Gene-phenotype-environment associations: evidence for a role of
natural selection?
Gene flow is thought to constrain divergence by swamping locally adapted
alleles (but see ), and while phenotype-environment associations can
indicate responses to natural selection , they may also reflect
phenotypic plasticity and non-random dispersal . Evidence for selection
is therefore strengthened when observed trait divergence has a genomic
component (i.e., “gene-phenotype-environment” associations).
Importantly, all traits we measured had substantial additive genetic
variation, suggesting high levels of potential for traits to respond to
selection. This was further reflected at the genomic level, where we
were able to successfully recover similar genomic architecture as
previously reported for this species. Using the mapped genomic
architecture, we show that while divergence of most traits was not
linked to a genomic component, relative spine lengths was. Specifically,
genomic regions that were associated with pelvic spine length (that were
longer in deeper waters), were divergent across water depth
strengthening the conclusion that the observed divergence in pelvic
spine length is a result of divergent natural selection – likely due to
predation.
Mapping the causative pathways from genes to phenotypes is a notoriously
challenging task , but identifying the genes on regions involved in
gene-phenotype-environment associations provide insight into the
molecular mechanisms underlying adaptation. On the genomic regions
associated with the pelvic spine length - depth relationship, we found
two genes (ULK2 and CCNB1) with haplotypes coding for alternate amino
acids. These are functionally interesting as ULK2 is a serine/threonine
kinase which, along with its homolog ULK1, interacts with the master
regulator of metabolism (mTOR) and regulates apoptosis in response to
starvation, and CCNB1 codes for Cyclin B1, a major kinase
regulator which activates mitosis and regulates the dynamics of the cell
cycle. Whilst these findings suggest candidate molecular functions
underlying responses to natural selection, explicit follow up studies
are needed to get at the causal relationships with pelvic spine length
variation.
Many of the regions of the genome that were correlated with
environmental variation in our study were not linked to observed trait
divergence. This is not surprising given that we only measured a
selected subset of traits within specific functional categories. An
understanding of what biological functions these regions are associated
with can therefore provide hypotheses about targets of divergent natural
selection for future studies. In our study, the biological processes
implicated in genomic regions included developmental processes, such as
development of nervous and sensory systems. Notably, visual response to
abiotic stimulus and perception of sound were enriched biological
functions that were associated with the community structure of
invertebrates. This is particularly interesting for Mývatn stickleback
due to potential for influencing ability to respond to prey stimulus,
and because the sensory drive hypothesis posits that an organisms’
communication system should be especially sensitive to ecological
variation .
Although adaptation can occur across exceptionally short time frames ,
selection often acts multifariously and environments are rarely stable
temporally, resulting in fluctuating selection pressures . Mývatn is a
highly dynamic ecosystem, where multiple dimensions of its ecology and,
importantly, stickleback population size fluctuate substantially through
time . Our environmental, phenotypic and genomic data were collected at
a single point in time, and may therefore not accurately reflect the
selective environment experienced by Mývatn stickleback within or across
generations. Whilst using ecological data collected at the same (single)
time point as genomic and phenotypic data is common practice in studies
that investigate selection in wild populations , tracking
gene-phenotype-environment associations through time would allow
inferences on how patterns of phenotypic and genomic variation
withstand, or respond to, temporal fluctuations in selective pressures.
In context of Mývatn stickleback, this would also facilitate the
exploration of the spatiotemporal balance between adaptive divergence
and gene flow.