Long-term history dependence of growth rates of E. coli after nutrient shifts.

According to a widely accepted paradigm of microbiology, steady-state growth rates are determined solely by current growth conditions1-3 and adaptations between growth states are rapid, as recently recapitulated by simple resource allocation models4. However, even in microbes overlapping regulatory networks can yield multi-stability or long-term cellular memory. Species like Listeria monocytogenes5 and Bacillus subtilis "distinguish" distinct histories for the commitment to sporulation6, but it is unclear if these states can persist over many generations. Remarkably, studying carbon co-utilization of Escherichia coli, we found that growth rates on combinations of carbon sources can depend critically on the previous growth condition. Growing in identical conditions, we observed differences in growth rates of up to 25% and we did not observe convergence of growth rates over 15 generations. We observed this phenomenon occurs across combinations of different phosphotransferase (PTS) substrates with various gluconeogenic carbon sources and found it to depend on the transcription factor Mlc.

Computational analysis of GAL pathway pinpoints mechanisms underlying natural variation.

Quantitative traits are measurable phenotypes that show continuous variation over a wide phenotypic range. Enormous effort has recently been put into determining the genetic influences on a variety of quantitative traits with mixed success. We identified a quantitative trait in a tractable model system, the GAL pathway in yeast, which controls the uptake and metabolism of the sugar galactose. GAL pathway activation depends both on galactose concentration and on the concentrations of competing, preferred sugars such as glucose. Natural yeast isolates show substantial variation in the behavior of the pathway. All studied yeast strains exhibit bimodal responses relative to external galactose concentration, i.e. a set of galactose concentrations existed at which both GAL-induced and GAL-repressed subpopulations were observed. However, these concentrations differed in different strains. We built a mechanistic model of the GAL pathway and identified parameters that are plausible candidates for capturing the phenotypic features of a set of strains including standard lab strains, natural variants, and mutants. In silico perturbation of these parameters identified variation in the intracellular galactose sensor, Gal3p, the negative feedback node within the GAL regulatory network, Gal80p, and the hexose transporters, HXT, as the main sources of the bimodal range variation. We were able to switch the phenotype of individual yeast strains in silico by tuning parameters related to these three elements. Determining the basis for these behavioral differences may give insight into how the GAL pathway processes information, and into the evolution of nutrient metabolism preferences in different strains. More generally, our method of identifying the key parameters that explain phenotypic variation in this system should be generally applicable to other quantitative traits.

Variation in the modality of a yeast signaling pathway is mediated by a single regulator.

Bimodal gene expression by genetically identical cells is a pervasive feature of signaling networks and has been suggested to allow organisms to hedge their 'bets' in uncertain conditions. In the galactose-utilization (GAL) pathway of Saccharomyces cerevisiae, gene induction is unimodal or bimodal depending on natural genetic variation and pre-induction conditions. Here, we find that this variation in modality arises from regulation of two features of the pathway response: the fraction of cells that show induction and their level of expression. GAL3, the galactose sensor, controls the fraction of induced cells, and titrating its expression is sufficient to control modality; moreover, all the observed differences in modality between different pre-induction conditions and among natural isolates can be explained by changes in GAL3's regulation and activity. The ability to switch modality by tuning the activity of a single protein may allow rapid adaptation of bet hedging to maximize fitness in complex environments.

Publisher Correction: Extensive signal integration by the phytohormone protein network.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Extensive signal integration by the phytohormone protein network.

Plant hormones coordinate responses to environmental cues with developmental programs1, and are fundamental for stress resilience and agronomic yield2. The core signalling pathways underlying the effects of phytohormones have been elucidated by genetic screens and hypothesis-driven approaches, and extended by interactome studies of select pathways3. However, fundamental questions remain about how information from different pathways is integrated. Genetically, most phenotypes seem to be regulated by several hormones, but transcriptional profiling suggests that hormones trigger largely exclusive transcriptional programs4. We hypothesized that protein-protein interactions have an important role in phytohormone signal integration. Here, we experimentally generated a systems-level map of the Arabidopsis phytohormone signalling network, consisting of more than 2,000 binary protein-protein interactions. In the highly interconnected network, we identify pathway communities and hundreds of previously unknown pathway contacts that represent potential points of crosstalk. Functional validation of candidates in seven hormone pathways reveals new functions for 74% of tested proteins in 84% of candidate interactions, and indicates that a large majority of signalling proteins function pleiotropically in several pathways. Moreover, we identify several hundred largely small-molecule-dependent interactions of hormone receptors. Comparison with previous reports suggests that noncanonical and nontranscription-mediated receptor signalling is more common than hitherto appreciated.

Polymorphisms in the yeast galactose sensor underlie a natural continuum of nutrient-decision phenotypes.

In nature, microbes often need to "decide" which of several available nutrients to utilize, a choice that depends on a cell's inherent preference and external nutrient levels. While natural environments can have mixtures of different nutrients, phenotypic variation in microbes' decisions of which nutrient to utilize is poorly studied. Here, we quantified differences in the concentration of glucose and galactose required to induce galactose-responsive (GAL) genes across 36 wild S. cerevisiae strains. Using bulk segregant analysis, we found that a locus containing the galactose sensor GAL3 was associated with differences in GAL signaling in eight different crosses. Using allele replacements, we confirmed that GAL3 is the major driver of GAL induction variation, and that GAL3 allelic variation alone can explain as much as 90% of the variation in GAL induction in a cross. The GAL3 variants we found modulate the diauxic lag, a selectable trait. These results suggest that ecological constraints on the galactose pathway may have led to variation in a single protein, allowing cells to quantitatively tune their response to nutrient changes in the environment.