Abstract
A major challenge in population genetics is to understand the local adaptation process in natural population and so to disentangle the various evolution forces contributing to local adaptation. The experimental studies on local adaption generally resort to altitudinal gradients that are characterized by strong environmental changes across short spatial scales. Under such condition, the genetic differentiation of the functional trait (measured by the Qst) as well as the genes coding for trait (measured by Fstq) are expected to be mainly driven by selection and gene flow. Genetic drift and mutation are expected to have minor effect. Theoretic studies showed a decoupling between Qst and Fst under strong gene flow and / or recent selection. In this study, I tested this hypothesis by combining experimental and modelling genomic approach in natural population of Fagus sylvatica separated by ~3 kilometres and under contrasted environments.Sampling was conducted in south-eastern France, a region known to have been recently colonised by F.sylvatica. Four naturally-originated populations were sampled at both high and low elevations along two altitudinal gradients. Populations along the altitudinal gradients are expected to be subjected to contrasting climatic conditions. Fifty eight candidate genes were chosen from a databank of 35,000 ESTs according to their putative functional roles in response to drought, cold stress and leaf phenology and sequenced for 96 individuals from four populations that revealed 581 SNPs. Classical tests of departure of site frequency spectra from expectation and outlier detection tests that accounted for the complex demographic history of the populations were used. In contrast with the mono-locus tests, an approach for detecting selection at the multi-locus scale have been tested.The results from experimental approaches were highly contrasted according the method highlighting the limits of those method for population loosely differentiated and spatially close. The modelling approach confirmed the results from the experimental data but revealed that up to 95% of the SNPs detected as outliers were false positive. The multi-locus approach revealed that the markers coding for the trait are differentially correlated compared to the neutral SNPs. But this approach failed to detect accurately the markers coding for the trait if no a priori knowledge is known about them. The modelling approach revealed that genetic changes may occur across very few generation. But while this genetic adaptation is measurable at the trait level, the available method for detecting genetic adaptation at the molecular level appeared to be greatly inaccurate. However, the multi-locus approach provided much more promise for understanding the genetic basis of local adaptation from standing genetic variation of forest trees in response to climate change.