Metabolic Chaos during Microbial Stress Responses

Bacteria mount a physiologic stress response to survive hostile or stringent conditions. The genes that comprise the stress response can be defined in two ways: a transcriptional network composed of genes with expression changes during stress, and a phenotypic network of genes whose deletion lowers the bacteria's fitness during stress. On first thought, one would expect these two networks to be the same, or at least have significant overlap. The genes important for surviving stress should be up-regulated, and genes not essential for tolerating the stress should be down-regulated. However, studies from multiple microbes and conditions have discovered that the transcriptional and phenotypic stress responses are nearly disjoint; genes with expression changes during stress can be deleted without changing fitness, and phenotypically important genes are rarely differentially expressed. Although several studies have described this striking phenomenon, it remains unclear why bacteria appear to have two separate stress response networks.

We discovered that fitness and expression changes can be linked using COBRA models. During nutrient depletion, genes with expression changes are often adjacent to phenotypically important genes. Even the magnitude of the fitness or expression changes are linked — the largest fitness changes are closest to the largest expression changes, and vice versa. The expression changes are much closer to the fitness changes than expected if both sets of genes were randomly distributed throughout the network. By contrast, the response to antibiotic stress is largely uncoordinated. Changes in fitness and expression appear randomly distributed across the network, and the magnitudes of the nearest fitness and expression changes are uncorrelated. Using a novel “enzyme coupling” framework, we show how modularity and redundancy in the metabolic network enable the split between transcriptional and phenotypic stress responses – in both the coherent starvation response and the incoherent antibiotic response.