To perform pyrolysis mass spectrometry (PyMS), a small sample of colony material is thermally degraded in a vacuum, thus yielding mixed volatile products whose composition reflects that of the organism. Mass spectrometric analysis of the products yields spectra that can be compared statistically; consequently, fine differences in cell composition can be detected [1]. Typing by PyMS is rapid, inexpensive, automated, and non-species-specific and has a high throughput (100-150 isolates per day) [2]. It has been used successfully to type isolates recovered during nosocomial outbreaks of Clostridium difficile infection [3] and now is a routine typing method for this organism in the United Kingdom. However, PyMS is a fingerprinting method; no permanent type designation has been assigned, and variations between different analytical batches prevent interbatch comparisons of spectra. This difficulty in routine typing of C. difficile is overcome by including representative isolates from any outbreak with further batches of isolates recovered in the same hospital. Our aim, for two reasons, was to develop the capability to predict C. difficile serotypes based on the scheme of Delmee et al. [4] from PyMS data. First, these serotypes correlate well with pathogenicity; isolates of groups G and H are commonly associated with nosocomial outbreaks, and serogroup F strains do not produce toxin A [5]. Furthermore, the results of outer membrane protein (OMP) profiles [6] and SDS-PAGE analysis of whole-cell protein [7] correlated well for these serogroups, thus indicating an underlying subdivision of the species that affects many of their properties. Second, assignment of a predicted serotype would, in effect, produce a permanent type designation for isolates. Nosocomial outbreaks of cross-infection due to most organisms tend to be of limited duration and scope, and the advantages of permanent type designations in these outbreaks are minimal. However, those investigators involved in the analysis of community outbreaks rightly criticize fingerprinting methods for their inability to assign the permanent type designations that are required for much broader temporal and geographic surveillance. Although permanent type designations
Read full abstract