Résumé
In the last 15 years, Quantitative Trait Locus (QTL) analyses have allowed locating genes controlling major agronomic traits on most crop genomes. The QTL approach has two major drawbacks: it obliges to develop a mapping population and, because linkage disequilibrium (LD) is strong in a mapping population, the confidence interval on QTL position is large with populations of normal size. It is generally not possible to determine which among candidate genes co-segregating with the QTL is responsible for the phenotype. Association studies on natural populations have proved to be a solution to this problem now commonly used in human and, more recently, plant genetics. The method, however, relies on the fact that LD decreases faster in natural populations because of a higher number of recombination events. Very little is known in rice about the decay of LD in relation with physical distance. Using a core collection of 217 accessions, we assessed the evolution of LD on the short arm of chromosome 6 around three genes of interest (wx controlling grain amylose content; Hd1 and Hd3a, two genes involved in heading date). Using the available rice sequences, we defined 52 SSR markers around the targeted genes, genotyped them using fluorescence technology and plotted multiallelic D' against physical distance. We tested several ways to deal with the very high number of alleles we observed. We repeated the same analyses for the indica and japonica sub-populations of the core collection. The distance, to which LD decayed, varied according to the region studied and could reach 350kb. Nevertheless this distance would allow using association studies to reduce the number of candidate genes underlying a QTL. (Texte intégral)