Abstract : The genomes of prokaryotes have long been considered as very optimized and stable. However, recent studies have shown that they undergo high rates of changes, notably by Lateral gene transfers (LGT). Several bacterial genomes may contain more than 15% of recently acquired genes and trees built using different genes often support different phylogenies. This raises the question of the internal coherence of genomes and of the lasting of genes interactions during evolutionary times. Several authors now consider that LGT occur so often that the very concept of species history does not apply to procaryotes. It has been however suggested that certain genes, assuming key functions, may have undergone less lateral transfers during their history. We have tested this hypothesis using phylogenetic methods, and we have reconstructed a consensus phylogeny of bacteria using complete genomes data. Our results suggest that certain genes retain congruent phylogenetic information allowing the building of the bacterial tree. The nature of the recently acquired genes also raises several problems Indeed, several authors have noticed that these genes often display a tendency to be A+T-rich in comparison to their hast genomes. We have analysed the structuring of the G+C content of genes along the genome of all sequenced prokaryotes species and shown that certain mechanisms, probably related to the replication process, affect the G+C content of genes near the replication terminus and may thus be responsible for an overestimation of the amount of recently acquired genes. If, the importance of lateral transfers in prokaryotic evolution has been largely documented, we think that it may have been overestimated because of methodological problems. When detected by a phylogenetic method, recently acquired genes show characteristics close to non-functional and parasitic sequences (such as phage). This suggests that certain categories of genes may be specialised in lateral transfer.