Population structure and conservation units
After excluding the candidate adaptive loci and loci deviating from HWE,
we kept 2,663 and 1,711 SNPs for the NIDGS and SIDGS, respectively. With
these sets of neutral loci, we determined that there is a significant
pattern of IBD for NIDGS (r = 0.136, p = 0.039) but not
for SIDGS (r = 0.556, p = 0.090). IBD was not significant
for the NIDGS adaptive dataset (r = 0.018, p = 0.396), or
for the SIDGS adaptive dataset (r = -0.378, p = 0.950).
PCA results for the total and neutral datasets in both species were
identical (Figure S8, Supporting information). From the NIDGS PCA
analysis, there were four axes for the total and neutral datasets and
three axes for the adaptive dataset explaining a significant proportion
of the variance (Figures 5A, 5B and S8, Supporting information). Both
the neutral and adaptive datasets identified similar populations as the
most distinct, particularly Lower Butter (LB), Rocky Top (RT) and
Tamarack (TA) (Figure S8, Supporting information). Mud Creek (MC) was
only distinctive at the neutral level. Considering potential groups, at
the first two PCs, Lower Butter (LB) was clearly separated from an
eastern group (MC, TA, PV, LV) and a western group (all other
populations). Based on these data, we defined three ESUs for NIDGS: ESU1
represented by the western group, ESU2 represented by Lower Butter, and
ESU3 represented by the eastern group (Figure 6A). From the SIDGS PCA
analysis, we considered as significant one axis for the total the
neutral and the adaptive datasets, the first two identified Olds Ferry
(OF) as the most distinctive populations, while the adaptive dataset
showed a clear separation of Paddock (PA) (Figure 5C and 5D, and S8,
Supporting information). Based on these data we defined two ESUs for
SIDGS: ESU1 represented by Olds Ferry, and ESU2 represented by all other
populations east of the Weiser River (Figure 6B).
The PCA neutral results for both species agree with the STRUCTURE
analyses (Figure 6). For NIDGS, STRUCTURE suggested separation of NIDGS
into two main groups corresponding to a western group (ESU1 and ESU2)
and also an eastern group (ESU3) (Figure 6A, K = 2). Here, Lower
Butter was included in the western group, although there was a high
proportion of admixture with the eastern group as well. Lower Butter
(ESU2) was differentiated from all other populations at all Kvalues starting from K = 3 (Figure 6A). The STRUCTURE analysis
indicated a second peak in the marginal likelihood for K = 6
using the Evanno et al. (2005) method, suggesting the presence of
fine-scale population structure within NIDGS (Figures 6A, and S9A, S9B
and S10A, Supporting information). At K = 6, the eight
populations of ESU1 split into three geographically distinct populations
or MUs, and the four populations of ESU3 split into three MUs, with Lost
Valley showing evidence of admixture (Figure 6A). In terms of adaptive
variation, within ESU1 we identified two AUs from eight populations,
corresponding to the same structure found for MUs, ESU2 (Lower Butter)
was also a distinct AU, and for ESU3, only Tamarack was adaptively
differentiated (Figure 6A). These results are also reflected in the
pairwise F ST estimates, where there is no
significant differentiation among 7 of the 8 populations of ESU1, even
at the adaptive level (Figures S11A and S11B, Supporting information).
Rocky Top (ESU1), Lower Butter (ESU2), and Mud Creek (ESU3) were the
NIDGS populations with the highest neutral average pairwiseF ST, with 0.09, 0.08 and 0.08, respectively,
considering an overall average of 0.06, while for the highest adaptive
average pairwise F ST was found for Rocky Top
(ESU1) and Lower Butter (ESU2), with 0.26 and 0.20, respectively,
considering an average of 0.10. The SIDGS STRUCTURE results also
mimicked those from the neutral PCA where individuals were found to be
divided in two groups, with a separation of Olds Ferry from all other
SIDGS localities at K = 2 (Figure 6B). This resulted in the
definition of two MUs, according to our criteria. In terms of adaptive
differentiation, Paddock was identified as a separate AU.
In terms of population differentiation, averageF ST between NIDGS and SIDGS was 0.143, compared
to 0.05 and 0.100 within NIDGS and SIDGS for the same dataset,
respectively. At the intraspecific level and considering the NIDGS and
SIDGS datasets separately, average neutral pairwiseF ST was highest for Olds Ferry (0.13),
considerably higher than the average value of 0.04 (Figure S11C,
Supporting information). However, as seen in the SIDGS adaptive PCA,
Paddock (PA) was the most distinct population and also had the highest
average pairwise F ST (0.24), considerably higher
than the average value of 0.15 (Figure S11D, Supporting information).
Considering the levels of genetic diversity, for the IDGS dataset, no
NIDGS population showed significantly decreasedH O compared H E, but Lower
Butter did show significantly higher H O thanH E (Table 4). Contrarily, for the same dataset,
two out of four SIDGS populations showed significantly lowerH O than H E. SIDGS
populations tended to have significantly lower H Ethan NIDGS populations (Table S1, Supporting information). At the
intraspecific level, for NIDGS we observed an average of neutralH O of 0.237, which was significantly lower than
the average H E of 0.280 (Table 4). Values of
heterozygosity for the adaptive dataset were lower than the neutral
dataset, and H O (0.199) was lower thanH E (0.190). Of the 10 populations for which we
estimated H E, four had significantly lower
neutral H O than H E, while
only two showed lower H O thanH E for the adaptive dataset (Table 4).
Considering the comparisons among populations, Steve’s Creek/Squirrel
Valley (SS, H E = 0.305), Tamarack (TA,H E = 0.296) and Mud Creek (MC,H E = 0.286) showed significantly higher neutral
HE (Table S1, Supporting information) than all other
NIDGS populations. For the adaptive dataset, Rocky Top (RT,H E = 0.294), Tamarack (TA,H E = 0.270) and Lower Butter (LB,H E = 0.263) showed the highest HEvalues, also significantly higher than most other NIDGS populations
(Table S1, Supporting information). Comparing the HEestimated for the neutral and adaptive datasets, Mud Creek (MC), Rocky
Top (RT), and Steve’s Creek/Squirrel Valley (SS) had significantly lower
adaptive than neutral H E (Table S1, Supporting
information). For the SIDGS, we observed an average of neutralH O of 0.231, which was significantly lower than
the average HE of 0.276. Values of heterozygosity for
the adaptive dataset were similar to the neutral dataset, again withH O (0.232) lower than H E(0.275). Of the four populations for which we estimatedH E, three presented significantly lower neutralH O than H E, while none
showed lower H O than H Efor the adaptive dataset (Table 4). Considering the comparison among
populations, Olds Ferry (OF, H E = 0.287) and
Paddock (PA, H E = 0.283) showed significantly
higher neutral H E than the other three SIDGS
populations (Table S1, Supporting information). For the adaptive
dataset, Paddock (PA, H E = 0.410) showed
significantly higher HE compared to all other SIDGS
populations, as well as when compared to its neutral HEestimate (Table S1, Supporting information).