The rates and patterns at which nucleotide substitutions occur vary significantly across the genome sequence of vertebrates. A prominent example is the difference in the rate of evolution of functional sequences versus nonfunctional (neutrally evolving) sequences, which is explained by the influence of natural selection on functional sequences. However, even within neutrally evolving sequences there is striking variation in the rates and patterns of nucleotide substitutions. Unraveling the underlying processes that induce this variation is necessary to understand the basic principles of variation in neutral substitution profiles, which in turn is crucial for the identification of regions in the genome where natural selection acts. This research question builds the main focus of the present thesis. I have studied the causes and consequences of variation in different patterns of nucleotide substitutions. In particular, I have investigated substitutional strand asymmetries in mammalian genes and could show that they result from the asymmetric nature of DNA replication and transcription. Comparative analysis of substitutional asymmetries then suggested that the organization of DNA replication and the level of transcription are conserved among mammals. Further, I have examined the variation in CpG mutation rate among human genes and could show that beside DNA methylation also GC content plays a decisive role in CpG mutability. In addition, I have studied the signatures of GC-biased gene conversion and its impact on the evolution of the GC isochore structure in chicken. By comparison of the results in chicken to previous results in human I found evidence that karyotype stability is critical for the evolution of GC isochores. Finally, beside the empirical studies, I have performed theoretical investigations of substitution rates in functional sequences. More precisely, I have explored the temporal dynamics of estimates of the ratio of non-synonymous to synonymous substitution rates dN/dS in a phylogentic-population genetic framework.

Comprendre les déterminants du taux d’adaptation est une question primordiale en évolution moléculaire. En particulier, l’influence de la taille efficace de population sur la sélection positive, ainsi que la nature des changements d’acides aminés qui mènent à de l’adaptation sont des questions encore débattues. Pour y répondre, la méthode DFE-α, dérivée du test fondateur de McDonald & Kreitman, est un outil puissant pour mesurer le taux de substitution adaptatif. Elle est néanmoins sensible à certains biais. Au cours de cette thèse, nous avons identifié deux biais majeurs de cette méthode, les fluctuations de long-terme du régime de sélection-dérive via des fluctuations démographiques, et la conversion génique biaisée vers GC (gBGC). Via des simulations, nous avons montré que divers scénarios plausibles de fluctuations démographiques peuvent mener à une sur-estimation du taux de substitution adaptatif. Nous avons aussi obtenu des indications empiriques que le régime de sélection-dérive récent ne reflète pas le régime de sélection-dérive de long-terme chez diverses espèces animales, ce qui représente une violation d’une hypothèse forte de la méthode DFE-α. D’autre part, nous avons montré que la gBGC entraîne une sur-estimation du taux de substitution adaptatif chez les primates et les oiseaux. Via un jeu de données de neuf taxons de métazoaires et un total de 40 espèces, nous avons d’une part initié une analyse visant à identifier la nature des changements d’acides aminés qui mènent à l’adaptation, et montré que les changements radicaux sont soumis à une plus forte sélection purificatrice que les changements conservatifs. D’autre part, nous avons pu évaluer le lien entre la taille efficace et le taux de substitution adaptatif tout en prenant en compte les deux sources de biais explorées précédemment. Nous avons mis en évidence pour la première fois une relation négative entre le taux de substitution adaptatif et des traits d’histoire de vie représentatifs de la taille de population de long-terme. Ce résultat va à l’encontre de l’hypothèse canonique d’une adaptation plus efficace en grandes populations. / Understanding the determinants of the adaptive substitution rate is a central question inmolecular evolution. In particular, the influence of the effective population size N e on positiveselection as well as the nature of amino acid changes that lead to adaptation are still debated. TheDFE-α method, which was derived from the seminal McDonald & Kreitman test, is a powerful toolfor estimating the adaptive substitution rate. However, it is sensitive to various sources of bias. Inthis thesis, we identified two major sources of bias of this test, long-term fluctuations of theselective-drift regime through demographic fluctuations, and GC-biased gene conversion (gBGC).Using simulations, we showed that under plausible scenarios of fluctuating demography, the DFE-αmethod can lead to a severe over-estimation of the adaptive substitution rate. We also showed thatpolymorphism data reflect a transient selective-drift regime which is unlikely to correspond to theaverage regime experienced by genes and genomes during the long-term divergence betweenspecies. This violates an important assumption of the DFE-α method. Our results also indicate thatgBGC leads to an over-estimation of the adaptive substitution rate in primates and birds. Using adataset of nine metazoan taxa for a total of 40 species, we started an analysis aiming at identifyingthe type of amino acid changes that are more prone to adaptation, and evaluated the link between N eand the adaptive substitution rate while accounting for the two sources of bias previously explored.We reveal for the first time a negative relationship between the adaptive substitution rate and life-history traits representative of long-term N e . This result is in contradiction with the widespreadhypothesis that adaptation is more efficient in large populations.