Characterization of the USDA Poa pratensis collection usingRAPD markers and agronomic descriptors

Characterization of germplasm collections is critical to assesscollection diversity and enhance utilization. A Poapratensis L. germplasm collection of 228 accessionsrepresenting 26 countries, along with 17 commercial check cultivars,was characterized using 86 random amplified polymorphic DNA(RAPD) markers and 17 agronomic descriptors. The Dicesimilarity coefficient used for RAPD data ranged from 0.56 to 0.95and average Euclidean distance used for agronomic data ranged from0.28 to 2.52. No two accessions had a similarity of one or a distanceof zero, showing there were no duplicate entries. Cluster analysis ofRAPD data using the unweighted pair-group method usingarithmetic averages (UPGMA) revealed 11 accessions withparticularly low similarity values. These were subsequently found tobe misidentified Poa species (one each ofP. alpina,P. compressa,P. glauca,P. urssulensis and sevenP. trivialis). ForRAPD data, 62% of the entries were in one large cluster with46 additional clusters containing one to 13 accessions. For agronomicdata, 89% of the entries were in four main clusters. Thisclustering pattern for RAPD and agronomic data suggested uniquegenotypes were generally under represented in the collection. Theagronomic-based clusters showed some broad separation byaccession origin, but in general, origin did not correspond closelywith the clustering pattern. The correlation between the RAPD andagronomic-based distance matrices, excluding misidentifiedaccessions, was highly significant (P<0.01) (n= 234, r = − 0.14). However, the correlationrepresented a relatively small fraction of the total variation,indicating that both molecular and agronomic characterizations wereneeded to assess overall diversity.