In this study, we used molecular and morphological data to quantify differentiation between populations from coastal areas of the Baltic Sea and inland populations that have been isolated since the last deglaciation. In general, we find that Baltic coast and coastal lake populations are similar genetically and morphologically. Inland populations, on the other hand, show greater genetic and morphological divergence. The recent introduction of predators into some inland populations also appears to have altered the evolution of body shape and spine length as inland populations with predators varied considerably in spine length but showed much greater conservatism in body size than inland populations without predators. Taken together, these results suggest that nine-spined stickleback populations in northern Sweden are strongly influenced by a combination of recent glacial activity, isolation/drift, and natural selection.
Our results support the hypothesis that inland populations of nine-spined sticklebacks were isolated from coastal populations following the recent deglaciation of the Baltic Sea. Our first line of support for this hypothesis comes from the physical location of these sites which were among the first areas to be exposed after the retreat of the Scandinavian Ice Sheet circa 11 - 10 k cal. ybp [22, 23]. A plausible scenario is that these populations originated from Baltic stock and were isolated as the land uplifted from glacial rebound forming lakes and ponds [24, 25]. Additional support for this hypothesis is venerated by the observation that few natural populations of nine-spined sticklebacks are found in northern Sweden above or below the highest coastline exposed after the last glacial maximum .
In agreement with the hypothesis for isolation of inland populations, we found lower levels of genetic differentiation, as measured by Nei's D
ST and standardized F
ST, in coastal sites and very high levels of genetic differentiation including the fixation of different alleles in many inland populations. Accordingly, we found higher levels of genetic variation as measured by H
O, π, and h
s in coastal populations than in inland populations. These last two findings are consistent with a pattern of isolation and genetic differentiation of populations through genetic drift and/or founder effects. Based on our results, these decreases in H
O do not appear to be due to a recent bottleneck in inland populations, with the exception of VST where significant results of simulated genetic bottlenecking were detected. We also found that populations from coastal lakes appear to be genetically indistinguishable from those emanating from the Baltic Sea, suggesting a separation so recent that genetic differences have yet to become apparent. These results strongly suggest that coastal populations are closely related whereas inland populations have existed in isolation for quite some time, and that genetic drift is responsible for the divergence in allelic frequencies and fixation of certain alleles.
Our support for a reduction in genetic variability among inland populations echoes findings in freshwater populations of the closely related three-spined stickleback (Gasterosteus aculeatus) that generally have reduced genetic variability [29–31] as a result of isolation from presumably ancestral anadromous populations . Similar models for post-glacial colonization by nine-spined sticklebacks have also been proposed and confirmed for North American populations [8, 33]. A recent study of nine-spined sticklebacks by Shikano and colleagues demonstrated similar patterns of isolation from anadromous ancestral populations of this species in Europe . In their study, coastal populations from the Baltic sea had high levels of genetic variation as measured by allelic richness and heterozygosity when compared to populations from freshwater systems near the Baltic while higher levels of genetic differentiation, as measured by F
ST and D
A, were prevalent in freshwater but not in coastal populations . The authors concluded that these patterns of genetic variation and genetic differentiation are consistent with postglacial recolonization of freshwater habitats, and subsequent isolation reducing variation in these populations through genetic drift and founder effects. Since Shikano and colleagues' study encompassed sampling from a much larger geographic area than studied here, it appears that this pattern of isolation is common in the Baltic region and that recent glacial history has greatly affected the current distribution of these fishes.
Our results show many morphological similarities between Baltic coast and coastal lake populations of nine-spined sticklebacks, mirroring the pattern of low levels of genetic differentiation. Given that coastal lakes have been isolated from the Baltic due to land uplift over a relatively short period of time (< 100 yrs), it is not surprising that there has been little differentiation in morphology compared to the relatively greater morphological divergence in inland populations. Moreover, piscivorous fish such as pike and perch are not currently detected in two of the three coastal lakes  but were likely present in these habitats prior to and during lake formation. Therefore morphological divergence in the absence of predation may be expected in these populations in the future, but perhaps not over the short time since isolation from the Baltic.
Inland populations of nine-spined sticklebacks displayed higher morphological diversity as compared to coastal populations. We also found evidence of morphological differences in both body size and spine length with respect to the presence of fish predators. Additionally, we found highly divergent morphological variation in some inland populations. For example, the populations with the two most divergent morphologies, BN and RT, show the greatest differences in mean body size. The BN population also has a reduction in the number of dorsal spines and dorsal spine length, and all but one individual completely lacked pelvic spines. This pattern of pelvic spine loss and an overall reduction in spines has been demonstrated in several species of stickleback [16, 19, 34] and may be due to an ecological escape from predation pressure, an ion deficiency related to calcification and bone deposition and/or increased invertebrate predation pressure in these populations [16, 20, 35].
Nine-spined sticklebacks that exist in ponds where they are the only fish species can obtain much larger body sizes than their coastal counterparts, presumably because of the absence of fish predation combined with interspecific competition for resources and/or fecundity selection [11, 12]. Confirming this prediction, we observed a trend for individuals hailing from inland populations without predators to be larger, on average, than both Baltic populations and inland lakes with predators. A similar pattern of larger size encountered in populations that either lack predators or have non-gape limited predators has been shown in three-spined sticklebacks  and brook sticklebacks (Culaea inconstans) , strongly suggesting that predation limits body size in this and other species of sticklebacks.
Despite having a smaller range in body size than their counterparts that hail from predation-free lakes, nine-spined sticklebacks from predator lakes have much greater variation in spine length. Although spine lengths are significantly smaller in inland populations compared to coastal populations, there is a trend for inland populations with predators to show longer spines (per unit of body length), on average, than inland populations without predators. The implication, therefore, is that spine length has increased in these populations as a result of the recent introduction of predatory fish. However the great variation in spine length exhibited among these populations within this group suggest that these populations may be responding to predation pressure differently. This great variation spine length could be caused by differing predation regimes within these lakes due to differences in predator communities, population densities of predators, differences in the relative exposure time to predators, and/or the recent extinction of predators within at least one lake. Although our results strongly suggest a role of predation, we cannot rule out that other ecological factors not measured in this study such as water chemistry may affect morphological variability in predator lakes.
PST - FST comparisons
Quantitative comparisons between morphological and neutral genetic divergence as estimated by Q
ST - F
ST, have been used as metric to investigate the potential for natural selection to influence morphological variation in many populations or species [38, 39]. The analogous comparison based solely on phenotypic data (i.e., P
ST - F
ST) has been highly criticized, particularly because it is difficult to tease apart the variance in phenotype attributable to environmental or genetic effects in wild populations. Thus, some advocate that Q
ST - F
ST comparisons should only be performed under controlled conditions in common garden experiments . While such approaches are preferable, we argue that there can be some value to P
ST - F
ST comparisons in natural populations. For example, in a recent meta-analysis that compared estimates of Q
ST - F
ST from different types of studies, estimates of P
ST from wild populations do not yield higher estimates than studies that use either broad or narrow sense estimates of additive genetic variation . Secondly, studies of P
ST in wild populations show meaningful variance among populations where a common garden approach may not be easily applied [27, 29, 41]. For example, in order to quantify additive genetic variation in our study of 18 natural populations with all potential crosses and multiple family groups taken to the F
2 generation would be a feat of herculean proportions in terms of time, scale and expense. Finally, in light of the criticism of P
ST as a substitute for Q
ST, it should also be kept in mind that common garden estimates of Q
ST may be inappropriate to compare to F
ST because the genetic basis of the phenotypes on which selection may potentially act may be partly genetic but non-additive (i.e. epigenetic), or environment-dependent. Thus we believe that P
ST - F
ST studies in natural populations do have some merit although we advocate caution in its interpretation.
Acknowledging the aforementioned concerns, we compared P
ST estimates for both spine length (PC1) and body size (PC2) within groups to estimate the relative influence of natural selection and genetic drift on the evolution of morphological phenotypic variation within our groups of interest. Despite the great morphological similarity within the Baltic coast group, it demonstrated P
ST > F
ST in both spine length and body size. Given the high gene flow experienced within this group, the most likely explanation is that these slight morphological differences could be explained by phenotypic plasticity in response to local environmental variation. However it should be noted that our F
ST estimates are so low that any variation in morphology would likely be greater than genetic divergence and so this result should be viewed cautiously. In coastal lakes, P
ST ≈ F
ST for these estimates making distinctions between selection and genetic drift equivocal. In inland lakes without predators, we found a robust F
ST - P
ST pattern of convergent evolution and stabilizing selection on reduced spine length strongly suggesting that reduced spine length is advantageous in single species lakes, potentially to aid escape of fish from invertebrate predators. A robust pattern of divergent selection and local adaptation on body size is also evident in inland no-predator lakes. Finally, great variation in body size and spine length within the inland predator group may also be indicative of divergent selection on these traits in these populations but confidence intervals are wide making a realistic determination difficult. Taken together, these results show opposite and robust patterns of convergent versus divergent selection on spine length based on the presence or absence of predation, strongly implicating the recent exposure to predators as a significant factor shaping phenotypic differences between these populations.
The degree to which our main focal phenotypic characters, body size and spine length, differ in their phenotypic response to selection is not currently known in nine-spined stickleback but may vary among different populations due to different environment conditions and standing levels of genetic variation [3, 5]. Studies on other fish species show that fish predators potentially can induce phenotypic changes in body shape and morphology without necessarily changing the background genetic structure [42–44]. For example, one study showed that cues of a predatory fish induced a deeper body and longer dorsal spines in a sunfish (Lepomis gibbosus) . The genetic component of the traits studied here remain unknown as previous studies investigating genetic differences among populations of nine-spined stickleback using a common-garden approach have not yet reported heritablities for these traits [13, 45]. Among three-spined sticklebacks, morphological traits of predator-naive and predator-sympatric populations demonstrate high values of heritability for body shape (h
2 = 0.67, 0.94) and length (h
2 = 0.92, 0.82) and moderate values for relative spine length (h
2 = 0.34, 0.39) . Other studies of three-spined sticklebacks have shown that body size is moderately heritable (h
2 = 0.42)  or that heritability may vary from negligible (h
2 = 0.007) to moderate (h
2 = 0.313) due to environmental factors such as rearing salinity . Taken together, these results suggest that in most cases morphological traits such as body size and spine length are moderately to highly heritable in three-spined sticklebacks, but that the ratio between genetic and environmental variance (and hence the heritability) may not be equal across populations.