Results
Direct genetic effects. Seventy-one genes were
differentially expressed between brains of SB/SB and SB/Sbworkers and twenty-three between brains of SB/SB and SB/Sbgynes (False Discovery Rate [FDR] < 0.1; Figure 2A).
Eighty-six genes were differentially expressed between gasters (the
abdominal segments following the post-petiole) of workers and 215 genes
between ovaries of gynes with different supergene genotypes (FDR
< 0.1; Figure 2A). In all tissues and castes where the gene
expression of SB/SB and SB/Sb was compared, the
differentially expressed genes tended to be upregulated in SB/Sbindividuals (Figure 2E) and more of the differentially expressed genes
were located within the supergene region than expected by chance (Figure
2I).
Genes that were differentially expressed in one tissue or caste because
of direct genetic effects were frequently differentially expressed in
the other tissue or caste (Figure 3A). A majority of differentially
expressed genes in both worker and gyne brains were differentially
expressed in other sample types, and eight of these genes were
differentially expressed in all four direct genetic effect comparisons
(Figure 3A). Four of these eight genes had functionally annotated
homologs: growth arrest specific 8 (gas-8 ), nose
resistant to fluoxetine-6 (nrf-6 ), and two Harbinger
Transposase derived-1 (harbi1 ) genes (Supplementary Data C). To
further characterize the direct genetic effects on transcription, we
used a generalized linear model to identify genes that were
significantly differentially expressed by genotype when including
samples of both castes and tissues and modeling the effects of genotype,
social environment, caste, and tissue (McCarthy, Chen, & Smyth, 2012).
This analysis revealed 145 significant genes that were differentially
expressed by the direct genetic effects of genotype (FDR <
0.1; 115 differentially expressed genes at FDR < 0.05, 77 at
FDR < 0.01; Figure 3C). Gene Ontology functional enrichment
tests (Fisher’s Exact Test, P < 0.01) for these direct genetic
effect genes (FDR < 0.1) revealed that locomotion ,reproduction , and numerous metabolism terms were overrepresented
(Supplementary Data A).
Indirect genetic effects . The number of genes expressed
differentially by indirect genetic effects (comparison of SB/SBadults reared in different social environments from egg through two
weeks of adulthood) was lower than the number of genes differentially
expressed by direct genetic effects (comparison of SB/SB andSB/Sb adults from the same social environment) in all comparisons
(worker brain, worker gaster, gyne brain, and gyne ovaries), when using
an FDR of 0.05 or 0.01 as a cutoff for significance. At the least
stringent statistical cutoff FDR of 0.1, this was also true for brains
of workers and gynes but not for abdominal tissues (Figure 2A,B). Genes
differentially expressed by indirect genetic effects tended to be
upregulated in individuals from polygyne colonies as compared with
individuals from monogyne colonies in the gasters and ovaries of workers
and gynes respectively (Figure 2F, positive log2expression ratios). In contrast to direct genetic effects, genes
differentially expressed by indirect genetic effects were not located
within the supergene region more frequently than expected by chance
(Figure 2J). The transcriptional signature of indirect genetic effects
was also much less pronounced than that of direct genetic effects across
sample types; on average, the fold-change was significantly lower for
genes differentially expressed by indirect genetic effects than for
genes differentially expressed by direct genetic effects for 3 of 4
caste and tissue combinations (worker brains, worker gasters, and gyne
ovaries at FDR < 0.1; Welch Two-Sample t-tests, P <
0.01; Figure 2E,F). These findings each mirror the previous findings
using solely gyne data (Arsenault et al., 2020). Our analysis of age-
and size-matched worker samples helps to rule out the possibility that
differences in gene expression could result from age-induced differences
(Lucas, Romiguier, & Keller, 2017) because the age of gynes can vary at
the time of nuptial flight (Arsenault et al., 2020; Nipitwattanaphon,
Wang, Dijkstra, & Keller, 2013). While there are well-documented
behavioral differences between workers in monogyne and polygyne colonies
(Gotzek & Ross, 2008; Keller & Ross, 1998; Ross & Keller, 1998, 2002;
Trible & Ross, 2016; Zeng, Millar, Chen, Keller, & Ross, 2022), we
observed only few genes that were differentially expressed in brains of
individuals exposed to these different social environments, with none
observed in workers and only five in gynes (FDR < 0.1; Figure
2B). Thus, pronounced indirect genetic effects of the fire ant supergene
were primarily detectable in abdominal tissues, in contrast to the
direct genetic effects, which were prevalent in abdominal tissues and in
the brain.
To quantify the indirect genetic effects that occurred from eclosion
until the pupal stage, we compared SB/SB workers that always
experienced a polygyne social environment and SB/SB workers that
experienced a monogyne social environment until the pupal stage and then
a polygyne social environment through the first 14 days as adults. This
analysis revealed no significant differences in gene expression (FDR
> 0.1; Figure 2C). To quantify the indirect genetic effects
that occurred during the first 14 days as adults, we comparedSB/SB workers that always experienced a monogyne social
environment and SB/SB workers that experienced a monogyne social
environment until the pupal stage and then a polygyne social environment
during the first 14 days as adults. This analysis also revealed no
significant differences in gene expression (FDR > 0.1;
Figure 2C).
The proportion of differentially expressed genes (FDR < 0.1)
that overlapped between castes or tissues was 22 times higher for the
set of genes influenced by direct genetic effects (Figure 3A) than those
influenced by indirect genetic effects (Figure 3B). To further
characterize the indirect genetic effects of monogyne versus polygyne
social environments, we utilized a generalized linear model to identify
genes that were significantly differentially expressed when including
samples of both castes and tissues in a model that considered the
effects of genotype, social environment, caste, and tissue (McCarthy et
al., 2012). This analysis revealed 700 significant genes that were
differentially expressed by the indirect genetic effects of the social
environment at a lenient statistical cutoff (FDR < 0.1; 7
differentially expressed genes at FDR < 0.05, 0 at FDR
< 0.01; Figure 3D).
The genes differentially expressed in the direct genetic effects
generalized linear model (FDR < 0.1) exhibited higher absolute
log2-transformed expression ratios than genes
differentially expressed in the indirect genetic effects generalized
linear model (Welch Two-Sample t-test; P < 0.01; Figure 3C,D).
Gene Ontology functional enrichment tests (Fisher’s Exact Test, P
< 0.01) revealed an overrepresentation of genes differentially
expressed in response to indirect genetic effects (FDR < 0.1)
that were annotated with the terms response to stimulus ,oogenesis , and various neural development terms (Supplementary
Data A).
Direct + indirect genetic effects. Given that the
behavior of workers depends both on their own genotype and social
environment (due to the presence of SB/Sb workers only in
polygyne colonies), we wanted to test for a combinatorial effect of both
colony social environment and supergene genotype on transcription.
Analyses of these combined direct and indirect genetic effects were
performed by comparing gene expression of SB/SB individuals from
a monogyne social environment to SB/Sb individuals from a
polygyne social environment (Figure 2D). The numbers of the
differentially expressed genes (FDR < 0.1) in brains due to
these combined effects in workers (n = 78) and gynes (n = 22) were
similar to the sum of differentially expressed gene tallies from
separate analyses of direct and indirect genetic effects in each caste
(workers: n = 71 + 0 = 71; gynes: n = 23 + 5 = 28). In worker gasters,
many fewer genes were differentially expressed (FDR < 0.1) by
combined indirect and direct effects (n = 117) than the sum of tallies
from separate analyses of these effects (n = 86 + 139 = 225). This could
be explained by overlap between direct and indirect effect gene sets
(examined below), opposing directionality of direct and indirect effects
on expression, and/or noise in the data. Gyne ovaries offer a stark
contrast to the other sample types, in that there were over 2000 more
differentially expressed genes (FDR < 0.1) observed in
response to the combination of direct and indirect genetic effects (n =
3451) than in separate tallies of indirect and direct genetic effects (n
= 215 and 966, respectively).
A generalized linear model including samples of both castes and tissues
revealed that the overlap between direct genetic effects and indirect
genetic effects on differential gene expression was higher than expected
by chance (Fisher’s Exact Test; p < 0.05; Figure 4A). However,
when considering tissues and castes separately, the overlap between
direct and indirect genetic effects on gene expression was significantly
greater than expected by chance in only one of the four pairwise
comparisons (gyne brains), where it was small (Fisher’s Exact Test; p
< 0.05; Figure 4B-E). In gyne ovaries, the overlap was
significantly less than expected by chance (Fisher’s Exact Test; p
< 0.05; Figure 4C). The genes differentially expressed by both
direct and indirect genetic effects, considered separately, included a
zinc-finger containing transcription factor, limulus clotting
factor C , two transposase-derived HARBI1 proteins, and three
odorant receptor genes (Supplemental Data D).