Phylogeography and Population Genetics
In most populations, genetic diversities were not pronounced (Table 4; Figure 2). A few COI haplotypes are particularly common, and the majority of the other haplotypes differ by one or two mutations from one of the common haplotypes in each population and lineage (Figure 2). Most populations feature only a few COI haplotypes. This is the case especially within lineages ICOI, IICOIand IIICOI as well as the populations of IVCOI and VCOI along the Reykjanes Ridge. A notable exception is population 1072 (lineage VCOI) to the west of Iceland, where nine haplotypes were observed. Nucleotide diversities were roughly an order of magnitude larger in the ddRAD data compared to COI (Table 4). Statistics differed slightly between the two datasets for populations and lineages in IRAD, IIRAD, and IIIRAD, but not extensively (see Table 4). Therefore, we will focus only on the analyses of the dataset which included all five lineages.
Observed heterozygosity across all ddRAD sites was close to 0.1 for all lineages and populations (Table 4). The level of inbreeding differed among lineages. Within lineages IRAD-IIIRAD and IVRAD,the inbreeding coefficient FIS was ≤ 0.14 for each population, but 0.18 for VRAD (Table 4). Due to the Wahlund effect, FIS increased when populations were grouped into lineages IRAD, IIRAD and IIIRAD, or all together (up to 0.27).
Most lineages appear to be geographically restricted and separated from each other. Lineage ICOI is the most widely distributed lineage ranging from the south-east to the north-east of Iceland around a distance of 900 km. However, the population farthest into the Nordic Seas (station 1159) is well-differentiated from the others with at least five mutations separating the observed COI haplotype from all other haplotypes (Figure 2). Among and within all lineages, populations are strongly differentiated from each other. In the nuclear ddRAD data, population differentiation within lineages is not well resolved in the PCA and Structure analyses (Figure 3), however, the coancestry analyses with fineRADstructure grouped most individuals by populations within lineages, suggesting low but detectable levels of population differentiation (Figure 3). Population differentiation is particularly strong in COI, where the vast majority of populations are differentiated by significant FST values larger than 0.7, often over 0.9 (Table 5). High and significant FST values were also observed among several geographically close populations (<100 km apart), e.g. station 879 compared to 880 or 881. A few instances of geographically distant but genetically similar populations in COI were observed as well: 869 compared to 873 (210 km; lineage ICOI), 880 and 881 compared to 1172 (~450 km; lineage ICOI), and to a lesser degree 83 compared to 1072 (431 km; lineage VCOI) (Table 5). ddRAD FST values were lower with all values within and among populations of lineages IRAD-IIIRAD being ≤ 0.29 (Table 5). Isolation-by-distance was observed in COI, though only a relatively small fraction of the genetic differentiation was explained by geographic distance (correlation coefficient rY1 = 0.56; p = 0.001; determination of genetic differentiation by geographic distance = 31.3 %), but not in the ddRAD data (correlation coefficient rY1 = 0.36;p = 0.15; determination of genetic differentiation by geographic distance = 13.2 %). The high levels of genetic differentiation are already observed at low geographic distances, which may explain the low correspondence between genetic differentiation and geographic distance.
The discordance between mitochondrial and nuclear data impacts the inferred geographic distribution of genetic lineages. It is noteworthy that ICOI extends much farther to the north-east than IRAD, whereas IICOI is much more restricted to the near shelf of Iceland than IIRAD, which extends farther to the east into the Arctic Ocean. The difference at station 1219 is the most pronounced, which has been assigned to IICOI, but in ddRAD appears to be a hybrid between IRAD and IIIRAD, (Figure 2), which otherwise occur farther south or north, respectively.
Demographic parameters (Tajima’s D and Fu’s Fs) are slightly negative for most populations and lineages in COI. Notable exceptions are IVCOI (Tajima’s D and Fu’s Fs positive) and ICOI-IIICOI jointly analyzed (Fu’s Fs is positive). In the nuclear data set, most single populations had positive values close to zero (thus also lineages IVRAD and VRAD), however, lineages IRAD, IIRAD, IIIRAD and IRAD-IIIRAD each had negative values ranging from -0.13 to -0.59. The three EBSP analyses yielded slightly diverging outcomes concerning the long-term trends. While in COI a slightly decreasing population size over time was suggested, one of the ddRAD-based EBSPs suggested long-term stability and the other a temporal increase in population size. However, all three analyses suggest that population sizes were strongly reduced shortly before the present followed by a rapid recovery of population sizes, likely an expansion following a bottleneck (Figure 5). It is difficult to estimate when this bottleneck may have occurred due to the lack of a conclusive specific substitution rate for isopods.