Genetic patterns within each species
The patterns of genetic diversity along the threespine genome have
previously been described in studies of divergence between marine and
freshwater threespine population pairs (Hohenlohe et al. 2010, Chan et
al. 2010, Jones et al. 2013, Roesti et al. 2015). FSTscores typically cluster in several broad peaks in comparisons among
freshwater and marine environments, with pronounced peaks around theEda locus (chr4; Hohenlohe et al. 2010) and the Pitx1locus (chr7; Chan et al. 2010), which are involved in freshwater
adaptation. Additionally, broad peaks found at three inversions (chr1,
11 & 21) have also been associated with freshwater adaptation (Jones et
al. 2012; Roesti et al. 2015). Unexpectedly, as we compared two marine
populations, we identified some of these characteristic patterns of
marine-freshwater divergence in this study (Fig. S3). A possible
explanation is that the northern and southern populations differ in the
degree to which they receive gene flow from freshwater populations. In
the south, threespines were sampled from an isolated stream that drained
directly into the ocean, while the northern threespines were sampled
from a lake connected to an estuary (Tables S1). Counterintuitively, the
patterns we found probably came from freshwater alleles in the southern
population, as a previous study of the lake in the north found no
evidence of hybridization between ‘anadromous’ and freshwater
populations (Drevecky, Falco and Aguirre, 2013), and a study of marine
populations in the North-West Pacific found a higher frequency of
freshwater associated alleles at the EDA locus in Oregon than
Alaska (Morris et al. , 2018). However, to test such hypotheses
about introgression, we would have to look at the frequency of the
low-plate EDA allele and the frequencies of inversions in Oregon
and Alaska and contrast this with nearby freshwater populations. An
alternative explanation is that the some of the patterns of
marine-freshwater adaptation may also be pleiotropically connected to
thermal regulation, as has been suggested for the EDA locus
(Morris et al. , 2018). Whether it is differential gene-flow or
pleiotropic adaption, we have found that the genomic landscape of
geographically diverse marine threespines is strikingly similar to the
marine-freshwater landscape.
In contrast to the patterns found in threespines, no large peaks of
FST were present along the tubesnout genome (Fig. 2).
Instead, there were several small and narrow FST peaks
suggesting that the tubesnout genome has been shaped by processes that
do not leave strong genetic signals, such as genetic drift or polygenic
adaptation (Rockman 2012, Stinchcombe and Hoekstra 2008, Yeaman 2015).
As the Null-W test is designed to detect linked clusters of
FST outliers, this also explains the lack of any
signatures of convergent evolution. Since the patterns of
FST were not strongly heterogeneous in tubesnout, it is
unsurprising that no significant matches to threespine were found.
The genetic patterns present in the ninespine stickleback were likely
the result of a strong genetic bottleneck and isolation between the
northern and southern populations, as on average, genetic divergence was
high and genetic diversity was low in all four populations (Table 1,
Fig. 2). Southern populations were sampled from two prairie lakes, which
were formed when a larger post-glacial lake dried up, isolating these
ninespine populations and presumably causing a genetic bottleneck
(Tufts, 2018), similar to the founder-effect observed in Nordic
populations (Shikano et al. , 2010). In contrast, the northern
populations were sampled from lakes close to the sea, which potentially
has provided several opportunities for gene flow from the marine
populations. A phylogeographic study separated ninespine populations
from the Atlantic coast and Great Lakes regions into two post-glacial
lineages, with evidence suggesting that the divergence time among these
lineages may be much older than the last glacial maximum (Aldenhovenet al. , 2010). Presumably, the prairie lake populations are part
of this Great Lakes lineage (Tufts, 2018) and therefore should be highly
diverged from the Northern populations. The extreme genetic divergence
among these populations is likely to be the result of long-term genetic
isolation combined with a strong genetic bottleneck in the southern
populations, not adaptation to latitude.
Comparing the genome scans of all species reveals three distinct
patterns, suggesting that the balance between the evolutionary processes
has differed among these species. The FST Manhattan
plots (Fig. 2A) show different patterns, which can be interpreted as the
result of three distinct evolutionary scenarios: local adaptation
(threespine), genetic bottlenecks (ninespine) and a weak or polygenic
selection and/or drift (tubesnout). This does not imply that the
ninespine has not experienced selection or that the threespine has not
been affected by drift, just that the patterns of diversity in the
genome have been more strongly affected by different processes in each
species.
A major caveat to these results is that very few populations were
sampled per species. Pool-seq mixes alleles across a population, which
means that the basic sampling unit is a population, in effect each
species had only 2-4 data points. The comparisons made in this study may
have been underpowered to detect any shared genetic patterns. However,
the presence of threespine peaks in previously identified regions
undergoing adaptation (Fig. S3) shows that strong genetic patterns were
detectable, thus only subtle patterns of genetic diversity were lost.
The lack of this pattern in tubesnout may be due to the lack of an
evolutionary history of repeated colonization followed by gene-flow from
freshwater populations, which can lead to complex genomic architecture
for adaptive traits (Tigano and Friesen, 2016; Faria et al. ,
2019). All things considered; this study demonstrates the diversity of
genetic patterns that can be identified from genome scans of wild
species, even with a limited number of populations.