From chestnut rose, a promising fruit crop of the Rosa genus, powdery mildew disease-resistant and susceptible genotypes and their F 1 progeny were used to isolate nucleotide-binding-site (NBS)-encoding genes using 19 degenerate primer pairs and an additional cloning method called overlapping extension amplification. A total of 126 genes were harvested; of these, 38 were from a resistant parent, 37 from a susceptible parent, and 51 from F 1 progeny. A phylogenetic tree was constructed, which revealed that NBS sequences from parents and F 1 progeny tend to form a mixture and are well distributed among the branches of the tree. Mapping of these NBS genes suggested that their organization in the genome is a ''tandem duplicated cluster'' and, to a lesser extent, a ''heterogeneous cluster.'' Intraspecific polymorphisms and interspecific divergence were detected by Southern blotting with NBS-encoding genes as probes. Sequencing on the nucleotide level revealed even more intraspecific variation: for the R4 gene, 9.81% of the nucleotides are polymorphic. Amino acid sites under positive selection were detected in the NBS region. Some NBS-encoding genes were meiotically unstable, which may due to recombination and deletion events. Moreover, a transposon-like element was isolated in the flanking region of NBS genes, implying a possible role for transposon in the evolutionary history of resistance genes.