown. In eusocial insects, interactions between caregiving worker nurses and larvae regulate larval development and resultant adult phenotypes. Here, we begin to characterize the social interactome regulating ant larval development by collecting and sequencing the transcriptomes of interacting nurses and larvae across time. We find that the majority of nurse and larval transcriptomes exhibit parallel expression dynamics across larval development. We leverage this widespread nurse-larva gene co-expression to infer putative social gene regulatory networks acting between nurses and larvae. Genes with the strongest inferred social effects tend to be peripheral elements of within-tissue regulatory networks and are often known to encode secreted proteins. This includes interesting candidates such as the nurse-expressed giant-lens, which may influence larval epidermal growth factor signaling, a pathway known to influence various aspects of insect development. Finally, we find that genes with the strongest signatures of social regulation tend to experience relaxed selective constraint and are evolutionarily young. Overall, our study provides a first glimpse into the molecular and evolutionary features of the social mechanisms that regulate all aspects of social life.

Fig 1. Social regulation of gene expression between ant nurses and larvae.

(A) Cartoon depicting positive gene regulation (i.e. activation) between larvae and nurses, where gene 1 is expressed in nurses and genes 2 and 3 are expressed in larvae. After the expression of gene 1 increases, the expression of gene 2 increases as a result of the social interaction of nursing (depicted in [B]). This can occur if gene 1 itself codes for a protein passed to larvae, if the mRNA transcript is passed directly, or if gene 1 activates the expression of some other gene in nurses, which in turn is passed as mRNA (or codes for a protein that is passed) to larvae. Following the increase in expression of gene 2, the expression of gene 3, which is shown to be activated by gene 2, also increases. While we have depicted a time-lag in this social regulation of gene expression, the time lags are likely too short to observe in our data, as larvae were collected every 3–4 days across development. Therefore, correlated transcriptome dynamics over development (see Fig 2) would reflect mechanisms shown here. (B) Gene regulatory networks act between and within individuals engaged in social interactions. Blue boxes are genes expressed in larvae, and red boxes are genes expressed in nurses. Solid lines depict regulatory interactions within tissues (here, within larvae or within nurses), while dashed lines represent social connections (nurse-larva or vice versa).

Fig 2. Nurse and larval transcriptomes show strong signatures of gene co-expression across larval development.

Plots (A-D) depict the expression profiles of individual genes (light lines) as expressed in (A) nurse head, and (B) nurse abdomens, as well as (C) larvae, shared with nurse heads, and (D) larvae, shared with nurse abdomens. Dark lines indicate the median expression values of all genes sorted into modules, with pre-defined expression profiles of modules depicted in plot insets. Colors indicate the pre-defined expression profile (i.e. module) that genes have been sorted into. Only the five shared modules containing the most nurse-expressed genes are shown for clarity. Larval expression profiles are divided by the nurse tissue they are shared with, such that (C) depicts larval gene expression shared with nurse heads (A), while (D) depicts larval gene expression shared with nurse abdomens (B). Note that nurse heads and larvae shared inversely-related expression profiles, and that this algorithm does not reveal the direction of regulation as it is simply correlation-based. (E) Stage-specific nurses have more genes than random nurses in modules shared with larvae than do random nurses, reflecting more broad-scale co-expression across development. “Connection type” refers to the tissue that the number of genes was calculated in (i.e. larva → nurse head indicates the number of genes expressed in larvae that are in modules shared with nurse heads), though directionality is not determined in this algorithm. Error bars indicate 95% confidence intervals derived from systematic drop-1 jackknifing of nurse samples. N = 10944 genes total.

Fig 3. Genes encoding secreted proteins such as giant-lens are important for social gene regulation.

(A) Genes encoding for proteins that are secreted in Drosophila melanogaster exhibit higher social connectivity (i.e. more strongly socially regulate larval expression) in nurse heads than genes encoding for non-secreted proteins (P-values from Wilcoxon test). (B) The protein giant-lens is one of the genes coding for secreted proteins with the highest social connectivity in nurse heads. Based on our data, giant-lens expressed in stage-specific nurse heads (red) appears to inhibit the expression of the homolog of human EGFR substrate 8 (eps8) expressed in worker-destined larvae (blue). The expression of giant-lens in nurses of a given colony was negatively correlated to the expression of eps8 in larvae of the same sampled colony (rho = -0.270, P < 0.001, N = 25 colony/stage pairings after removing missing samples). Expression at stage i is equal to log₂(expression_i/expression₁), i.e. the ratio of expression at the given stage to expression at L1.

Fig 4. Highly social genes tend to be less evolutionarily constrained.

Selective constraint, estimated from whole-genome polymorphism data, is (A) positively correlated with within-tissue connectivity (Spearman correlation; head: rho = 0.122, P < 0.001; abdomen: rho = 0.217, P < 0.001), but negatively correlated with (B) social connectivity (head: rho = -0.090, P = 0.009; abdomen: rho = -0.150, P < 0.001) and (C) sociality index (head: rho = -0.132, P < 0.001; abdomen: rho = -0.223, P < 0.001), where sociality index is the difference between social and within-tissue connectivity per gene. Each point in (A-C) indicates a single gene, as expressed in nurse heads or abdomens. Lines are trendlines from linear model. (D) Sociality index differs according to estimated evolutionary age (GLM; LRT; χ² = 57.357, P < 0.001), as ancient genes tended to have lower sociality indices than all other categories (Tukey’s post-hoc test; ancient—insect: P < 0.001, ancient—hymenoptera: P < 0.001, ancient—ant: P < 0.001, all other comparisons P > 0.05). Individual points depict average values across nurse heads and abdomens for all genes within each estimated evolutionary age class, indicated by labels on points. Error bars depict 95% confidence intervals from bootstrapping. Numbers in parentheses indicate number of genes in each age class.