Results

Sequencing and transcriptome assembly

In total, we obtained over 787 million high-quality paired-end 150-bp reads, with an average of 39.4 million reads per tissue type (Supplemental Table 1; accession numbers will become available upon publication). After filtering, we retained over 766 million reads (Supplemental Table 2). An average of 88.4% of these reads mapped to the N. lecontei genome. Our genome-guided de novotranscriptome assembly identified 132,243 contigs. After further filtering, we retained 58,353 contigs. Across our 77 libraries, average mapping rate to these contigs was 85.5% (range: 74.9-91.6%). The provisionally annotated transcriptome contained 16,714 contigs representing 9,304 predicted insect genes. Using these 16,714 contigs as the transcriptome, a high mapping rate was maintained across all libraries (mean: 76.7%; range: 60.5-88.6%).
Decoupling of gene-expression profiles increases with ecological dissimilarity of N. lecontei life stages
Consistent with our prediction that differences in gene-expression traits would be most pronounced between ecologically dissimilar life stages (Figure 1B), stages separated by major and complete metamorphosis were clearly distinct as shown in the first principle component (PC1) of our principle component analysis (Figure 2A). By contrast, there were no clear distinctions between cryptic and aposematic larvae or between males and females along either of the first two gene-expression PCs. PC2 primarily separated tissues within each life stage (Figure 2A). Along this axis, gene expression in larval and adult heads was clearly distinct from expression in other tissues, as was expression in the adult antennal tissues. Based on these results, we analyzed gene-expression decoupling among stages/sexes in two ways: all tissues combined and heads only (Supplemental Tables 3 and 4, respectively).
Regardless of whether we looked at all tissues or heads only, both the magnitude of differential expression (log fold-change) and significance (Benjamini-Hochberg adjusted p-value) increased in accordance with ADH predictions across metamorphic transitions (Figure 2B for all tissues). Specifically, the percentage of DEGs increased with ecological dissimilarity of the life stages compared: 2.2% of genes were differentially expressed between cryptic and aposematic larvae, 24.7% between aposematic and dispersing larvae, and 25.1% between dispersing larvae and adult males (Supplemental Table 3; heads only: 2.9%, 17.6%, and 31.1%, respectively, Supplemental Table 4). The proportion of DEGs increased concurrently with the extremity of metamorphosis (Fisher’s exact tests; P < 1 x 10-20) except for the all-tissue dispersing larvae vs. male comparison (P =0.59).
Despite males and females being more morphologically dissimilar than the feeding and dispersing larval stages, the percentage of DEGs between sexually dimorphic adults (all tissues: 7.0%, Supplemental Table 3; head only: 6.5%, Supplemental Table 4) was far less than that observed across both the major and complete metamorphic transitions (Fisher’s exact tests; P < 1 x 10-208)(Figure 2B). In contrast, we observed a higher percentage of DEGs between the sexes than between the two feeding larval stages (Fisher’s exact tests;P < 1.0 x 10-20). Taken together with the PCA, these results indicate that gene-expression decoupling between the sexes is more or less on par with the minor metamorphic transition between cryptic and aposematic larvae and far less pronounced than the decoupling observed between more extreme life-stage transitions. Overall, these transcriptome-wide analyses support the ADH prediction that genetic decoupling increases as the ecological demands of the different life stages become more dissimilar, as well as the prediction that genetic decoupling tends to be more pronounced between developmental stages than between the sexes.

Variation in decoupling among different types of gene-expression traits reflects changes in ecology

Consistent with our prediction that genes that mediate changing ecological interactions will exhibit the most pronounced decoupling, many of the top differentially expressed genes between life stages/sexes corresponded to candidate genes that were related to their ecological differences (Supplemental Table 5). Among the top differentially expressed genes between cryptic and aposematic larvae were genes potentially involved in immunity, metabolism and detoxification. For example, esterase FE4-like (elevated in bodies of aposematic feeding larvae), is a gene thought to be involved in resistance to organophosphate insecticides and may be essential to ingesting and sequestering toxic pine resins. The esterase FE4-like gene was also among the top genes downregulated in the non-feeding dispersing larvae. The change from an aposematic feeding larva to a dispersing larva was also accompanied by a pronounced decrease in expression ofCameo2 , a gene that is thought to play a role in carotenoid-based pigmentation (Y. Li et al., 2014) and has been linked to larval body color in N. lecontei via QTL mapping (Linnen et al., 2018). Putative pigmentation and detoxification genes were also among the top differentially expressed genes between larvae and adult males (Supplemental Table 5).
The top differentially expressed genes between males and females for each tissue appear to be related to the sex-specific tasks of the adults (Supplemental Table 5). For example, females, which must properly provision their eggs, had elevated expression for avitellogenin-like gene, which is associated with egg yolk formation, hormone regulation, lifespan, and foraging behavior (Ihle, Page, Frederick, Fondrk, & Amdam, 2010; Munch, Ihle, Salmela, & Amdam, 2015; Nunes, Ihle, Mutti, Simoes, & Amdam, 2013; Seehuus, Norberg, Gimsa, Krekling, & Amdam, 2006; Wheeler, Ament, Rodriguez-Zas, & Robinson, 2013) There was also increased expression for many major royal jelly protein family genes, known to play a role reproductive maturation and tending to young (Buttstedt, Moritz, & Erler, 2013; Dobritzsch, Aumer, Fuszard, Erler, & Buttstedt, 2019; Drapeau, Albert, Kucharski, Prusko, & Maleszka, 2006). Among the genes with unusually high expression in males were three chemosensory genes (OR54 ,OR22 , and OBP9 ) that were particularly high in the antennae and may play a role in mate finding.

Chemosensory genes vary in decoupling across life stages

Consistent with our a priori predictions, the magnitude of decoupling for manually curated chemosensory genes was highly variable (Figure 3A). While several chemosensory genes had similar levels of expression across life stages (e.g. those falling along the dotted lines), others were among the most decoupled genes in the transcriptome (e.g., those falling at the edges of the transcriptome-wide cloud of points). This contrasted with a family of housekeeping genes that have a similar family size—the ribosomal protein L genes (RPLs). As expected, these genes were highly coupled across development and had nearly identical expression levels in comparisons between the all sexes or stages (e.g., those that fall on the dotted line). Figure 3B directly compares variation in decoupling for these two categories of genes. Consistent with the ADH, gene-expression decoupling for chemosensory genes was both more variable (i.e., wider distribution) and significantly higher than RPLs in 3 out of 4 comparisons (Figure 3B and Supplemental Table 4; Mann-Whitney U test P < 1.0 x 10-3). The exception was the aposematic vs. dispersing larvae comparison representing the major metamorphic transition (P = 0.90). Overall, these results are consistent with a scenario in which selection has favored coupling for some chemosensory genes and stage-specific decoupling for others.
To gain additional insight into how chemosensory gene expression changes across the life cycle, we created a heat map of stage-specific expression for each gene (Figure 3C). Whereas some chemosensory genes were expressed only during a single life stage, others were expressed at moderate to high levels in multiple life stages. We further analyzed these genes by identifying those that were in the highest 10% of expression for each stage/sex. This analysis revealed several chemosensory genes that were highly expressed in a single stage/sex while others were highly expressed in multiple stages and/or both sexes (Figure 3D and Supplemental Table 6). Together, these patterns provide clues into chemosensory gene function. For example, chemosensory genes that were highly expressed by both feeding larvae and females (OBP5 and OR25 ) could be involved in detecting host-plant cues. Likewise, genes highly expressed in both sexes of reproductive adults (OBP2 , OBP8 , and OBP10 ) may play roles in conspecific mate recognition. Genes expressed by all three stages (OBP4 , OBP9 , and OR55 ) may be important for detecting conspecific or host cues that are critically important throughout the life span. In contrast, genes with high expression in only a single stage may only be essential for a single task or ecological pressure. For example, OBP13 is expressed at high levels in feeding larvae with low levels of expression in dispersing larvae an almost no expression in adults. Each adult sex also has specific chemosensory genes with OR20 and OR41 expressed almost exclusively in female antennae while OR22 , OR51 , and OR54 expressed almost exclusively in male antennae. Notably, although all larvae were males, we did not identify any genes that were highly expressed in larval and adult males to the exclusion of adult females, suggesting once again that ecology is a better predictor of gene-expression patterns than sex in this species.