NUMERICAL EVALUATION OF THE SPECIES GROUPINGS AMONG THE CLASSICAL PROPIONIBACTERIA

Numerical analyses were used to evaluate the species groupings among t he classical Propionibacterium species based on 147 strains from a dai ry and an anaerobic environment. A data set of 74 phenotypic character s, determined using standardized API systems, were included. The Jacca rd and Sokal and Michener coefficients as well as dendrogram distances were used to broaden the method of delineating the clusters. One-, tw o- and three-dimensional plots were created and each strain clustered in relation to its closest relatives. Numerical data at similarity lev els of S-J = 0.43 and S-SM = 0.71, respectively, as well as PCR/RFLP d ata, revealed a separation of the 4 classical Propionibacterium specie s, confirming the major species as documented by Cummins and Johnson ( 1986). Each major grouping consisted of 2 or more clusters and many we re linked at a lower level to loose lying strains. From the dendrogram groupings, it was also clear that even though these strains could be placed in 1 of the 4 classical species using the present identificatio n key, the clustering positions were in many cases different. The stra ins with the lower similarities were identified to a specific species, but in many cases this did not agree with the clustering position. In the cases of the strains clustered away from the type strains, identi fication was confirmed using the PCR/RFLP data and this aided in confi rming the identification status in terms of to which of the 4 classica l species the strain was affiliated.