Challenges in microbial ecology: building predictive understanding of community function and dynamics

Abstract

The importance of microbial communities (MCs) cannot be overstated. MCs underpin the biogeochemical cycles of the earth's soil, oceans and the atmosphere, and perform ecosystem functions that impact plants, animals and humans. Yet our ability to predict and manage the function of these highly complex, dynamically changing communities is limited. Building predictive models that link MC composition to function is a key emerging challenge in microbial ecology. Here, we argue that addressing this challenge requires close coordination of experimental data collection and method development with mathematical model building. We discuss specific examples where model-experiment integration has already resulted in important insights into MC function and structure. We also highlight key research questions that still demand better integration of experiments and models. We argue that such integration is needed to achieve significant progress in our understanding of MC dynamics and function, and we make specific practical suggestions as to how this could be achieved.

Figure 1

Linking MC research questions with data and modelling. Research areas, plotted according to their complexity and temporal or spatial scale, form the link between data of different forms (magenta) and modelling formalisms (dark green). Pattern emergence (light green), that is, collective behaviour obtained by up-scaling from individual description to population level, can be predicted by modelling and tested experimentally. Abbreviations: DS, dynamical systems of deterministic, mechanistic nature, implemented as discrete or continuous time models (difference equations, ODEs); (d) FBA, (dynamic) flux balance analysis; SDS, stochastic dynamical systems, such as Markov chains, random walks, birth–death processes; IBM, individual-based models; PDE, partial differential equations, these are deterministic structured population models (for example, according to space or traits); diffusion processes, probabilistic counterparts of ODEs and PDEs.

Figure 2

(a) The human large intestine and the rumen and caecum in herbivorous animals harbour dense MCs dominated by anaerobic microbes that cross-feed metabolites extensively. In recent models, these are approximated by a small number of functional groups (Muñoz-Tamayo et al., 2010; Kettle et al., 2015). (b) Comparison of this model to a fermenter experiment with a pH shift from 5.5. to 6.5 after 9 days for metabolic products (dashed lines are experiment data, solid lines are model results). (c) Comparison of temporal species dynamics between model predictions and data. The experimental data consists of phylogenetic groups (16S rRNA gene sequencing), simulations refer to functional groups; approximate correspondence between the two is indicated by colour coding (for example, Lachnospiraceae equivalent to B5 and B8, shown in green, Bacteroides belonging to B1 shown in blue and purple).

Figure 3

MC analysis and predictive modelling in wastewater treatment and nutrient recycling facilities (WWTP). (a) Carbon and nutrients are eliminated from wastewater by (micro-)biological activities before discharging the cleaned water. WWTP processes are optimised through controlling MC conditions (for example, anaerobic, aerobic) and partial recycling of the MC biomass. (b) Overall system behaviour and MC biomass are predicted using growth kinetic and dynamic models. (c) Functional metabolic models are derived from pure culture physiology or metagenomic data. (d) Although the WWTP system is largely engineered, the MCs form mostly spontaneous, consisting of ‘core' assemblages and ‘passenger' groups as a function of constant wastewater input, after (Saunders et al., 2016). (e) The assemblage process depends on taxonomic relatedness and ecological interactions, and can lead to a drift of MCs over time. (f) MCs in WWTP occur mostly in flocs and or biofilms (picture courtesy of J van der Meer, University of Lausanne) that display immense microdiversity. Both, interactions and spatial organisation are factors that generate niches, gradients and foster co-existence.

Figure 4

Spatial patterns can emerge from local metabolic and mechanical interactions between individual microbes. Such self-organised patterns facilitate, for example, cooperation (a), competitive co-existence (b) or formation of fruiting bodies (c). (a) Cross-feeding between G (green) and R (red) strains facilitates exclusion of cheater C (blue) over time. Cross-feeding was engineered in yeast strains and visualised by fluorescence tagging, adapted from (Momeni et al., 2013). (b) Spiral patterns emerge from chasing in space: Colicin producer C kills sensitive S. S outcompetes resistant R, which outcompetes C in turn. Spatial structure facilitates dynamic co-existence. Experimental results on agar plates adapted from (Kerr et al., 2002) and simulation of spatial system from (Szczesny et al., 2014). (c) Experimental observation (Reichenbach et al., 1968) and simulation (Janulevicius et al., 2015) of circular aggregates formed by social motility of myxobacteria as an intermediate step in the development of fruiting bodies.

Figure 5

(a) (Left) Chemostat competition study between marine, nitrogen-fixing Cyanothece sp. and a non-nitrogen-fixing Synechococcus species (source: Department of Aquatic Microbiology, University of Amsterdam). (Middle) Schematic drawing of a chemostat. (Right) At high nitrate levels, the nitrogen-fixer (Cyanothece) is competitively excluded by the non-nitrogen-fixer (Synechococcus). Symbols are measurements; lines are model predictions (after Agawin et al., 2007). (b) (Left) Study in Winogradsky column microcosms. (Middle) Schematic of the vertically layered structure of a mature Winogradsky column. Principal microbial types are found in different layers, their ecological activities and the associated core chemical reactions are illustrated. As a result opposing gradients of sulphide and oxygen develop. (Right) Microbial activity leads to a transient drop in redox potential in the overlying water, and a long-term drop in the sediment, at high levels of added cellulose (‘high C'). Low levels of added cellulose (‘low C') induce only a short-term reduction in redox potential.