Résumé
Leaf scald (LS) is a major sugarcane disease, resulting in cane loss and sugar yield decrease when growing susceptible cultivars. This disease is caused by a xylem invading bacterium, Xanthomonas albilineans (Xa), and is mainly transmitted by aerial contamination in tropical humid climates. Cultivar resistance is the most efficient methods used to limit pathogen spread and yield loses. In order to tag alleles controlling LS resistance, a genome-wide association study was undertaken based on a survey 189 interspecific sugarcane cultivars (Saccharum spp.) representing the worldwide diversity. This panel of cultivars was phenotyped for LS resistance under natural infestation conditions, in Guadeloupe (French West Indies) in two field trials, each consisting of a complete randomized block design with 3 replications. Two successive crop cycles were analyzed for each trial. Disease Severity (DS) was calculated by measuring symptoms on 10 stalks per replication (scale 0-5 related to the amount and length of foliar necrotic lesions) with DS =[sigma]symptoms/50. The DS obtained was normalized using arc-sinus root-square transformation. BLUPs (Best Linear Unbiased Predictors) were then calculated for each cultivar. The cultivar panel was analyzed with AFLP and DArT markers resulting in 2576 polymorphic markers (AFLP + DArT). Genetic structure of the population was analyzed by a Principal Component Analysis using a subset of markers not tightly linked and a Tracy-Widom test (SASV9.2) for the significance of each component. Associations between LS resistance and each of the 2576 polymorphic markers were tested using BLUPs values in a General Linear Model (GLM) taking into account population structure using TASSEL software. Based on permutation tests, markers were declared significant at a genome-wide type 1 error of 0.05. LS phenotyping results shows that clone x crop-cycle interaction or clone x trial interaction are both highly significant compared to residual effects (P<0.001), but clone effects always appear highly significant compared to these interactions (P<0.001). Congruency of these findings with the markers detected in each trials and crop cycles (respectively 9, 7 ,14 and 12 markers detected for trial one crop cycle 2 and 3, trial 2 crop cycle 2 and 3) is discussed. Four markers were detected in both trials and shows that quantitative dissection of alleles controlling LS resistance is feasible in a humid ecology such as the one of Guadeloupe. 1600 additional markers should be tested soon for association. These should open the construction of pan-genomic-information based models designed to predict the behavior of breeding material toward Xa resistance infection to save traditional tedious indexation efforts. (Résumé d'auteur)