Noise in gene expression determines cell fate in Bacillus subtilis

Abstract

Random cell-to-cell variations in gene expression within an isogenic population can lead to transitions between alternative states of gene expression. Little is known about how these variations (noise) in natural systems affect such transitions. In Bacillus subtilis, noise in ComK, the protein that regulates competence for DNA uptake, is thought to cause cells to transition to the competent state in which genes encoding DNA uptake proteins are expressed. We demonstrate that noise in comK expression selects cells for competence and that experimental reduction of this noise decreases the number of competent cells. We also show that transitions are limited temporally by a reduction in comK transcription. These results illustrate how such stochastic transitions are regulated in a natural system and suggest that noise characteristics are subject to evolutionary forces.

Fig. 1

The regulation of competence in B. subtilis. (A) The comK regulatory network. Arrows and perpendiculars represent positive and negative regulation, respectively. For simplicity, factors shown in gray were not considered in our modeling. (B) The kinetics of growth in competence medium [in absorbance units (AU) measured in a Klett colorimeter]. (C) Competence development, determined microscopically with strains carrying a comK-cfp* fusion. The dashed lines in (B) and (C) represent T₀.

Fig. 2

Detection of single RNA molecules (comK and comK-M2) by FISH. (A) Schematic diagram depicting the endogenous comK (left), comK-cfp* (middle), and comK-M2 (right) reporters, all controlled by the comK promoter (PcomK). Multiple specific fluorescent probes bind to each mRNA molecule [comK (green) or comK-M2 (red)], yielding distinct fluorescent signals. The comK-cfp* construct identifies competent cells. comK-cfp* designates an in-frame fusion of CFP to comK, and M2 designates the RNA with 32 repeat sequences. (B) Differential interference contrast (DIC) images and (C) pseudo-colored fluorescence images taken at T₀ for the WT strain, in which the comK mRNA was hybridized to six FISH probes [C6–tetramethyl rhodamine (C6-TMR)] that bind to the comK open reading frame. Dots correspond to individual mRNA molecules. Scale bars, 4 μm. (D and E) Kinetics of the population means of mRNA molecules per noncompetent cell before and after T₀ for the WT (blue circles, BD4379), the rok (red circles, BD4380), and the comK (purple circles, BD4382) strains. The WT and the rok strains (D) were hybridized to C6-TMR to detect comK mRNA molecules. The comK strain (E) was hybridized to a probe (PM2-Alexa 594) that binds to the M2 probe-binding sequence. Error bars were obtained by bootstrapping.

Fig. 3

Noise in comK transcription is mainly intrinsic. (A) Intrinsic and extrinsic noise were measured by detecting the mRNA from two coexpressed genes (comK and comK-M2) controlled by the comK promoter (26). Uncorrelated gene expression in individual cells is indicative of intrinsic noise. (B) DIC images and (C) pseudo-colored merged fluorescence images taken at T₀ showing hybridization to comK-M2 (red) and comK (green) mRNA with the PM2-Alexa 594 and C6-TMR probes, respectively. Scale bars, 4 μm. (D) Distribution of comK and comK-M2 mRNA molecules for the WT strain (BD4379) at T₀, showing weak correlations between production of the two mRNA molecules (r = 0.15). (E) Correlation coefficients throughout growth for the same strain. Error bars indicate SE.

Fig. 4

Noise reduction in comK expression lowers the percentage of competent cells. The leftmost column depicts comK mRNA distributions predicted by the model for the WT (A), rok (B), and low-noise (C) strains at T₀. The middle column shows ComK protein distributions at T₀ assuming a high rate of translation in the WT and rok strains [(A) and (B)] and a lowered rate of translation in the low-noise strain (C). The vertical red lines show the predicted threshold beyond which the positive autoregulatory loop of comK would be activated, resulting in competence [the threshold in the rok strain changes because of increased gene expression (see SOM)]. The rightmost column shows CFP fluorescence images from the three strains, taken at T₂ and overlaid on DIC images. All three strains expressed the comK-cfp* fusion, thus fluorescing when competent. The lowest panel in the column shows a microscopic field for the low-noise strain selected to show one competent cell, although the frequency of such cells was less than 1%. Scale bars, 4 μm.