Résumé
In rice, a deep and thick root system is generally considered to be a useful trait for maintaining yield under water stress in a broad range of conditions, notably in rainfed ecosystems. A large natural variation in root architecture is observed in rice due to the specific adaptation of groups of rice cultivars to contrasting hydrological conditions (irrigated, rainfed, or upland). Understanding the genetic and molecular mechanisms controlling the development of the root system and its adaptive plasticity in response to water availability and soil environment would help in improving yield stability of rice crops confronting drought situations. This paper reviews current knowledge on QTLs detected for constitutive and adaptive root development, the present status of QTL integration on a consensus QTL map linked to the rice physical map through sequenced markers, and QTL validation in near-isogenic or substitution lines. A key step is now to link QTLs to genes, which can be done through several approaches. We are presently exploring some of them: looking for QTLs for physiological parameters derived from models, fine-mapping QTLs for root depth using meta-analyses and recombinant genetics, searching for known root architecture and stress-response genes by direct genetics in an insertion mutant collection, and searching for orthologs in rice of genes involved in root development in other species. We will combine these approaches with association studies between polymorphism within validated candidate genes and phenotypic variation that will confirm the interest of the genes in the target plant material, and, in addition, will give access to a range of alleles at the genes. We think that these combined approaches could hasten the discovery of important genes and alleles involved in root traits, provided that a high-throughput reliable phenotyping technique linked to field performance, which is presently missing, is developed. This will provide breeding programs with markers allowing the combination of favorable alleles at key loci for root traits. Although genotype building involving a few loci is now well under control, we are still lacking experience in how to accumulate alleles at many loci in the most efficient way. A marker-aided recurrent selection strategy could help achieve this goal.