Fig 1.
Conceptual image of IndiMeSH model components and methods.
a) IndiMeSH uses two-dimensional angular pore networks representing soil cross-sections (here a soil aggregate) as a backbone for the aqueous phase configuration, chemical diffusion and as a habitat for bacteria represented as individual agents. Angular pores enable dual occupancy of both the water and gas phase within individual pores. Growth and nutrient consumption is calculated using metabolic networks, which enables the triggering of fundamentally different growth strategies depending on local nutrient and redox conditions. b) For the same pore network, aqueous phase configuration governed by the matric potential (wet and dry conditions) determines the connectivity and thus bacterial dispersal opportunities as well as nutrient diffusion characteristics by limiting water films held within the angular pores.
Table 1.
Summary of genome-scale and reduced-scale metabolic network dimensions for all species.
For all species, the number of reactions and metabolites were reduced by an order of magnitude using the redGEM algorithm [36]. This reduced the computational burden significantly.
Table 2.
Evaluation of metabolic network reduction and comparability to genome-scale metabolic networks.
Genome-scale and reduced-scale metabolic networks were compared using flux variability analysis whilst exposed to different environmental conditions. A percentage of the maximum uptake rate (20%, 60% and 100%) for each nutrient was cross-combined with all other nutrient levels in order to generate 81 conditions in the case of four nutrients and 27 conditions in the case of three nutrients. Presented is the median percentage of all reactions having a minimal flux or maximum flux difference falling within the respective category (with standard deviation in brackets).
Fig 2.
Relative abundance of E. coli and S. enterica in the IndiMeSH simulations and original COMETS simulations and experimental results.
Combined relative abundance of initial inoculation ratios (99:1 and 1:99 of E. coli S. enterica, three replicates each, inoculated at 100 random nodes) for experimental results, COMETS simulations and IndiMeSH simulations after 48h of growth. Error bars represent one standard deviation.
Fig 3.
Distribution of obligate aerobic P. putida (cyan) and facultative anaerobic P. veronii (magenta) along the carbon-oxygen counter gradient from the central port to the peripheral ports.
Combined experimental results (boxplots, whiskers indicate minimum and maximum of data) and simulation results (shaded area with 95% of all cells, thick line represents mean) in a well connected, artificial pore network using IndiMeSH (left) and Monod parametrization (right). Segregation of facultative anaerobes into an anaerobically growing (central port, anoxic) and peripheral, aerobically growing subpopulation can be observed in the results using IndiMeSH which is absent in the results using Monod parametrization. The observed shift of obligate aerobes towards the periphery is due to the chosen canonical half saturation coefficient in IndiMeSH (0.05 mM) compared to the Monod half saturation coefficient in the original model (0.0063 mM).
Fig 4.
Response of bacterial population to a shift in boundary conditions represented as glucose perfusion into a soil aggregate cross-section.
(a) The pore network used for the simulation congruent to the soil aggregates used in the experiment. The distribution of glucose within the pore network is visualized prior to glucose perfusion. (b) Relative diffusion compared to a saturated pore network is calculated using the Millington-Quirk equation. The dashed orange line indicates the state of the pore network at the matric potential set experimentally and within the simulations, limiting nutrient diffusion in the aqueous phase whilst facilitating gaseous diffusion due to unsaturated pores. (c) Growth curve of the bacterial population showing a secondary growth phase after glucose addition with a decline in population size approaching the previously established stable population size due to the system carrying capacity.
Fig 5.
Strategic analysis of pore size distribution and hydration conditions.
Visualization of the degree of connectivity for all three pore networks with varying mean pore size at two hydration conditions. (a) Histogram of pore sizes taken from lognormal distributions with varying mean and standard deviations. (b) Resulting soil water characteristic curve including indications of matric potential taken as wet (-1 kPa) and dry (-10 kPa) conditions. (c) Calculated relative diffusion for all pore networks using the Milling-Quirk equation. (d) Bacterial growth dynamics depending on pore network and hydration condition for the whole simulation time of 120 h. (e) Bacterial consumption for the small pore size network and varying hydration conditions. (f) Spatial distribution of acetate production and consumption. Acetate is produced at the carbon rich center due to overflow metabolism (electron acceptor limited) and consumed at the periphery via aerobic respiration.