Discussion
Our experimental design leverages samples from two organs and the full
breadth of social chromosome genotypes and social forms to provide a
high-resolution snapshot of the transcriptomic effects of these factors
in fire ants, with the aim of exploring the direct and indirect genetic
effects of the supergene. Our focal samples, young queens (gynes)
embarking on nuptial flights, arguably are the most relevant caste/life
stage on which to conduct such a study. These potential future
reproductives were collected while engaged in a pivotal activity that
immediately precedes nest founding (monogyne form) or joining (polygyne
form), thus marking the initiation of a queen’s reproductive life. Fire
ant workers are obligately sterile, though essential to colony fitness,
while males are short-lived and appear to play no meaningful role in
colony social life. Inclusion of SB/SB gynes from both monogyne
and polygyne colonies facilitates direct comparison of the gene
regulatory consequences of the contrasting developmental environments of
the two colony types, while sampling two organs of particular relevance
to the alternate social syndromes (brains mediating queen
dispersal/nest-founding behaviors and ovaries enabling reproduction)
from individual ants, with appropriate biological replication, allows us
to begin linking individual facets of the complex phenotype of polygyny
to putative genetic mechanisms. Expression profiling of two organs also
provides the opportunity to screen for loci with putative constitutive
regulatory effects, that is, genes differentially expressed in multiple
cellular contexts and thus possibly mediating body-wide effects ofSb . Finally, inclusion of the rare homozygous Sb/Sbgenotype allows exploration of the potential genetic causes of the
documented, presumably endogenous effective lethality of this genotype.
The genome-wide perspective. Considering the striking
number of individual- and colony-level differences between the monogyne
and polygyne forms of S. invicta ―e.g., in worker behavior towards
queens, queen cuticular chemical profiles and reproductive trajectories,
worker gene expression patterns, and many other biological features
(Eliyahu et al., 2011; Keller & Ross, 1998; Trible & Ross, 2016; Wang
et al., 2008)―we predicted that differences in the social environment in
which gynes are reared would be reflected in gene expression differences
well beyond those that can be explained by social chromosome genotype
alone. Although we observed no significantly differentially expressed
genes (DEGs) directly attributable to colony social form in our
multifactorial-model analysis (FDR < 0.01), polygyne-rearedSB/Sb gynes did exhibit many more DEGs in ovaries when compared
to monogyne-reared SB/SB gynes (205 DEGs) than when compared to
polygyne-reared SB/SB gynes (43 DEGs; Figure 3). This result
mirrors the previously observed small, but statistically significant,
higher weight of monogyne-reared, as compared to polygyne-reared,SB/SB gynes (all SB/SB gynes, however, are far heavier
than polygyne-reared SB/Sb or Sb/Sb gynes) (C. DeHeer et
al., 1999). Gyne weight in fire ants reflects amount of storage reserves
accumulated during sexual maturation and profoundly affects prospects
for successful independent founding (e.g., the ability to produce
workers solitarily), versus colony joining, as an individual
reproductive strategy (C. J. DeHeer, 2002). Our gene expression results
point to an interaction between the effects of natal colony type and
supergene-mediated gene regulation that is superimposed on the small
effect of natal colony type alone.
We predicted that gene expression differences associated with social
chromosome genotype would preferentially involve genes located within
the supergene region (Nipitwattanaphon et al., 2013; Wang et al., 2013).
The reason is that extensive sequence divergence exists between theSb and SB haplotypes of the supergene in U.S. populations
of S. invicta (estimated at 1.4 SNPs per Kb; (Pracana, Priyam, et
al., 2017; Wang et al., 2013)), contrasted with the rest of the genome,
which recombines freely and is subject to constant high levels of
admixture between the social forms (Ross, Krieger, Keller, & Shoemaker,
2007; Ross, Shoemaker, Krieger, DeHeer, & Keller, 1999; Shoemaker,
Deheer, Krieger, & Ross, 2006), mainly via polygyne-derivedSB/Sb gynes mating with monogyne-derived SB (haploid)
males (Ross & Shoemaker, 2006). Consequently, linkage disequilibrium
outside of the supergene region is minimal in both social forms (Ross &
Shoemaker, 2018; Yan et al., 2020). Our data expand upon the previous
finding that genes differentially expressed in association with social
chromosome genotype are disproportionately localized within the
supergene region (Nipitwattanaphon et al., 2013; Wang et al., 2013),
illustrating the importance of supergene cis -regulatory
evolution. Nonetheless, many of the genes differentially expressed in
association with social chromosome genotype are located elsewhere in the
genome, illustrating that trans -acting effects, or other
epistatic interactions involving the supergene, are also widespread and
potentially of major significance in the evolution of the supergene and
the alternate social forms in fire ants. This is not surprising given
the prevalence of trans-acting effects documented in supergene systems
mediating sexual dimorphism (Parsch & Ellegren, 2013; Wijchers &
Festenstein, 2011) and, strikingly, even in the relatively small
(~100kb) supergene regulating mimicry in Papilio
polytes butterflies (Kunte et al., 2014).
An important, related point is that the majority of supergene-driven
DEGs in our multifactorial-model comparisons show higher expression in
individuals carrying a copy of the Sb supergene haplotype, both
within and outside of the supergene region (Figures 2, S2). Preliminary
indications of preferentially elevated gene expression in the presence
of the Sb haplotype have been reported (Wang et al., 2013), but
the broad scope of this phenomenon was not established previously. Thus,
the observed “degenerative expansion” (Stolle et al., 2019) of the
Y-like Sb supergene haplotype is, somewhat paradoxically,
accompanied by an increase in transcriptional activity for manySb alleles at loci within the supergene. This Sbupregulation could be driven by a combination of cis- regulatory
evolution, increased chromatin accessibility, and/or recentSb- specific gene duplication events that appear to have occurred
in conjunction with the expansion of the Sb haplotype (Fontana et
al., 2019; Stolle et al., 2019).
Although such gene duplication within the inversion-derived Sbhaplotype likely explains some of the widespread Sb -mediated
upregulation observed in our analyses, we think it is unlikely to be the
sole driver of this phenomenon for three reasons. First, while Fontana
et al. (2019) show a general increase in transcript levels from genes
with Sb- specific paralogs, they found no significant correlation
between gene expression level and copy number, implying that regulatory
element evolution may play a larger role than copy number in observed
expression differences between SB and Sb haplotypes inS. invicta . Second, we detected several genes in our analysis
that exhibited differential expression without significant
allele-specific expression, indicating that their increased
transcription is facilitated by both the Sb and SBhaplotypes rather than being a simple consequence of possession of extra
copies of the Sb -linked variants. Finally, while a number of our
candidate genes have undergone gene duplication events, the increase in
gene expression we observe in Sb -carrying individuals is
significantly higher than one would expect from a simple gene
duplication event (or even several), with log2-fold
changes greater than 5 in many duplicated genes (Table 1). Ultimately,
the Sb -upregulated DEGs we report reflect an increase in the
number of identical or highly similar transcripts in Sb- carrying
individuals that collectively influence downstream pathways and
phenotypes, no matter the specific causal genetic architecture.
Our comparisons of supergene effects on gene expression in the brains
and ovaries revealed some key differences in the profiles of the two
organs. In particular, ovary DEGs are less concentrated in the supergene
than brain DEGs (Figure 2D). This indicates that regulatory effects of
the supergene may be more indirect in the ovaries than the brain,
involving more diffuse sets of genetic pathways regulated largely bytrans -acting effects. Moreover, many more DEGs are detected in
ovaries than the brain in comparisons of the two homozygous classes
(SB/SB versus Sb/Sb ; Figure 2), suggesting that the
transcriptional consequences of supergene homozygosity are particularly
pronounced in the ovarian tissues of Sb/Sb gynes. Because these
individuals do not survive to reproduce in U.S. populations, it was not
possible to gauge the effects of these expression differences with
respect to fecundity; however, maturing Sb/Sb gynes have been
shown to suffer curtailed development in terms of other physiological
markers associated with onset of oogenesis compared to gynes of the
other two genotypes (Hallar et al., 2007).
The supergene perspective. Given that minimal levels of
recombination are observed within the Sb haplotype (Ross &
Shoemaker, 2018; Wang et al., 2013; Yan et al., 2020), its constituent
genes are effectively inherited as a single Mendelian element and are in
near-complete linkage disequilibrium. Nonetheless, expression profiles
and associated selective pressures are expected to vary among supergene
genes in relation to their specific functions. Indeed, we observed
spatially heterogeneous gene expression patterns across the supergene
region in S. invicta gynes, with genes expressed more highly inSb than SB , as well as genes showing the opposite pattern,
clustered together more commonly than expected by chance.
We highlighted two regions of the supergene with these distinctive gene
and allele-specific expression patterns (Figure 5,S5). The first, here
denoted region 1, exhibits minimal differential expression in theSB/SB-SB/Sb comparison, while showing extensiveSB/SB -biased expression in the SB/SB-Sb/Sb comparison.
Concurrently, this region shows disproportionate SB -biased allele
expression in heterozygous SB/Sb individuals. These results point
to diminished expression of the alleles within the Sb haplotype,
compensated by means of overexpression of the SB versions of
those genes. We hypothesize that genes subject to this SB dosage
compensation in region 1 (and elsewhere) contribute to the observed
inviability of Sb/Sb queens (Hallar et al., 2007). Dosage
compensation is a commonly observed phenomenon in sex chromosome systems
(Disteche, 2012; White et al., 2015) and has been documented recently in
the supergene of the white-throated sparrow (Sun, Huh, Zinzow-kramer,
Maney, & Yi, 2018). A second region of the supergene, here denoted
region 2, contains genes exhibiting strong Sb -biased gene
expression and Sb allele-specific expression (Figure 5),
indicating upregulation driven largely by expression of theSb -linked alleles. We hypothesize that genes in region 2 (and
other genes conforming to this regulatory pattern) act as functional
drivers of the polygyne syndrome and thus are important candidates of
major effect on this social polymorphism.
Gene-specific, functional observations. We leveraged our
sampling scheme to search for genes exhibiting differential expression
patterns that parallel specific phenotypic differences observed between
gynes of each social chromosome genotype. Specifically, we classified
DEGs into four categories: brain-specific, ovary-specific,Sb/Sb -specific, and constitutive (Table 1). Among eight detected
brain-specific DEGs, Ribonuclease H-like is notable for having a
human ortholog linked to splicing and neurological disfunction (Bin et
al., 2016). Additionally, among the twelve detected ovary-specific DEGs,sphingosine-1-phosphate lyase is notable for having aDrosophila ortholog involved in sphingolipid metabolism that is
linked to reproductive deficiencies (Phan et al., 2007). This gene also
stands out as the only characterized gene in the ovary-specific category
with lower expression in heterozygotes than SB/SB homozygotes.
Among eight detected Sb/Sb- specific DEGs (those with differences
in both organs but solely in the Sb/Sb -SB/SB comparison),pre-mRNA-splicing factor Slu7-like and peptide deformylase,
mitochondrial-like are notable for their putative impacts on the
production of properly spliced and processed proteins. Finally,
among the thirteen detected constitutive DEGs (those with differences in
all comparisons of Sb- supergene presence/absence) we identified
notable genes with orthologs involved in metabolism (NADH
dehydrogenase ), odorant perception (pheromone binding protein
GP-9-like , also referred to as an odorant binding protein) (Leal,
2013), and brain development and sperm motility (growth
arrest-specific protein-8 ) (zur Lage, Newton, & Jarman, 2019).Gas-8 may be implicated in differences in traits as diverse as
gyne dispersal behavior (C. DeHeer et al., 1999) and male reproductive
capacity (Lawson, Vander Meer, & Shoemaker, 2012). Interestingly, one
of the constitutive supergene-effect loci we identify appears to be
located outside of the supergene region: nose-resistant to
fluoxetine protein 6-like is affiliated with defects in lipid transport
to the ovaries and with lifespan in Caenorhabditis elegans and
humans (Brejning et al., 2014; Choy & Thomas, 1999). This gene is a
candidate for direct, trans -regulation by the Sb supergene
and may contribute to the diminished early fecundity of polygyneSB/Sb queens relative to SB/SB queens (C. J. DeHeer, 2002;
Keller & Ross, 1993). However, given recent evidence that duplicate
genes may arise by translocation into the Sb supergene from other
genomic locations (Fontana et al., 2019), improved Sb assemblies
and annotation will be necessary to further explore this hypothesis.
Pheromone binding protein GP-9-like belongs to one of the best
studied gene families in S. invicta : the insect odorant binding
protein (OBP) family (Keller & Ross, 1998, 1999; Pracana, Levantis, et
al., 2017). These genes are thought to produce proteins that function as
molecular carriers, transporting odorant molecules to their receptors in
some canonical study systems, though OBPs likely also function outside
of the chemosensory system as general chaperone molecules (Leal, 2013).
In S. invicta , OBPs are of particular interest, as it has long
been known that the fire ant social polymorphism is marked by fixed
amino acid differences at an OBP, General protein-9 (encoded by the geneGp-9 ) (Ross, 1997). We found that the most strongly
differentially expressed OBP-encoding gene in our data was not the
widely studied Gp-9 , but a paralog, OBP12 , that exists as
duplicated tandem genes on the Sb haplotype and as a single-copy
gene on the SB haplotype (Fontana et al., 2019; Pracana,
Levantis, et al., 2017). This Sb -specific duplication ofOBP12 corresponds with a unique increase in gene expression andSb allele-specific expression among S. invicta OBP genes
(reads from both Sb paralogs evidently mapped to the singleOBP12 encoded in the SB reference genome in our analyses;
Figure S4). Our data are largely consistent with the recent findings by
Dang et al. (2019) regarding increased expression of OBP12 in
antennae of Sb -carrying workers, though we also observed
increased OBP12 expression in both the brains and ovaries ofSb -carrying gynes, which may mitigate against the notion of
antennal-specific neofunctionalization (Dang, Cohanim, Fontana, Privman,
& Wang, 2019). Additionally, the >12-fold increase in
expression we observed for OBP12 in Sb -carrying
individuals compared to SB/SB individuals is much higher than the
simple doubling one might expect if the duplicate was expressed
similarly to its progenitor (Table 1, Figure S4). This points to effects
of gene regulatory evolution. Our findings are consistent with the
hypothesis that duplication of OBP12 in the Sb haplotype
and subsequent Sb- upregulation of the derivative OBP12 may
facilitate some differences in sensory functions and response behaviors
between monogyne and polygyne fire ants (Keller & Ross, 1998; Trible &
Ross, 2016).
Conclusion. We find that the gene regulation underlying
a major alternate form of social organization in fire ants is more
complex than the simple inheritance patterns or broad structural
differences in the implicated supergene may superficially suggest. As
hypothesized (Y. C. Huang & Wang, 2014), gene expression evolution
appears to play a large role in the evolution of the complex phenotype
of polygyny in S. invicta . We show that possession of theSb inversion-mediated haplotype by a queen leads to changes in
gene expression both within and beyond the boundaries of the
constitutive supergene inversions, that such supergene-mediated effects
interact with those of the developmental environment, and that genes
differentially expressed in queens bearing different supergene genotypes
are heterogeneously distributed along the supergene. Further work
exploring the interplay of protein sequence, gene expression, and
regulatory element evolution is necessary to better understand the
genetic causes of a fundamental shift in fire ant colony social
organization with widespread behavioral, ecological, and economic
consequences (Tschinkel 2006).
Acknowledgements
We thank Dietrich Gotzek for the idea to sample gynes embarking on
mating flights, Bob Schmitz and Nick Rohr for library preparation,
Yannick Wurm and Rodrigo Pracana for providing linkage mapping and OBP
annotation information, and the Georgia Advanced Computing Resource
Center, a partnership between the University of Georgia’s Office of the
Vice President for Research and Office of the Vice President for
Information Technology. This work was supported by U.S. NSF grants to
B.H. and K.R. (1755130) and K.R. (1354479) and U.S. Federal Hatch funds
to B.H. and K.R.