Quantifying the bias introduced by the use of different targeted regions of the 16S rRNA gene for the characterization of microbial communities using amplicon-sequencing

FEDERICO M. IBARBALZ; MARIA VICTORIA PÉREZ; EVA L. M. FIGUEROLA; LEONARDO ERIJMAN

Rosario

Congreso; IV Argentinean Conference on Computational Biology and Bioinformatics & IV Conference of the Iberoamerican Society for Bioinformatics; 2013

BackgroundParallel high-throughput sequencing has transformed the field of microbial ecology by allowing the extensive characterization of microbial community structures of highly diverse ecosystems. Amplicon sequencing of the 16S ribosomal RNA gene (16S rRNA) is one of the most widely used approaches. Prior to parallel sequencing, a particular set of "universal" primers is used to target one or more of the nine hypervariable regions (V1-V9) present in the 16S rRNA gene. No matter the sequencing depth, the results are inevitably skewed by the bias introduced by the unequal amplification of each member of the community, which in turn depends on the selected set of primers [1; 2]. We have recently showed that bacterial community structure in activated sludge, used for wastewater treatment, depends primarily on the type of wastewater, even when comparing sets of data obtained using different 16S rRNA regions. Nevertheless, patterns of bacterial distribution were noticeable affected by the bias introduced by the targeted 16S rRNA region [3].The aim of this work was to assess the degree of dissimilarity among communities that would overcome the bias introduced by the use of primer sets targeting different 16S rRNA regions. We hypothesized that a transition from 16S region-based clustering to composition-based clustering would occur within a gradient of communities with gradually increasing dissimilarity.To test this hypothesis, we performed 454 pyrosequencing of a series of twelve monthly samples from activated sludge taken from a sewage treatment plant (STP) located in the northern suburbs of Buenos Aires, Argentina. Additionally, two activated sludge samples from a different STP located in the southern suburbs of Buenos Aires were also included in the analysis. DNA amplification was performed using two pairs of universal primers, targeting the V1-V3 regions (E. coli 8-534 position) and the V4 region (E. coli 563-924 position). Due to technical replicates, every sludge sample was analyzed four times, two at each region. Raw reads were filtered, aligned and taxonomically classified using Mothur v.1.22.2. Distance calculations and statistical analysis were performed using the Fast UniFrac online tool.ResultsProteobacteria was the dominant phylum detected by both set of primers. Reads from V1-V3 region highlighted the presence of Actinobacteria and Bacteroidetes, while Firmicutes and Acidobacteria had higher abundances in reads from the V4 region. UniFrac and Bray-Curtis distance matrices showed significant correlation between regions (Mantel test), even though, on average, V4 distances were lower than those from V1-V3. The alpha diversity metrics determined with both regions were similar.The percentage of change (C%=100-Similarity%) between samples within the time series were, on average, 43.5% for V1-V3 and 34.9% for V4. C% between different STPs was 73% and 60% for V1-V3 and V4, respectively. Despite these large differences in bacterial composition, when both data sets were analyzed together, samples clustered into two separate groups according to the 16S rRNA region.ConclusionsWe showed that V1-V3 and the V4 region of the 16S rRNA can reach similar estimation of alpha diversity, but fail in providing an appropriate representation of beta diversity due to the biases introduced by the unequal amplification of the different community members. These results highlight the need to combine different strategies for the study of complex microbial ecosystems.