Discussion
Our study suggests that direct genetic effects (i.e., SB/SBversus SB/Sb genotype) have a greater influence on patterns of gene expression than indirect genetic effects (i.e., being raised in a monogyne social environment containing only SB/SB workers or a polygyne social environment containing both SB/SB and SB/Sbworkers). This contrasts with two earlier studies, which concluded (i) that fire ant worker gene expression profiles (whole body) are more strongly influenced by indirect genetic effects associated with differences in the genetic composition of monogyne and polygyne colonies than by the direct genetic effects of their own supergene genotype (Wang et al., 2008) and (ii) that, in worker antennae, the number of genes differentially expressed in response to direct genetic effects is similar to the number differentially expressed in response to indirect genetic effects (Dang et al., 2019). These contrasting results may stem from differences in experimental design and statistical approach. In our experiments, we used body parts of single individuals (6-8 replicates per tissue/caste/genotype/social environment) (Arsenault et al., 2020). By contrast, Wang et al. (2008) used pools of 7-10 individuals for each of the 20 replicates and Dang et al. (2019) pools of 36-55 individuals for each of the 4 replicates. Because our analyses showed that indirect genetic effects have a lower effect on the magnitude of changes in gene expression than direct genetic effects, it is possible that using pools of individuals (i.e., because of lower variance) or more replicates increased the likelihood of detecting genes influenced by indirect genetic effects compared to direct genetic effects. The possibility that the variance in the indirect genetic effects experienced by different individuals is higher than the variance in direct genetic effects experienced by different individuals is an intriguing hypothesis that warrants further study.
Whether direct or indirect genetic effects more greatly impact transcription is a multidimensional question: the degree of difference in expression and the number of affected loci are both relevant. By some metrics we do observe a greater number of genes undergoing expression changes in response to indirect than direct genetic effects (worker gaster and queen ovaries and combined sample GLM, each at FDR < 0.1 but not more stringent cutoffs), offering apparent agreement with prior findings (Wang et al., 2008). However, our study suggests general agreement between tissue types and castes on the following: gene expression ratios are generally larger for direct than indirect genetic effects and indirect genetic effects on transcription appear more tissue- and caste-specific than direct effects. Moreover, we detected no significantly differentially expressed genes in the brains of workers in response to indirect genetic effects of the supergene, despite the presence of many differentially expressed genes arising from direct genetic effects with the same statistical power. This stands in contrast to the prediction of pervasive indirect genetic effects stemming from prior results from whole workers in S. invicta(Wang et al., 2008) and is more in line with results from a recent study of indirect genetic effects on brains in the clonal raider ant (Kay et al., 2022a).
The overrepresentation in the supergene region of genes differentially expressed due to direct genetic effects, but not indirect genetic effects, is consistent with cis -regulatory effects of supergene evolution and/or differences in chromatin accessibility between theSB and Sb haplotypes. Prior analysis of the gyne data alone also pointed to such effects (Arsenault et al., 2020). Evolutionary divergence of supergene-encoded proteins (Martinez-Ruiz et al., 2020; Pracana et al., 2017a; Pracana, Priyam, Levantis, Nichols, & Wurm, 2017b) likely also make trans- regulatory contributions to the observed direct genetic effects, but these are not expected to disproportionately impact expression within the supergene (Signor & Nuzhdin, 2018). Moreover, a large-scale analysis of multiple tissue types in humans revealed that cis- regulatory sequence variants often influence transcription in a constitutive fashion across tissues (Ardlie et al., 2015). We found that direct genetic effects on gene expression patterns were often consistent across tissues, offering apparent support to the hypothesized cis -regulatory impact of evolution of the fire ant social supergene.
It has previously been noted that differentially expressed genes arising from direct genetic effects (Arsenault et al., 2020) and combined direct and indirect genetic effects (Martinez-Ruiz et al., 2020) of the fire ant supergene are overwhelmingly more highly expressed in SB/Sbthan SB/SB individuals. Remarkably, this holds true for genes residing outside the supergene region, suggesting thattrans- mediated gene upregulation is caused by the inversion-derived Sb haplotype (Arsenault et al., 2020; Martinez-Ruiz et al., 2020). The genetic change(s) responsible for widespread gene upregulation in response to the Sb haplotype remains unknown, but effects of cis- regulatory evolution and gene duplication (Fontana et al., 2020) on trans -regulatory factors could be of relevance. Remarkably, we also observed more gene upregulation than expected by chance due to indirect genetic effects of the Sb haplotype – that is, higher expression of differentially expressed genes in SB/SB individuals cohabitating withSB/Sb individuals than those cohabitating with only SB/SBindividuals. This suggests the Sb haplotype may lead to gene upregulation through both direct and indirect genetic effects, despite a largely non-overlapping sets of genes being influenced.
The lack of significant gene expression changes associated with cross-fostering SB/SB worker pupae from a monogyne to a polygyne social environment until two weeks post-eclosion suggests thatSB/SB worker tolerance of multiple SB/Sb queens may not require a transcriptional response (at least not at a scale detectable with our methods). A possible explanation for this finding is that the presence of SB/Sb workers may induce passive habituation ofSB/SB workers to the Sb chemical greenbeard and thereby induce them to accept SB/Sb queens. Such a behavioral change without a detected change in brain gene expression is consistent with findings from a study of variation in aggressive behavioral response to social cues in honey bees (Rittschof, 2017) and a study of indirect genetic effects on brain gene expression in the clonal raider ant (Kay et al., 2022a). Thus, a growing body of evidence suggests that context-specific behavioral variation can arise from variation in the social environment without requiring fundamental changes to gene regulation at a detectable level in whole brains. Nevertheless, our results do suggest that indirect genetic effects accrue through the full course of development in S. invicta (when the social environment is consistent from egg to two-weeks as adults), as detected in abdominal tissues. Establishing a more fine-scale time course of the social supergene’s effects on development, both direct and indirect, remains an outstanding problem.
Queens of S. invicta face a unique selective pressure during reproductive development in polygyne colonies, which likely explains why their ovarian tissues exhibit the greatest observed indirect genetic effects. As they mature in polygyne colonies, SB/SB gynes become targets of spiteful greenbeard-mediated elimination by Sb -bearing workers (Gotzek & Ross, 2008; Keller & Ross, 1998; Trible & Ross, 2016). The SB/SB gynes that survive in polygyne colonies until mating flights have been found to be slightly, but significantly, lighter than SB/SB gynes embarking on flights from monogyne colonies (DeHeer et al., 1999; Keller & Ross, 1995), reducing their independent colony-founding potential (DeHeer, 2002). The differential expression between the SB/SB gynes of monogyne and polygyne colonies could be associated with morphological and/or physiological differences associated with these weight differences (Arsenault et al., 2020).
Some of the conclusions we reached are similar to those of a genome wide association study (GWAS) of direct and indirect genetic effects conducted in laboratory mice (Baud et al., 2021) despite the very different approach of mapping trait variation as opposed to transcriptional variation. Effect sizes of loci associated with direct genetic effects on trait variation were found to be higher than effect sizes of loci associated with indirect genetic effects on trait variation (Baud et al., 2021), mirroring the larger gene expression ratios we observed for direct genetic effects than for indirect genetic effects. In contrast to this GWAS study (Baud et al., 2021), we observed a modest but significant overlap in the genes with significant indirect and direct genetic effects on transcription in S. invicta .
Among the genes that exhibit a significant transcriptional response to both direct and indirect genetic effects are a zinc-finger domain-containing transcription factor, which represents a candidate trans-regulatory protein that may contribute to the widespread direct genetic effects observed outside the supergene region in S. invicta (Arsenault et al., 2020; Martinez-Ruiz et al., 2020). Also responsive to direct and indirect effects is limulus clotting factor C , which encodes an enzyme linked to antibacterial hemolymph clotting (Iwanaga, 1993; Iwanaga, Miyata, Tokunaga, & Muta, 1992) and, notably, three odorant receptor genes. Odorant receptors play an important role in chemical communication and therefore in social organization. There are also many documented amino acid substitutions between odorant receptor proteins encoded by the SB and Sbhaplotypes of the fire ant supergene (Cohanim, Amsalem, Saad, Shoemaker, & Privman, 2018).
As in other social animals (Avalos et al., 2020; Baud et al., 2021; Gempe et al., 2012; Linksvayer, 2006; Vojvodic et al., 2015), direct and indirect genetic effects are pronounced in S. invicta . By examining direct and indirect genetic effects of the fire ant supergene on gene regulatory dynamics in samples from distinct tissues and castes we show that direct genetic effects on transcription are generally more detectable than indirect genetic effects and that direct genetic effects are more consistent across biological contexts than indirect genetic effects. Our results also add evidence to the notion that behavioral plasticity in response to the social environment need not be accompanied by detectable differences in brain gene expression (Kay et al., 2022a). Comparative study of other supergene systems (Kay et al., 2022b) will help to test the generality of our findings across taxa.