Figure 2. Principal component analysis of 276 Clethrionomys glareolus individuals sampled from 12 populations across Europe using 2,476 SNPs based on genotype likelihood. With percentage of variation explained for each component displayed on the axes, together the four components explain 20.1% of variation. Each circle represents an individual, colours correspond to Regions and sites Abbreviations - C: Central, E: Eastern, N: North, S: South, W: Western, .fi: Finland, .se: Sweden, .pl: Poland, .cz: Czechia, .de: Germany, it: Italy, .fr: France, .ro: Romania
PC1 explained 9% of the variation and sorted the populations by geography, correlating to both longitude (R2= 0.83) and latitude (R2=0.49). PC2 (5%) separated West and East within Regions; in the Northern populations the population west of the baltic sea (N.se) from the the three Northern populations east of the Baltic sea (NE1-3.fi), and within Central Europe the three geographically close populations (C1.d3, C2-3.cz) from the two Eastern European populations (CE.pl and SE.ro). The third and fourth components together explained 5.5% of genetic variation, and separated two populations from the other ten: the northern (N.se) population beyond the Baltic Sea, and the southern population beyond the Alps (S.it).
In the admixture analysis (Figure 1), the lowest variance of likelihood was found for 2 or 3 ancestral populations (Figure S2). AssumingK =2, the populations in the southern range of the distribution are separated from the other populations. Increasing to K =3 additionally separated the three populations east of the Baltic (NE1-3.fi), increasing to  K=4 separated S.it beyond the Alps, and to K=5 the Swedish population beyond the baltic (N.se). AssumingK =10 (having the next lowest variance of likelihood), clusters mirrored sampling locations, except for the two pairs of populations with the least geographic distance, the southern Finnish populations forming a single cluster, and the Central populations (C2 and C3.cz) with only some degree of admixture suggested.
A similar pattern emerges from the pairwise population FST estimates (PPF), which revealed moderate to high levels of differentiation between populations (Table S4). PPF ranged between 0.035 (C1.cz vs C2.cz) and 0.555 (N.se and SW1.fr) with an average fixation index of 0.269 (SD = 0.12). PPF corresponded well with geographic proximity of populations (Mantel tests genetic and geographic distance correlation: r = 0.47, p = 0.002), suggesting a strong spatial pattern of isolation by distance. In accordance with previous results, the population from Sweden (east of Baltic sea) was more similar to the central European populations than to the North-Eastern populations.