DOES ECOSYSTEM SIZE DETERMINE AQUATIC BACTERIAL RICHNESS? COMMENT

In their paper ‘‘Does ecosystem size determine aquatic bacterial richness?’’ Reche et al. (2005) observed a significant correlation between lake surface area and lake bacterial OTU (operational taxonomic unit) richness in 32 lakes. The authors propose that this relationship corroborates one of the predictions of the island-biogeography theory, i.e., that larger islands support more species than smaller islands (MacArthur and Wilson 1967). The results of Reche et al. (2005) have already been cited in support of a positive relationship between habitat size and bacterial taxonomic richness (Bell et al. 2005, Dolan 2005). We argue that the study by Reche et al. (2005) does not provide support of a causal relationship between bacterial richness and habitat size, since their conclusions are biased by incorrect merging of data sets that are not comparable and because the methods used to determine bacterial richness are not adequate. The significant correlation between lake area and bacterial OTU numbers obtained by Reche et al. (2005) was based on data from three separate papers (Lindström and Leskinen 2002, Zwart et al. 2002, Reche et al. 2005). Treated separately, the data from Lindström and Leskinen (2002) and Zwart et al. (2002) show no significant correlation between lake area and number of OTUs detected (P 1⁄4 0.40 and P 1⁄4 0.20 respectively; linear correlations of log-transformed data) while the data from Reche et al. (2005) are almost significantly correlated (P 1⁄4 0.072). When these three data sets are merged, the correlation becomes significant (P , 0.001), as reported by Reche et al (2005). In two of the data sets (Lindström and Leskinen 2002, Reche et al. 2005), OTU richness was determined by counting the number of bands formed in denaturing gradient gel electrophoresis (DGGE). This method is commonly used in microbial ecology to obtain an image of microbial community structures (e.g., Forney et al. 2004, Loisel et al. 2006). Some of the limitations of DGGE are briefly discussed by Reche et al. (2005), for instance they acknowledge that DGGE in the best case only reflects the most dominant taxa. However, the consequence of bacterial community structures being skewed with few abundant and many rare taxa, a quite likely scenario (Acinas et al. 2004, Venter et al. 2004, Gans et al. 2005), is not addressed. If a low fraction of the present populations was detected, a change in the number of DGGE bands could reflect a change in rankabundance of populations (i.e., in the number of populations above the threshold of detection) rather than a change in richness (Forney et al. 2004). Thus, the number of DGGE bands may provide a biased estimate of richness, since it also depends on the evenness of a community. Furthermore, a recent study combining numerical simulations with laboratory experiments (Loisel et al. 2006) demonstrates that the number of bands or peaks obtained using DGGE and similar methods is not related to the richness of communities. Instead, the number of OTUs detected saturates around 35. Thus, the number of DGGE bands is a poor estimator of community richness. In the third data subset used by Reche et al. (2005), i.e., the data from Zwart et al. (2002), OTU richness is represented by the number of unique bacterial 16S rRNA sequences obtained from seven different studies screening clone libraries from nine lakes. Bacterial OTU numbers as determined by DGGE band numbers appears to saturate around 35 as reported by Loisel et al. (2006). In contrast, the number of unique sequences obtained from clone libraries from similar communities can be much greater. In the data set compiled by Zwart et al. (2002) the number of unique sequences per lake ranged from six to 125, the average being 60. Therefore clone library data are not comparable with DGGE data, and accordingly these two types of data should not be merged. Further, the available data (six lakes) from the original references in Zwart et al. (2002) show that the clone libraries screened range largely in size, at least from 45 to 350 clones per lake. Since the number of OTUs, or unique bacterial 16S rRNA sequences, from each lake was not corrected for sample size (i.e., number of clones picked) we suspect that the reported OTU richness reflects the effort spent by the respective researcher rather than community richness. Therefore, these data are not suitable for analysis of how lake surface area or other parameters are related to bacterial richness. Manuscript received 10April 2006; revised 20 January 2006; accepted 20 June 2006. Corresponding Editor: S. Findlay. 1 Limnology/Department of Ecology and Evolution, Evolutionary Biology Centre, Uppsala University, Sweden. 2 E-mail: Eva.Lindstrom@ebc.uu.se