Application of numerical systematics to the phenotypic differentiation of legionellae.

A total of 163 strains, including 106 strains of Legionella pneumophila, 28 strains of Tatlockia micdadei, and 29 strains of other legionellae (including members of the proposed genus Fluoribacter), were studied. Ten tests which together could distinguish the genera previously proposed were identified. These tests included catalase-peroxidase, gelatinase, hippurate hydrolysis, starch hydrolysis, medium browning, acetoin production, oxidase, medium fluorescence, colony fluorescence, and the bromcresol purple spot test. T. micdadei strains were strongly catalase positive and bromcresol purple spot test positive and produced acetoin but otherwise were usually inert in the other tests. L. pneumophila and Fluoribacter species could usually be distinguished by strength of catalase activity, blue-white colony fluorescence (if present), and differences in frequency of hippurate hydrolysis, starch hydrolysis, yellow-green medium fluorescence, and, to a lesser extent, oxidase activity. With a simple algorithm and computer program, the overall accuracy was 98.8%.