Eight strains each of Bacillus subtilis, Bacillus pumilus, Bacillus licheniformis, and Bacillus amyloliquefaciens were analyzed by using pyrolysis gas-liquid chromatography. Statistical analysis with canonical variates gave four well-separated groups, which represented the four species. Further analysis of the same strains by deoxyribonucleic acid-deoxyribonucleic acid hybridization and API identification systems confirmed the discrimination obtained with pyrolysis gasliquid chromatography. However, analysis by biochemical tests performed in the classical way gave only three groups since it was not possible to achieve separation of the strains representing B. subtilis from those of B. amyloliquefaciens when these tests were used.Pyrolysis, a process whereby molecules are thermally degraded in an inert gas atmosphere, has enhanced the use of conventional gas-liquid chromatography by enabling nonvolatile compounds to be analyzed. Pyrolysis gas-liquid chromatography (PGLC) was fmt proposed as an approach to microbial differentiation by Oyama (15) during the development of a system aimed at detecting life on Mars. However, its potential in microbiology was not appreciated until Reiner (17) was able to distinguish different species of Mycobacterium and different serotypes of Escherichia coti in a reproducible manner. Since then, PGLC has been used in the differentiation of numerous types of bacteria (10, 18, 22) and fungi (5,231. The recent application of PGLC to aerobic sporeformers by Oxborrow et al. (12)(13)(14) indicates that, providing the cultural and chromatographic conditions remain constant, PGLC can be applied usefully to the characterization of bacilli.The variation between pyrograms of the same strain and the high level of redundancy found in PGLC data require the application of data processing techniques capable of highlighting significant variations in the heights of specific peaks. Several methods for handling data in this t Present address: School of Chemistry, The University, Newcastle-upon-Tyne, NE1 7RU, United Kingdom.way have been described (10, 16), but as yet there is no agreement on the best statistical approach, and much work remains to be done in this field. This paper reports on the usefulness of lowresolution PGLC when coupled to multivariate data analysis for differentiating closely related groups of bacteria and provides evidence for the separation of Bacillus antyloliquefaciens from Bacillus subtilis.
MATERIALS AND METHODS