Within- and between-population comparisons suggest independently acting selection maintaining parallel clines in Scots pine (Pinus sylvestris)

Abstract Parallel clines in traits related to adaptation in a species can be due to independent selection on a pair of traits, or due to selection in one trait resulting in a parallel cline in a correlated trait. To distinguish between the mechanisms giving rise to parallel adaptive population divergence of multiple traits along an environmental gradient we need to study variation, correlations, and selective forces within individual populations along the gradient. In many tree species, budset timing (BST) forms a latitudinal cline, and parallel clinal variation is also found in other seedling traits, such as first-year height (FYH) and fall frost injury (FFI). In this study, we set up a common garden experiment with open pollinated progeny from natural populations of Scots pine (Pinus sylvestris), with one large sample from single population (500 families) and smaller samples from across a latitudinal gradient. BST, FYH and induced FFI were first measured in a greenhouse. The seedlings were then planted in the field, where survival and height were measured at the age of 9 years as fitness proxies. We compared between- and within-population variation and genetic correlations of these three seedling traits, and estimated selection gradients at the family level in our main population, taking into account the potential effects of seed weight. Between-population genetic correlations between seedling traits were high (0.76–0.95). Within-population genetic correlations in the main population were lower (0.14–0.35), as in other populations (0.10–0.39). Within population, extensive adaptive variation persists in the seedling traits, in line with rather weak selection gradients, yet maintaining the clines. Although our sampling does not cover the whole cline equally, the results suggest that the individual clines in these traits are maintained by largely independently acting selection, which results in fewer constraints in adaptation under changing climate.

The positive effect of seed weight on seedling height in early years has been reported many times (e.g., Mikola 1985, Wennström et al. 2002, Reich et al. 1994).In Maritime pine heritable genetic variation was observed for seed weight (Zas & Sampedro 2015).Ramirez-Valiente et al. 2020 found evidence for selection for larger seed mass in an analysis where multiple populations from different latitudinal origin were combined to estimate selection.This dependency should be kept in mind when interpreting selection effects on FYH.
Considering the speed of current environmental change, the long generation time of Scots pine obviously slows down adaptation (e.g., Savolainen et al., 2004).The negative correlation between mother tree's age and first year seedling height, can result from younger trees having slightly heavier seed.Through the observed selective advantage of tall seedlings, alleles contributed by young trees could have some competitive advantage.This association might have some influence on shortening the generation time and accelerating adaptation.

Supplementary Material: Early effects of selection
Here we examined in more detail the mortality/selection effects on the very first years of the seedlings lives.We first compared the distributions of BST (budset timing) and FYH (first year height) between groups of seedlings scored dead and alive at different timepoints (at 2010 inventory, at 2011 inventory, and between these two inventories).Each population was analyzed separately.Differences in group means were tested with Welch Two Sample (twosided) t-test.
We found that by 2010, in the southernmost population samples, there was a trend of surviving plants to be shorter and have earlier budset than those that died, consistent with selection by frosts.In the focal population and other Finnish populations, the survivors were taller, suggesting general vigor.During the year after planting (2010-2011), the trend was mostly similar.The high overall mortality within this year is likely due to random environmental factors that do not correlate strongly with the focal traits (see Lönnroth 1925).We also took a closer look at the relative importance of selection during the very first years versus later years.For this, we repeated the two-step Lande-Arnold analysis (Lande & Arnold, 1983;Phillips & Arnold, 1989, see main text) by omitting seedlings that died before planting in the field.We found that the signals of selection were weaker with this subdata, indicating that selection on these traits acts rather early.Young Scots pine seedlings are susceptible to frost damage at early ages (Luoranen et al. 2018) and importance of early selection has been found also e.g. in Pinus contorta (Warwell & Shaw, 2019).Table S3.Standardized linear selection gradients (β) from the linear model and standardized quadratic and correlational selection gradients (γ) from the full quadratic model, excluding seedlings that died already before planting to field (dead in 2010 inventory).β represent the strength of directional selection on the traits.Diagonal γ (in bold) describe the strength of stabilizing/disruptive selection on the traits.Off-diagonal γ express the selection on the pairwise correlation between traits.Above: gradients when family fitness was defined through both survival and height at field.Below: gradients with mean survival of the family as the fitness.Significance from the respective multiple regression model: .P<0.10, * P < 0.05, ** P < 0.01, *** P < 0.001.
for linear terms for directional selection) and gamma (coefficients for quadratic terms for stabilizing and correlational selection).Estimates for stabilizing selection were doubled (Stinchcombe et al. 2008).To evaluate the relative importance of survival versus tree height as a fitness estimator, we also tested a model excluding the tree height: (Root) → Survival 2010 → Survival 2017 Further, we ran the full model with a subset of data (n=4463, omitting seedlings that died before field planting) to examine the significance of very early selection: (Root) → Survival 2017 → Height 2017 Directional selection was found for earlier BST and greater height (table S4, fig.S1).However, the very tallest seedlings did not have the highest fitness, resulting in stabilizing selection in FYH.When only survival was considered, the results were similar, but the coefficients were generally larger.When using only seedlings still alive at 2010 inventory/field planting (i.e., omitting selection on the first, seedling nursery years), directional selection on both traits was slightly weaker and no stabilizing selection was detected.The greenhouse block effect was statistically significant when early survival was included in the analysis, but not in later years.
The field block effect was statistically significant when tree height was included in the fitness estimate, but not when only survival was considered.Both block terms were however kept in all models for the sake of model comparison.The gamma term for interaction (i.e., correlated selection) was not statistically significant in any analyses.
At the family level (see main text) we did not observe clear signs of selection for timing of budset, nor stabilizing selection for FYH.The aster results here could thus be influenced by unobserved selection in FFI and SW.This result, however, is concordant with a hypothesis of southern late (and cold prone) alleles being selected against in the seedling population (see e.g., García-Ramos & Kirkpatrick, 1997).Southerly winds prevailing at the time of pollination could introduce nonlocal alleles, as female flowers can be receptive before local pollen shedding (Sarvas, 1962;Varis et al., 2009; see also Kling & Ackerly, 2021).

Figure S1 .
Figure S1.Variograms showing spatial autocorrelation in tree age, seed weight, and in the residuals of a linear model of weight vs. age in the Ranta-Halola subpopulation.
Selection gradients: βBST = directional selection in timing of bud set; βFYH = directional selection in first year height; γBST BST = stabilizing selection in timing of bud set; γFYH FYH = stabilizing selection on first year height; γBST FYH = selection on correlation between timing of bud set and first year height.

Figure S2 .
Figure S2.A graphical representation of selection on BST and FYH.The dots are the actual measures of the traits in the greenhouse.The contour lines represent the approximation of the individual selection surface from the aster analysis.Highest fitness is found among seedlings that are taller than average (but not the tallest) and have early budset timing.

Table S1 .
Welch Two Sample (two-sided) t-test for differences between the means of budset timing (BST) of dead and live seedlings at different timepoints.Population means were calculated from individual seedling data.The p-values are not corrected for multiple testing.

Table S2 .
Welch Two Sample (two-sided) t-test for differences between the means of first year height (FYH) of dead and live seedlings at different timepoints.Population means were calculated from individual seedling data.The p-values are not corrected for multiple testing.