Strain typing is an important tool for infection control investigations in hospitals, particularly to confirm suspected outbreaks of disease. However, many clinical laboratories do not have the ability to perform strain typing studies on site, often because they lack either the specialized equipment or the technical expertise. PFGE is often considered the gold standard for epidemiologic studies of MRSA because of its high discriminatory power and the strong correlation between PFGE strain types and epidemiologic data (7
). Unfortunately, PFGE typing is performed in few clinical laboratories in the United States and sending isolates to a state health department or other reference laboratory can be time consuming and, in the latter case, expensive. The DL system provides a commercial alternative to PFGE for typing both MRSA and methicillin-susceptible Staphylococcus aureus
) that is simple enough to be performed in most clinical microbiology laboratories.
Our study confirms that DL, although sometimes less discriminatory than PFGE for typing MRSA, can provide information that is useful for infection control investigations in hospitals. Typing of 15 isolates can be completed in as little as 4 hours, and isolates that are unrelated by DL are typically unrelated by PFGE. Thus, DL can be used effectively to rule out outbreaks of MRSA in hospitals. Similar data regarding the ability of DL to discriminate among MRSA strains have been reported by Ross et al. (14
), Shutt et al. (15
), and te Witt et al. (17
). All three studies showed that DL is less discriminatory than PFGE.
A recent report by Limbago et al. (11
) describing the PFGE types of MRSA isolates obtained during a population-based study of invasive MRSA infections in the United States in 2005 to 2006 confirms that USA100 and USA300 are the predominant MRSA PFGE types in the United States (8
). In our study, USA100 isolates formed a single large cluster of ~93% similarity, just below the >95% cutoff value defined by DL for “relatedness.” The USA100 type strain pattern from the DL MRSA library was present in the main USA100 cluster, suggesting that at least the USA100 PFGE type could be predicted based on high similarity to the type strain in the DL library. USA300 and USA500 isolates clustered together on the DL dendrogram, which is consistent with recent data suggesting that USA300 isolates are derived from an USA500 ancestor (10
). These two PFGE types can be differentiated by using the similarity matrix and the pattern overlay function. Thus, by testing the appropriate USA100 and USA300 control strains and using the similarity matrix and pattern overlay functions, it may be possible to use DL to predict the likelihood that an isolate is USA100 or USA300. This would not be true of USA400, USA700, or USA800, however, since the type strains for these patterns clustered together.
The importance of using the similarity matrix and the pattern overlay functions to compare pairs of isolates that are suspected to have the same strain type is one aspect of the DL software that has not been emphasized in prior studies. In the current study, this aspect was particularly important for differentiating among USA300 and USA500 isolates, which often clustered together in DL dendrograms at >95% similarity. The PFGE patterns of USA300 and USA500 isolates are similar but <80% related using Dice coefficients and the unweighted pair group method for analysis (12
), even though both types belong to MLST clonal complex 8. Microbiologists have grown accustomed to using dendrograms for displaying PFGE data (12
) and often share dendrograms with infection control personnel as a simple way of conveying information about which isolates from a suspected outbreak of disease are related. While the dendrograms provided by the DL system are helpful for defining broad categories of relatedness among isolates, obtaining the greater detail needed for epidemiologic investigations to confirm strains that belong to the same type requires the use of both the similarity matrix and the pattern overlay function.
A unique feature of this study was the analysis of strains not previously typed by PFGE that were obtained from purported outbreaks of MRSA in hospitals. Among the MRSA outbreak isolates, the PFGE patterns of two of four sets reflected greater pattern diversity than did DL, more so than would be expected during a hospital outbreak lasting 2 or 3 months. The pattern overlay and matrix similarity functions also showed the isolates to be indistinguishable. Because the DL patterns were indistinguishable, it is likely that the strain typing data rather than the epidemiologic data were used to define the outbreaks. The diversity of the PFGE typing results using the interpretive criteria of Tenover et al. (16
) argues against the likelihood that these isolates were from a true outbreak of disease. The variability of the PFGE patterns is more consistent with the variations seen in an endemic MRSA strain over time. The differences between the DL and PFGE data for outbreaks A and B emphasize the importance of analyzing the strain typing results in conjunction with epidemiologic information rather than relying solely on the typing data to define an outbreak. This approach poses a potential problem for reference laboratories that receive bacterial isolates for typing without any accompanying epidemiologic data. Thus, the results of typing studies conducted with DL should be reported with caution if no epidemiologic data are available. In most cases, a group of isolates with different DL patterns (i.e., well below the <95% similarity cutoff value) will not be clonal, i.e., are not likely to be part of an outbreak, and can be reported as unrelated with confidence. However, a group of isolates with indistinguishable patterns, as shown here, will not always represent the same USA type or clonal group and thus may not indicate an outbreak. If epidemiologic data are available, and those data provide a link among a group of isolates (e.g., a point source for the organisms has been identified), indistinguishable DL patterns of the isolates provide useful additional evidence that an outbreak exists.
In conclusion, DL is useful for screening isolates of MRSA for potential outbreaks, especially when the analysis includes the use of both the similarity matrix and pattern overlay function to differentiate USA300 isolates from USA500 isolates. Isolates that are unrelated by DL are likely to be unrelated by PFGE, based on our data and those from previous studies (14
). However, additional typing may be necessary to confirm strain relatedness or to identify a specific USA strain type or other MRSA type. Thus, PFGE remains superior for discriminating among outbreak strains, confirming strain relatedness, and identifying specific USA types.