Comparison of PFGE and multilocus sequence typing for analysis of Klebsiella pneumoniae isolates

Klebsiella pneumoniae represents an important nosocomial pathogen causing urinary, respiratory and blood infections (Brisse et al., 2006; Podschun & Ullmann, 1998). Hospital outbreaks due to K. pneumoniae are frequent and especially feared when caused by multidrug-resistant strains, such as extended-spectrum beta -lactamase producers (Paterson & Bonomo, 2005). DNA-based strain typing methods are used to distinguish K. pneumoniae clinical isolates in order to understand transmission patterns and to help management of hospital infections. Molecular serotyping, based on PCR-RFLP of the cps operon responsible for capsular polysaccharide expression, has a higher discriminatory ability than traditional K typing (Brisse et al., 2004), and ribotyping is also applicable to K. pneumoniae (Brisse & Verhoef, 2001). Nevertheless, the most commonly used method is PFGE analysis of macrorestriction fragments (Arlet et al., 1994). The main advantage of PFGE lies in its high discriminatory power (Hansen et al., 2002), but PFGE is technically demanding and requires a high level of coordination (e.g. http://www.cdc.gov / pulsenet) to achieve inter-laboratory reproducibility. In contrast, multilocus sequence typing (MLST) provides unambiguous data that are suitable for global epidemiology and evolutionary studies (Maiden et al., 1998). A MLST method was previously developed for K. pneumoniae, and analysis of nosocomial isolates showed that MLST can discriminate among epidemiologically unrelated isolates (Diancourt et al., 2005). However, the discriminatory power of MLST was not compared to that of PFGE. In our previous study (Diancourt et al., 2005), 28 isolates belonged to 11 groups that were not distinguished by MLST nor by ribotyping. Among these 11 groups, 5 comprised isolates from distinct countries or separated by large sampling times. For these apparently unrelated cases the isolates could be suspected as being genotypically undistinguishable due to an insufficient discriminatory power of MLST and ribotyping, rather than due to an undocumented epidemiological link We report here on the comparison of the discriminatory power of PFGE with previously reported methods.