Considerations in evaluation of the applicability of DNA fingerprinting techniques for species differentiation.

Recently, Koeleman et al. (6) reported that amplified ribosomal-DNA (rDNA) restriction analysis (ARDRA), i.e., PCRrestriction fragment length polymorphism analysis of parts of the rRNA cistron, is not suited for (i) strain differentiation within Acinetobacter baumannii and (ii) species differentiation within the genus Acinetobacter, while random amplified polymorphic DNA (RAPD) analysis and selective restriction fragment amplification (AFLP kit, Keygene, Wageningen, The Netherlands) perform well for both purposes. On the basis of our experience in strain differentiation (1, 3, 5, 11) and species differentiation (2, 5, 7, 10) of strains of the genus Acinetobacter by different techniques, including those compared by Koeleman et al. (6), we are puzzled by some of the results reported. First, a study of the capacity of methods to differentiate at or below species level requires the inclusion of several strains per species. Since only 1 strain for each described genomic species was investigated (6), together with 13 clinical isolates, conclusions on the applicability of the techniques for species differentiation are impossible. Second, the conclusions with regard to the strain differentiation capacities of RAPD analysis and ARDRA are not surprising. RAPD analysis is generally applied for typing, including within the genus Acinetobacter (3, 8, 11), while ARDRA is aimed at species but not strain differentiation (2, 7, 9, 10) as it relies on the conserved nature of rDNA. ARDRA aims at overlooking intraspecific differences, which facilitates the recognition of species-specific patterns. Thus, the fact that ARDRA is not suited for typing is well established. Contrary to the conclusion of Koeleman et al. (6), data from the literature suggest that RAPD analysis is not suited for species differentiation, exactly due to the large degree of intraspecific variability of RAPD fingerprints. Applications of RAPD analysis therefore deal with strain differentiation, including for Acinetobacter (e.g., see references 3, 8, and 11). With respect to the species differentiation capacity of ARDRA, the conclusions of Koeleman et al. (6) are in apparent contradiction with reports which show that ARDRA allows for rapid and unambiguous species differentiation within the genus Acinetobacter (2, 7, 10) and other genera (9). We showed that all 202 strains belonging to 18 genomic Acinetobacter species can be distinguished, leaving only A. haemolyticus and DNA group 13BJ/14TU unseparated (2). ARDRA differentiates easily between the four species of the A. calcoaceticus-A. baumannii complex (2, 10), three of which are of clinical significance and which are difficult to distinguish phenotypically. Two explanations for the disagreement between the ARDRA results of Koeleman et al. (6) and those of others are possible. Although no original photographs of the gels are presented, visual interpretation of the digitized gel pictures shows that the densitometric patterns are similar to those described previously (2, 10). However, several additional weak peaks are present and the authors state that up to 10 restriction fragments per enzyme were obtained, which is significantly more than in other studies (2, 7, 10). This suggests that some bands either are due to incomplete digestion or are artifacts of the digitization procedure. The latter occurs when the software interprets background staining or spots as genuine fragments, and this requires careful comparison of the number of bands on the original gels with that on the digitized pictures. Furthermore, the calculation of similarity between densitometric curves by means of Pearson product-moment correlation, as done by Koeleman et al. (6), is not appropriate for ARDRA patterns (4), which consist of discrete bands—most often of comparable intensity—and which favor the application of band-matching coefficients such as the Dice coefficient (4). Heyndrickx et al. (4) also used combined restriction patterns for numerical interpretation with GelCompar software (Applied Maths, Kortrijk, Belgium) and showed this approach to be useful, on the conditions that the digital pictures are checked for artifacts and that the Dice coefficient is used. In conclusion, there are sufficient arguments to support the applicability of ARDRA (2, 7, 10) for species differentiation, whether or not computer analysis is used. In addition, the conclusion of Koeleman et al. (6) that RAPD analysis is applicable for species differentiation is also in contradiction with most literature and cannot be justified, since not enough strains were included. Finally, the typing and identification possibilities of AFLP have been established previously for Acinetobacter by using a large number of strains (5), but one should be cautious when comparing the capacities of a laborious, expensive, and commercialized technique (AFLP) with those of rapid and simple approaches such as ARDRA and RAPD analysis. We would regret it if the interpretation of Koeleman et al. (6) discouraged researchers from using ARDRA, since this method has been shown to be of great help in the elucidation of the ecology and taxonomy of species of Acinetobacter and many other genera.