2. Functional Stability and Community Structure Relationships in Bioreactor Communities.
Bioreactor systems were found useful in elucidating certain characteristics of microbial communities that would have been too complex to study in open or uncontrolled systems. In addition, these systems were also useful in demonstrating the need to use complementary molecular and analytical techniques targeting both community structure and function. In our earlier report, it was noted that one set of bioreactor communities was dominated by spirochetes. Since they generally constitute less than 5% of anaerobic digester communities, their dominance in our system was intriguing. Further studies designed to find the reasons for this dominance, showed that several other reactors inoculated from three different sources namely rumen, sediment, and anaerobic sludge and operated for six months under identical conditions were also dominated by spirochetes. Since all operational conditions were the same except the inoculum source, we suspect that the most likely reason for this enrichment of spirochetes is the low glucose loading rate (0.4 g/liter-day). Spirochetes are slow growing organisms and compete well under low substrate concentrations.
A. Cause and Effects of Shifts in Bioreactor Community Composition. In situ 14C studies of the reactors indicated that spirochete dominance resulted in a modified food web that converted most of the glucose to acetate and ethanol as opposed to acetate and propionate that was observed in other reactor communities dominated by streptococci. This observation was confirmed by studies conducted with RM-8, a spirochete isolate from these reactors, and by substrate utilization assays on reactor suspended solids. The above observations indicate that fermentative populations may have a significant impact on the carbon flow in methanogenic systems. In addition, it was also noted that the nature of the metabolic food-web that emerges out of the association of multiple populations is one of the factors that controls the functional stability of complex microbial systems. Among the many anaerobic bioreactor communities studied, systems that simultaneously funneled the substrate pulse through multiple intermediates were functionally more stable than systems that processed it in a sequential manner. This conclusion has implications for designing mixed community reactors that are functionally more stable.
Community characterization techniques including amplified ribosomal-DNA restriction analysis (ARDRA), terminal restriction fragment length polymorphisms (T-RFLP), and image analysis of phase contrast microscopy demonstrated that minor populations (both numerically and by volume) may play a significant role in response to perturbations. In one set of reactors dominated by spirochetes, a numerically minor member of the community that was undetected prior to a glucose perturbation, became dominant in response to the perturbation. This behavior adds to the notion that complex microbial systems sustain populations that may benefit from a change in the environmental conditions or substrate availability.
B. Linking Community Structure with Function. It is increasingly recognized that rDNA-based community characterization techniques (e.g., ARDRA, T-RFLP analysis) perform poorly when it comes to linking the structure to its function. Since these techniques are very powerful in detecting a full spectrum of the community without isolation, tools that can link community fingerprints to the function will complement them. Community structure and function data obtained from these bioreactors offered us the opportunity to test the usefulness of artificial neural networks (ANN, Almeida et al. 1998), for this purpose. ANN is a mathematical tool that can be used to relate complex data sets when the underlying relationships are not known. We found that ANN may be a useful in predicting structure-function relationships of the same community if large data sets are used for training the network. Using ANN, we were also able to identify important T-RFLP ribotypes that correlated to the accumulation of various intermediate products under perturbed conditions. Figure 1 (below) depicts one of the structures of ANN that was used for this purpose. With a limited number of training examples, its prediction capability was found questionable which is among the known disadvantages of ANN. The application of this technique to bioreactor communities is novel and we are still exploring its usefulness for various molecular data being collected from other studies.
|Figure 1. Example of an artificial neural network being used to relate
community structure with function in bioreactor communities.
Return to Community Analysis Thrust Group