Lack of structure in the gene pool of the highly polyploid ornamental chrysanthemum

Selection of clonally propagated chrysanthemums is mostly performed on F1 hybrids using phenotypic characteristics without the use of molecular information. We applied 448 amplified fragment length polymorphism markers to a set of 81 accessions, mainly from the European gene pool, covering the different horticultural types (cut, pot and garden varieties) and originating from the most important European chrysanthemum breeders. The average pairwise genetic similarity of 0.69 was moderate to rather high. Neighbour-joining clustering resulted in no grouping of the accessions, either by their common origin or their horticultural type, or by similarities in important phenotypic characteristics. The structure of the dendrogram could not be supported by bootstrap analysis. Furthermore, network analysis using SplitsTree, principal coordinate analysis via DARwin or analysis of the population with structure did not differentiate reliable and invariable clusters. Therefore, we tested the marker saturation by plotting the mean coefficient of variation for every pairwise similarity of the bootstrap analysis against the different numbers of markers. We showed that the number of markers is sufficient for a precise estimate of genetic similarity and that the lack of bootstrap support is not due to a low genetic diversity or a lack of marker information, but most likely resulted from the breeding history of the cultivars, involving repeated backcrosses and the exchange of genotypes between breeders.