< Back to Article
Spatial constraints and stochastic seeding subvert microbial arms race
Fig 2
Slow killing strains survive better in small spaces with equal starting abundance.
A. The fractional mean relative abundance of slow killing cells, 
, after 64 generations of simulation is plotted against the relative killing rate of the slow killer, κr, for different size simulations. Each line consists of 20 evenly spaced data points. Each data point is the mean of many simulations; for the Nmax = 9, 26, 68, 106, 1025, and 10,686 cell simulations we average across 750, 750, 500, 100, 50 and 20 simulations, respectively. For 106, 1025, and 10,686 cell simulations 
decreases rapidly with decreasing κr. Conversely, for Nmax = 9 cell and 26 cell simulations 
decreases linearly with κr. The standard error in 
from the simulations are shown along each trend. B. 
, the final proportion of the slow killing strain, is plotted against κr. Stochastic seeding means that the initial abundance of the slow killing strain was randomly selected from a binomial distribution with equal chance of either strain and the total cell count = 9 (
), thus modeling random attachment events from planktonic suspension. All simulations in A. are seeded in this manner. Deterministic seeding with an equal proportion is impossible for 9 total cells, so we averaged the trends where the initial abundance of the slow killing strain is deterministically set at 4 and 5 cells out of 9 (
). There are 20 data points in this trend line; each point is the average of 1000 simulations. These two lines describe simulated “small colony” sizes. The 1025 cell simulations are also stochastically seeded; the initial proportion is expected to be 0.5 +- 0.016. There are 20 data points in this trend line, and each is the average of 50 simulations. This line describes a simulated “large colony” (
). The error bars represent standard error. C. γ, the “Seeding Effect” is the difference between the stochastic and deterministic small colony seeding trends divided by the difference between the stochastic, small colony and the large colony trends. This quantifies how much of the “total finite size effect,” i.e., the difference between large and small simulations, is a result of stochastic seeding. Here, γ is plotted against κr, showing that if κr < 0.5, then γ > 0.5.
doi: https://doi.org/10.1371/journal.pcbi.1011807.g002