Comparative analysis of microarray, RISA and sequencing approaches for microbial community analysis

Various methods for the rapid analysis of the microbial community structure in different ecosystems compete with the more acceptable cloning and sequencing of the 16S rRNA gene (rrs). These methods often provide the benefit of rapid fingerprinting that can be used in statistical analyses of microbial (more often bacterial) community shifts due to either environmental perturbations or spatial and temporal shifts. With the reduction of cost for sequencing, the relative benefits will soon be reduced to the issue of analysis speed and the quantity of information produced. We examined the relationship between the community structures of different ecosystems as a function of the analytical method applied. Three methods were used: microarray (phylochip), the ribosomal intergenic spacer analysis (RISA) and the cloning and sequencing of the rrs gene. In addition, some samples were sequenced by pyrosequencing to remove the rrs amplification step from the sequencing bias. Correlations between the different analyses demonstrated that at a certain ecosystem level, they all provided sufficient information to differentiate between microbial communities. On the other hand, the relative sensitivity and sample separation power were not at all similar. The gel separation technique, RISA, was dependent on the number of bands produced and the likelihood of several bacterial species to be found in the same band. The cloning and sequencing approach provided the most informative data but in our case lacked significant clones to compete with high density microarrays, which had detect limits on the order of 10-4 (or one out of 10 000). The lack of resolution for the electrophoretic based method (RISA) reduced its correlation coefficients when compared to the two other methods. Pyrosequence data, although producing smaller sequences than the Sanger sequenced rrs clones, provided the largest number of sequences for statistical comparisons between samples. Therefore, the future of pyrosequencing (increased sequence length, multiple tags, etc) might eventually eliminate the use of other community structure analysis methods.