Results
We produced panels of 15,412 SNPs (7.6 % missing data) for 90
individuals of H. molleri and 33,140 SNPs (5.2 % missing data)
for 83 individuals of P. cultripes .
Genetic structure
STRUCTURE runs converged well for low K values but not for larger K
values (Table S2; Figures S3.1 and S3.2). The best-supported number of
genetic clusters (K) identified using STRUCTURE varied according to the
metric used (PI or ΔK) and marker type. In most cases, we found the best
support for two genetic lineages (K = 2), but some metrics identified
further substructure, with up to six genetic clusters (K = 6) when using
PI (Table S3; Figures S4.1 and S4.2).
Ancestries derived from both markers were spatially coherent at
different K values. That is, individuals from the same or nearby
localities shared similar ancestries and more admixed individuals
coincided with geographical shifts in cluster assignment (Figure 1). For
K = 2, both marker types were congruent in identifiying major
subdividions in each species: a northern and a southern lineage forH. molleri , and a central-western and a northeastern lineage forP. cultripes . From K = 3 to K = 8, the spatial patterns of
genetic structure for both species were largely congruent between marker
types in terms of admixture levels and ancestry group assignment
(Figures 1 and S5). Both markers generally agreed on the genetic
ancestry of localities or group of localities as sharing a singular
genetic ancestry, although the K value at which for a given assignment
to a cluster could differ between markers. For instance, for H.
molleri , the western-coastal populations from Portugal (dark purple,
Figure 1) formed a well-differentiated cluster at K = 3 with SNPs and at
K = 4 with microsatellites. Another example is the locality Ojos de
Villaverde, at the southeastern-most corner of the distribution ofH. molleri. This locality appeared well differentiated at K = 4
for SNPs (green), but at K = 5 in microsatellites (magenta) (Figure 1).
In P. cultripes , we observed the same phenomenon. For instance,
the localities from northwestern Portugal were very differentiated at K
= 4 with SNPs (green), but at K = 5 with microsatellites (green, Figure
1). Both markers agreed in localities within the northern half of the
Iberian Peninsula with nearly “pure” ancestries and no further
clustering after K = 4, and yielded very admixed localities in the
southern half of Iberia from K = 4 to K = 8, although the levels of
admixture and the ancestry assignments differed notably between markers.
In P. cultripes , for K = 7 and K = 8, microsatellites yielded
more admixed individual ancestries compared to SNPs (Figure S5), driven
by the more admixed southern localities (Figure 1). For H.
molleri , we could not quantify reliably these differences in admixture
levels between markers because the individuals analyzed for each dataset
were not all the same.
Genetic structure based on STRUCTURE analyses was highly congruent with
that inferred by model-free hierarchical clustering (Supplementary File
S1), which yielded well-supported clades for SNPs but less so in
microsatellite-based topologies.
Congruence in individual/locality ancestries between
microsatellites and SNPs
Both species showed higher intra-marker similarity (H. molleri ,
SSCs = 0.27 - 1.00; P. cultripes , SSCs = 0.77 - 1.00) than
inter-marker similarity (H. molleri , SSCs = -0.03 - 0.42;P. cultripes , SSCs = 0.55 – 0.89) (Figure 2). For
microsatellites, ancestries were very similar (SSCs close to 1) from K =
2 to K = 8 (except K = 7) for H. molleri and from K = 2 to K = 4
for P. cultripes . For SNPs, STRUCTURE results were almost
identical only from K = 2 to K = 4 for H. molleri , but up to K =
6 for P. cultripes . Larger K values were in all cases associated
with less consistent results across STRUCTURE runs. For most K values,
pairwise SSC values in microsatellite runs had a larger spread (i.e. a
greater range of values), especially at larger K values. This spread was
minimum for STRUCTURE results derived from SNPs, though at larger K
values (K = 4 to K = 8 for H. molleri ; K = 6 to K = 8 forP. cultripes ) they tended to converge into 2 or even 3 regions of
the parameter space (Figure 2). The similarity between
SNP-microsatellite runs did not follow a clear pattern along increasing
K. For H. molleri , SSCs were homogenously lower across all K
values than for P. cultripes , highlighting the distinct solutions
obtained between datasets. For this species, SSCs were maximum at K = 2
(0.89), and minimum at K = 4 (0.55). From K = 5 to K = 8, SSCs had a
small increase in the 0.58 – 0.68 range.
Microsatellites yielded more admixed ancestries at larger values of K
(i.e. K = 7 and K = 8; Figure S5) which seem to be driven by the more
complex patterns of genetic structure in the southern localities (Figure
1).
Genetic diversity
Correlation of genetic diversity between microsatellites and SNPs-based
measures was weak in both species (P. cultripes , Pearson’sr = 0.39, P < 0.001). Genetic diversity (sMLH)
from SNPs in H. molleri was highest in southwest Iberia and
decreased towards northern (β = -0.08; P < 0.001) and
eastern localities (β = -0.04; P = 0.02) (Figure 3; Table S4). We
did not detect a significant correlation of microsatellite diversity
with latitude (P = 0.63) or longitude (P = 0.10).
For P. cultripes , genetic diversity decreased with latitude for
SNPs (β = -0.07; P < 0.001) and microsatellites (β =
-0.09; P < 0.001). Longitude had a marginal effect on
diversity from SNPs (β = -0.02; P = 0.06) but not from
microsatellites (P = 0.93). Both markers agreed in diversity
being (1) extremely low in the north-eastern localities, in costal
France, both on the Atlantic and Mediterranean sides, (2) moderately low
in the Northern Plateau and along the Mediterranean coast and interior,
and (3) greatest in the central south-western localities (Figure 3;
Table S4). These south-western localities also showed the largest
complexity in genetic structure and patterns of admixture across K
(Figure 1).