More is Different: Metabolic Modeling of Diverse Microbial Communities

Abstract

Microbial consortia drive essential processes, ranging from nitrogen fixation in soils to providing metabolic breakdown products to animal hosts. However, it is challenging to translate the composition of microbial consortia into their emergent functional capacities. Community-scale metabolic models hold the potential to simulate the outputs of complex microbial communities in a given environmental context, but there is currently no consensus for what the fitness function of an entire community should look like in the presence of ecological interactions and whether community-wide growth operates close to a maximum. Transitioning from single-taxon genome-scale metabolic models to multitaxon models implies a growth cone without a well-specified growth rate solution for individual taxa. Here, we argue that dynamic approaches naturally overcome these limitations, but they come at the cost of being computationally expensive. Furthermore, we show how two nondynamic, steady-state approaches approximate dynamic trajectories and pick ecologically relevant solutions from the community growth cone with improved computational scalability.

Keywords: flux balance analysis; metabolic modeling; metabolism; microbial communities; microbial ecology; systems biology.

FIG 1

Comparison between single taxon and community models. (A) In a single taxon genome-scale metabolic model, there is a unique optimal growth rate and an associated defined flux subspace. Fluxes that cross the model boundary represent exchanges with the external environment and are all relative to the biomass of the organism. The optimal growth rate solution for a single-taxon model can be represented as a point along a line and always has a unique maximum. (B) In a community model, individual taxon models are separate compartments that are embedded within a larger extracellular compartment. Transports across the taxon boundaries are scaled by that taxon’s relative dry weight to maintain the mass balance and to fulfill the originally imposed flux bounds. Feasible growth rate solutions in a multitaxon community model form a polytope, and there is a suboptimal subspace containing achievable growth rate distributions. The feasible flux space for these community-scale models is greatly inflated, compared to single-taxon models.

FIG 2

Approaches to selecting optimal solutions from the community growth cone. Growth rates in complex microbial communities are often dictated by a trajectory from the inoculation time to the point of maximum growth (red circle) located in a suboptimal growth region (red shaded area) (A). A dynamic model will simulate the trajectory explicitly (B), but one may also try to directly find a point in the growth cone that is empirically close to the true, unknown trajectory. SteadyCom uses a perfect diagonal as the implicit trajectory (C), whereas ctFBA finds the shortest path between the inoculation point and the maximum growth regime (red circle) (D). Untargeted approaches for the large-scale validation of community-scale metabolic model predictions (E). Emergent properties, such as the predicted versus observed taxon abundances that are estimated from amplicon or shotgun sequencing or the relationships between community diversity and growth rates, can be validated. Peak-to-trough ratios from deep metagenomic shotgun sequencing or isolate sequencing can be used to estimate replication rates (proxies for growth rates), which can be compared to model-predicted growth rates. Supernatants from in vitro culturing that are sampled at various time points can be used to measure community-wide consumption and production fluxes with untargeted metabolomics, which are then compared to model-predicted fluxes.