A MULTIVARIATE-ANALYSIS OF PATHOGENIC VARIATION IN COLLETOTRICHUM-GLOEOSPORIOIDES INFECTING THE TROPICAL PASTURE LEGUME, STYLOSANTHES SCABRA

Multivariate statistical analysis was used to characterize and classif y pathogenic variation in isolates of Colletotrichum gloeosporioides t hat cause anthracnose disease of the tropical pasture legume, Stylosan thes scabra. A total of 182 isolates collected from field sites in Que ensland, Australia, over the past 15 years were tested for pathogenic variation on six differential genotypes of S. scabra using a seedling bioassay. Four reference isolates, representing the four pathogenic ra ces, were included in the bioassay for comparison. Data on the disease severity of 172 field and four reference isolates (set 1) were used t o classify the reference isolates into races and to determine if the f ield isolates belonged to an existing or new race. Linear discriminant functions were developed to classify the four reference isolates, and a cross-validation procedure was used to test the classification succ ess of placing these isolates into the four races. Isolate sr4 was cla ssified 76% of the time as race 1 and 17% of the time as race 4, isola te sr24 was classified 88% of the time as race 3, and isolates wrs20 a nd wrs32 were mainly classified as either race 4 or 4a. With one small cluster of weakly virulent isolates and the prior expectation of the four races, the field isolates were classified into five virulence gro ups using cluster analysis. Three of these clusters were associated wi th the existing races: race 1 in cluster 3, race 3 in cluster 1, and r aces 4 and 4a jointly in cluster 2. Cluster 1 isolates were avirulent on the differential cultivar Seca, cluster 3 isolates were virulent on 'Seca', and isolates in clusters 2 and 4 were virulent on accessions 36260 and Q10042. For an independent evaluation of the discriminant an alysis, additional data were obtained on 14 isolates (set 2), of which eight had been previously classified. The five set 1 clusters were us ed to develop linear discriminant functions to classify the isolates i n set 2. Five of the eight isolates common to both data sets were corr ectly classified; while isolates wrs20 and wrs32, previously in cluste r 2, were classified in cluster 4 in set 2. However, clusters 2 and 4 were close neighbors with no striking differences in the overall disea se severity levels on the six differentials for the isolates. In futur e analyses, three races, represented by the reference isolates sr4, sr 24, and wrs20 and/or wrs32, may be used to account for the existing ra nge of pathogenic variation. The usefulness of the multivariate approa ch to classify field isolates into races, in order to ascertain if iso lates belong to an existing or novel race, was discussed.