Fig. 2: Multivariate visualization and statistical differentiation of biofilm and planktonic bacterial communities.
Hierarchical clustering (Ward’s minimum variance method) organized all OTUs in the study according to distributions of standardized (z-scored) mean relative abundance across treatments. Throughout the 6-week experiment, biofilm communities were distinct from planktonic microbial communities while both clustered by time point (a). Biofilms displayed a successional trajectory not found in the planktonic community (b); data points are color coded by time point (2-weeks = blue, 4-weeks = gold, 6-weeks = red), and shaped by sample type and organism (open = planktonic, closed = biofilm; Triangle = algae, circles = coral, square = sand). Variance partitioning of three different PERMANOVA models (c) illustrate the relative influence of sample type, time point, benthic organism, and nutrient enrichment on bacterial taxonomic structure of planktonic and biofilm communities. Model 1 combines both planktonic and biofilm communities and emphasizes that sample type is the strongest driver of microbial community composition. Time point has the next largest influence on both planktonic (Model 2) and biofilm (Model 3) bacterial communities. Biofilm microbial communities are more strongly influenced by benthic organism (R2 = 0.123) than nutrient treatment (R2 = 0.063), while these parameters are equally influential in the planktonic community, explaining a much smaller variance (<1%). All tests and model terms shown are significant (p < 0.01).
