Measuring species richness based on microbial community fingerprints: the emperor has no clothes.

Danovaro and colleagues (3) recently compared microbial community diversity and richness estimates obtained using automated ribosomal intergenic spacer analysis (ARISA) with those obtained using analysis of terminal restriction fragment length polymorphisms (T-RFLP) of 16S rRNA genes. While community fingerprinting methods such as ARISA and T-RFLP are useful for comparative analyses, they are not useful ways to assess the richness or diversity metrics of complex communities (4). Such methods are inherently limited by their detection threshold or, more precisely, by their dynamic range (5). Simply put, the number of peaks detected in either T-RFLP or ARISA assays will grossly underestimate the actual richness of any community with a long-tailed rank abundance distribution. Microbial communities, including aquatic and terrestrial bacterial assemblages (1, 2, 8, 10, 11), are generally found to approximate long-tailed distributions, such as lognormal, power law, or log-Laplace distributions (7). When taken in conjunction with differences in the contents of microbial communities, this property makes counting peaks in fingerprint patterns an unproductive exercise. Danovaro and colleagues also conclude “that ARISA is more accurate than T-RFLP analysis on the 16S rRNA gene for estimating the biodiversity of aquatic bacterial assemblages.” The basis of the authors’ conclusion is that ARISA profiles contain more peaks than T-RFLP profiles generated from the same community sample. We assert that neither method has been shown to be more or less accurate based on the data presented in the paper. ARISA and T-RFLP analysis provide different levels of resolution. This leads to differences in the number of peaks and in the evenness of detected phylotypes but not to an increase in accuracy so far as estimating microbial diversity. Moreover, while the two methods partition the diversity in the community differently, neither method yields distinguishable categories that correspond to named taxonomic levels, such as species or genera. Consequently, neither method can be considered better from a taxonomic perspective. While this letter is proximally in response to Danovaro and colleagues’ paper, it is also intended to address a wider swath of the microbial ecology literature (see, for example, reference 6), in which authors fail to meaningfully discuss how the exclusion of rare taxa by the detection limits of such assays affects the interpretation of their data (9). For reasons we do not understand, investigators seem reluctant to acknowledge the limitations of such methods. So we will say what we all already know: the emperor has no clothes (http://www.andersen.sdu.dk /vaerk/hersholt/TheEmperorsNewClothes_e.html); current microbial community fingerprinting methods cannot provide reliable diversity indices.